By Elham Saeidinezhad

Last Thursday, more than $6 billion was stolen from Binance, the world’s largest crypto exchange. The first wave of the attack drained its wallets on ten blockchains in about half an hour. Binance answered with a choice: it would absorb the loss itself. That was the right decision. It is also the problem. A customer protected by a choice is protected only as long as the firm can afford to make it, and wants to.

I read Binance’s customer agreements to see what it was required to pay. They protect what customers own—the assets are held for them, kept apart from the exchange’s own and out of reach of its creditors if it fails. But ownership is a claim to assets that are still there. It says little about who replaces assets that are gone, and on that question the written terms leave most of the answer to the exchange.

That matters because the incentive to pay runs the wrong way. A healthy exchange pays to protect a valuable business, as Binance did. A weak one has less to protect and less to pay with, so the promise is weakest exactly when customers need it most. It is insurance written by the house it insures. The next exchange to be robbed may be smaller, weaker or less willing, and its customers could discover that the protection they counted on was never a contractual right.

The fix is not new. Lenders, bondholders and fund investors stopped relying on the other side’s goodwill long ago. Loan agreements, securitisations, fund rules and clearing houses settle five questions before anything goes wrong: who pays, from what money, which terms cannot change without consent, when holders must be told, and who takes over if the operator fails. This essay sets out what each of those answers would look like for a crypto exchange, and why each has to be priced as an option or insurance policy before the next theft.

Who actually paid

Binance puts the loss at about $6.3 billion; the blockchain-analytics firm Elliptic counts $6.43 billion. Binance said it would cover it from three sources, all of them its own:

• part of SAFU, its user protection fund;

• part of its futures insurance fund, the amount above what that fund must keep;

• its corporate treasury.

In its October 31 update, it said it had set aside more than $6.34 billion and replenished 90% of the affected assets. By November 1, it put the figure at about 95%. The rest is in less liquid tokens that take longer to buy back.

The names make this sound like insurance. It is not. SAFU is a reserve Binance created and funds, and it is held by Nest Clearing and Custody, one of Binance’s own Abu Dhabi entities. Binance has committed to restore it to about $1 billion, roughly a sixth of this loss, so on its own it could never have covered the theft. The futures insurance fund exists to cover losses from futures trading, not theft. No outside insurer has been reported as paying any part of the loss.

Part of the money may yet come back. Elliptic reports that about $2.85 billion of the stolen funds has been frozen or seized, mostly by stablecoin issuers. That is a protection too, but it is one held by third parties, and nothing in the customer’s contract provides for it.

So Binance’s customers were made whole with Binance’s own money. The question at the center of this essay is whether they had a right to it. When they placed their assets on the exchange, had they secured a right to be made whole after a theft, or were they relying on Binance to decide, when the time came, that paying was worth it?

What the contract says

I read two of Binance’s customer agreements: the terms for its Dubai entity and those for its Abu Dhabi entities.[^1] I wanted to know one thing. When assets held on the exchange are stolen, what is the customer entitled to?

The agreements get the first question right. Under the Dubai terms, customer assets remain the customer’s property; under the Abu Dhabi terms, they are held on trust for the customer. They are recorded separately from the exchange’s own assets, and the exchange may not lend or pledge them without the customer’s agreement. If the exchange failed, they would sit outside its estate. That is a real protection. A creditor of a failed firm waits in line and may recover only part of what it is owed. An owner takes its property back.

But ownership protects customers against the exchange’s creditors. It does not protect them against a thief or cyber security threats. Customer assets sit together in shared wallets, and under the Abu Dhabi terms each customer owns a proportion of the pool. If the pool is robbed and no one refills it, every customer’s share is worth less. Ownership tells customers what share is theirs. It does not tell them who refills the pool.

On that question, the agreements say much less. On my reading, they:

• make the exchange liable only for gross negligence, wilful misconduct, fraud or a material and continuing breach;

• exclude losses caused by the actions of third parties, and list “security breaches” among the exclusions;

• treat the failure of a sub-custodian as force majeure;

• limit the customer to damages, waiving “any claim in rem, injunction, and/or specific performance”;

• in the Dubai terms, cap liability at the fees the customer paid in the previous twelve months. (Binance says its Dubai clients’ wallets were not affected.)

Then there is SAFU. The Dubai terms say the fund covers certain losses, “or as may otherwise be determined by Binance.” The Abu Dhabi terms do not mention it at all. Binance’s own SAFU explanation says Binance decides which losses and which claims are eligible. A fund that pays when its owner decides is not a right to be paid.

Read together, the terms protect what customers own and leave the question of who replaces it largely to Binance. On my reading, Binance has committed to pay more than these agreements clearly require. In other words, customers were made whole by a choice, not a right.

Why a firm would choose to pay

But why would Binance absorb such a large loss? The answer begins with what an exchange sells. Customers come to a trading venue not only to store assets, but also because other customers are there: buyers need sellers, sellers need buyers, and a deep market makes it easier for both to trade. The assets customers hold on the platform and the trading activity they generate are central to the value of the business.

That value depends on trust. If Binance had passed the theft on to customers, the concern would not necessarily have stayed with those directly affected. Other users might have asked whether their assets could be next and decided to withdraw. As assets and traders left, trading activity could fall, making the platform less useful to the customers who remained. A decision to impose losses could therefore damage the business well beyond the amount stolen.

Seen this way, paying for the loss was not simply a cost. It was also a way to protect Binance’s future business. The company could spend money today to preserve customer confidence, trading activity and the value those relationships generate over time. Customers benefited from that decision, but the decision also served the firm.

Economists Benjamin Klein and Keith Leffler described the logic behind this kind of behavior in their work on reputation. A firm may have a reason to honor a commitment even when enforcement is uncertain: if customers stop trusting it, the future business it loses can be worth more than the cost of keeping the commitment. In their model, the incentive depends on the value of that future business. A second condition matters here too: the firm must be able to bear the immediate cost. Klein and Leffler, 1981.

Bitfinex offers a useful comparison. After nearly 120,000 bitcoin were stolen in 2016, the exchange initially spread the loss across customer accounts, reducing balances by approximately 36% and issuing tokens to record what customers had lost. Within eight months, those tokens had been redeemed at face value or exchanged for equity. The contrast is not that customers necessarily bore the final loss at Bitfinex, but that they had to bear it first—and wait for recovery. Bitfinex’s initial update; Bitfinex’s account of the later redemption.

The comparison illustrates why the firm’s financial condition matters. A firm with sufficient resources and a valuable business to protect may decide that absorbing a loss is the best way to preserve its future. A firm with fewer resources may be unable to do so immediately, even if it also wants to retain its customers.

Binance’s commitment protected customers this time. The question is how much protection customers can count on when the firm making that commitment is no longer healthy.

The incentive that inverts

Now imagine the same theft at a weaker exchange. It has lost assets, customers are deciding whether to withdraw, and the firm has limited resources with which to respond. It may still understand that protecting customers would preserve trust. The problem is that preserving trust takes money—and money paid to customers is no longer available to meet the firm’s other obligations.

Those obligations might include debts to lenders, payments owed to trading counterparties or other claims against the business. As the firm’s resources shrink, each group has a stronger reason to pay attention to where the remaining money goes. A payment that protects customers may preserve some value in the business, but it also reduces the assets immediately available to other claimants. The firm’s management must weigh those competing demands while its room to act is narrowing.

This is what makes the protection fragile. When the exchange is healthy, it can look at the cost of a theft against the value of a business it expects to operate for years. Absorbing the loss may be the best way to preserve that future. When the exchange is close to failure, that future is less certain and the resources available to protect it are smaller. Even if management wants to make customers whole, the firm may not be able to do so.

The result is an uncomfortable inversion. The incentive to go beyond the written terms can be strongest when the firm is strong and has a valuable business to protect. It can weaken as the firm deteriorates, just as customers become more dependent on it.

In some insolvencies, even a payment made before failure can later be challenged. Under US bankruptcy law, for example, an estate may seek to recover certain payments made to creditors in the 90 days before a filing, subject to specific conditions and defenses. In the Celsius bankruptcy, the litigation administrator pursued claims concerning certain customer withdrawals made before the filing. That does not mean every customer payment can be recovered, or that the same rules would apply to Binance. It shows that, in distress, receiving a payment and being able to keep it can become separate questions. US Bankruptcy Code, section 547; Celsius court opinion.

None of this means a weak firm will inevitably refuse to help its customers. It means the customer’s protection depends on the firm having both the resources and the incentive to provide it at the moment it is needed.

An incentive is not a right. If it fades, customers must look to the rights they can enforce under the applicable contract and law. In other words, the incentive that saved customers last week behaves in an awkward way. It is strongest when the firm is strong, which is when it is least needed. It is weakest when the firm is weak, which is exactly when customers need it most.

It is insurance written by the house it insures.

The terms can change before the crisis

There is another reason to look closely at the contract: the terms a customer accepts today may not be the terms that apply later. Both Binance agreements I read allow the exchange to make changes. The process differs between them, but under the Abu Dhabi terms, continued use of the service can count as acceptance. The agreement says Binance will generally try to give notice before amendments take effect, although some changes can take effect immediately. Binance Global Terms of Use, section 23.

The practical question is not simply whether a firm can update its terms. Businesses need to make changes as their services and regulatory obligations evolve. The question is what happens when a change affects a protection the customer relied on when placing assets with the firm. Does the customer receive enough notice to understand it? Can the customer withdraw before it takes effect? And what happens if withdrawals are already restricted?

Celsius shows why those questions matter. When the company entered bankruptcy in 2022, its terms of use had been revised several times. The bankruptcy court examined whether those terms and their amendments formed an enforceable agreement, and what they said about ownership. In its 2023 decision, the court concluded that, under the applicable terms, assets deposited in Celsius’s Earn program became Celsius’s property. The assets remaining in those accounts became part of the bankruptcy estate, subject to the defenses the court reserved for individual customers. The decision did not determine ownership of assets in Celsius’s separate Custody program. Celsius bankruptcy court decision.

That said, one needs to be careful about the comparison. Binance’s terms say that client assets remain the customer’s property or are held on trust for the customer; the Celsius Earn terms, on the other hand, transferred title to the company. They are different provisions, governing different arrangements. The lesson is that when a firm fails, the wording of its customer agreement—and whether customers actually agreed to changes in those terms—can make all the difference.

Other areas of finance offer a way to address this problem. Under the EU’s Payment Services Directive, for example, providers generally must propose changes to a payment-services framework contract at least two months before they take effect. Where the contract provides for acceptance by silence, the customer must be told of the right to terminate without charge before the change applies. The rule has exceptions, but its logic is straightforward— notice matters because it gives the customer a right to decide whether to stay. Payment Services Directive, Article 54.

A similar question about customers’ right belongs in crypto custody agreements. If a venue changes a term governing ownership, custody or liability, customers need more than a notice that the document has been updated. They need a meaningful opportunity to understand the change and move their assets before accepting it.

The protection that matters is the one a customer can enforce when things go wrong. That depends not only on what the agreement says today, but also on how it can change—and on the customer’s ability to leave before it does.

Half an hour against a calendar

There is a second problem, and it is about time. The first wave of the Binance attack took about half an hour. In that time, Binance wallets on ten blockchains were drained. But the mechanisms customers rely on to learn about a problem—and decide what to do—operate on a different clock.

In the Binance agreements I reviewed, I did not find a fixed deadline requiring the exchange to notify customers of an incident. The terms also give the exchange the ability to suspend withdrawals. Proof of reserves can provide a picture of assets at a particular moment, but a picture is not the same as a continuous warning system. It tells a customer something about the position when the snapshot was taken; it does not, by itself, tell them what happens next if a shortfall appears. The SEC has also cautioned that proof-of-reserves reports are not financial-statement audits. SEC investor bulletin.

The timing problem is not unique to crypto. I also read a first-lien leveraged-loan agreement from the US market. It had no cyber clause. Its notice provisions relied on whether a matter could have a “material adverse effect,” while financial covenants were tested at quarter-end.

Consider how that works after a ransomware attack. The borrower may lose revenue while systems are down. It may face costs to restore them, pay specialists and deal with interrupted operations. Cash is affected as those costs arise, but a financial test conducted at the end of a quarter may register the deterioration later. Meanwhile, notice to lenders can depend on management’s assessment of whether the incident meets the standard in the agreement. The economic damage and the contractual response do not necessarily match. In effect, customers bear much of the risk during an attack, while management retains broad discretion over how to respond. That discretion has economic value: customers grant the exchange flexibility while accepting fewer rights themselves. Yet the value of those rights—and the cost to customers of giving them up—remains largely unexamined.

Public-market disclosure has a similar gap, although it also places a constraint on management. Under the SEC’s cyber disclosure rule, a company generally has four business days to report a material incident after determining that it is material. It must make that determination without unreasonable delay after discovering the incident. The four-day filing clock therefore does not begin at discovery, but the company cannot simply postpone its assessment indefinitely. SEC cybersecurity disclosure rule.

Europe’s data-protection rules illustrate another way to start the clock. When a qualifying personal-data breach occurs, the GDPR generally requires the organization to notify its supervisory authority within 72 hours of becoming aware of it. That is not a rule requiring customers to be told about every theft: it concerns personal-data breaches, has exceptions, and the initial notification is to the regulator. But it shows that a reporting clock can be tied to awareness of an event rather than waiting for a company to complete its assessment of materiality. GDPR, Article 33.

The distinction is consequential. A regulator may need information to oversee the firm. A customer needs information to decide whether to leave assets in its custody. Those are different decisions, and a deadline designed for one does not necessarily give the other person a timely warning.

The question, then, is not just whether an incident must eventually be disclosed. It is who starts the clock, who receives the notice, and what the holder can still do when the notice arrives.

Ownership is only half the protection

Put the pieces together and a pattern appears. The contracts protect what customers own. They give customers almost no say in what happens when the firm, or the safety of their assets, starts to change.

Lenders in leveraged loan market learned long ago that this is not enough. They protect themselves in two ways. One is property: security over assets, and limits on competing claims. The other is control: information, limits on what the borrower may do, and a say in what happens next. Property protects lenders once the borrower has failed. Control protects them while there is still something to save.

J.Crew showed why both are needed. In 2016 it moved valuable intellectual property into subsidiaries outside its lenders’ covenants. It then used that property to back new secured bonds that ranked ahead of those lenders. The lenders’ collateral was real; it was simply moved. Loan agreements now often carry a “J.Crew blocker” to stop it happening again.

Binance’s terms cover the first part, property— customer assets belong to the customer, or are held on trust, and are kept separate from the exchange’s own. Control is another matter that is totally missing in Binance’s agreements with its clients. The one trigger the terms clearly provide, the power to suspend withdrawals, belongs to the exchange. That is exactly what happened after the heist. After the attack, Binance paused withdrawals even for its Dubai customers, “as a precautionary measure,” although it said their wallets were not affected. A pause can be the right call after an attack. But the clause that stops assets leaving during an attack also stops customers leaving the platform, and only the exchange decides when it starts and ends. Customers have no trigger of their own. In effect, customers have written an option in the exchange’s favour—without receiving an explicit premium for the risk they bear and the insurance they wrote for the exchange.

Tokenized real-world assets face the same split. Tokenization alone does not decide what a holder can do. Where the terms are negotiated, both questions can be answered from the start: what does the holder own, and what can the holder do if the arrangement comes under stress?

Before the next heist: five protections crypto can borrow from traditional finance—and price into Its contracts

After a theft, customers need answers to simple questions. Who must replace their assets? What money is there to pay them? How long can withdrawals stay frozen? And who takes over if the platform can no longer be trusted to run custody?

Other parts of finance answer these questions in advance, in contracts and rules. The two Binance agreements reviewed here leave far more to the exchange’s discretion. None of the clauses below would stop a hack. Each one changes what happens to customers afterwards.

Each row pairs one selected traditional-finance example with the two Binance agreements. The examples cover different kinds of firms and risks, and they are not a picture of market practice. The shading is the author’s reading of the documents, not a rating.

1. Say what customers are owed, and who pays first

If ten tokens disappear because custody fails, the agreement should say whether the customer is owed ten replacement tokens or a cash payment. If it is cash, the agreement should say how the tokens are valued, on what date, and when payment is due. EU fund law already works this way: a custodian that loses an asset must return the same asset or its value “without undue delay,” and it cannot contract out of that duty.

Loan agreements settle the valuation question in the same spirit. The leveraged-loan agreement reviewed here has a judgment-currency clause. If a debt must be converted into another currency, the clause fixes the exchange rate at the one on the business day before judgment. If the converted amount buys less than what was owed, the borrower must make up the difference as a separate obligation. A custody agreement could do the same for tokens— name the price source and the date, and require the venue to cover any shortfall between the cash paid and the cost of buying the tokens back.

The duty should also follow the assets when the venue hands custody to another firm. A custody contract filed with the SEC makes the custodian liable for its sub-custodians “as if it had committed such acts and omissions itself.” Both Binance agreements do the opposite: they treat a sub-custodian’s failure as force majeure.

Lenders insist on this too, through the covenant to guarantee and give security. When the borrower forms or acquires a new subsidiary, that subsidiary must sign up to the guarantee and pledge its assets within 90 days. The lenders’ claim follows the assets into the new entity rather than stopping at the old one. J.Crew showed what happens when it does not— assets moved to subsidiaries outside the lenders’ reach. A custody agreement could follow the same logic. Every firm the venue uses to hold client assets would be inside the venue’s liability, not outside it.

The agreement should then say which resources absorb the loss. One possible order is the venue’s own capital first, followed by a dedicated protection fund. Any unpaid balance would remain a claim against the venue.

Loan agreements write this order down. After a default, the agreement lists exactly how recovered money is applied- costs first, then interest, then principal, in a fixed sequence. They also protect the first-loss layer. A restricted-payments covenant limits the dividends and other distributions the borrower can make, so the equity that stands in front of the lenders cannot simply be paid out. A venue’s promise to put its own capital first would mean little without a similar limit on taking that capital out.

These are two separate commitments: a duty to repay, and a plan to fund the repayment. Customers need both, because a strong claim against a firm with no money may still recover very little. Lenders have always known this. A loan is not only a promise to pay- it comes with collateral, guarantees and covenants that keep the borrower able to pay.

2. Give customers a defined claim on the fund

A protection fund should come with rules customers can read before they deposit assets. The rules should say which losses qualify, how claims are made, who approves payment, and what happens if claims exceed the balance.

Lenders already write rules like these for the money they rely on. In a 2008 student-loan securitisation, the indenture sets the reserve’s initial deposit and its required balance, says when it is drawn on, and releases anything above the required balance. When money is short, loan agreements share it by formula, not by choice. In the leveraged loan market, for instance, it is a common practice that a lender that receives more than its ratable share must immediately share the excess with the others. A protection fund could follow the same rule—when claims exceed the balance, every eligible customer is paid the same fraction, and no one is paid first because they asked first.

The agreement should also set a minimum size, regular checks and a deadline for topping up any shortfall. What the fund holds matters too— a fund can lose much of its value at the very moment customers need it.

Asset-based lending offers a useful starting point. Lenders do not simply ask how much the borrower owns. They ask how much of it they could recover if things went wrong—and limit lending accordingly.

The revolving credit agreement reviewed here turns that principle into a formula. Its “borrowing base” counts 90% of eligible card receivables and 90% of the estimated liquidation value of eligible inventory, then deducts reserves for risks that could reduce recovery. Assets that do not meet the eligibility rules do not count.

The formula comes with three contractual safeguards:

First, checks become more frequent as the cushion shrinks. The borrower must certify its borrowing base monthly. If its unused borrowing capacity stays below a specified floor for five consecutive business days, certification becomes weekly.

Second, a shortfall creates an obligation to act. If outstanding borrowings exceed the permitted amount, the borrower must repay the excess within one business day. Reporting the gap is not enough; it must close it.

Third, the party protecting the lenders can tighten the safeguards. The lenders’ agent may introduce additional reserves or narrow asset eligibility on five business days’ notice. The borrower does not control how much its own collateral counts for.

For a crypto exchange protection fund, the lesson is the following— the agreement should define which assets count toward the fund, how they are valued, how often coverage is checked, and when the exchange must replenish a shortfall. An independent party acting for customers should have authority to tighten those requirements as risks rise.

A headline and snapshot of fund balance might tell customers how much is there today. However, these clauses would establish the dynamics of what happens when it falls short and what must remain available regardless of any circumstances, including a heist.

This is common practice in asset-based lending— lenders define what counts as backing, how much it is worth after haircuts and reserves, and what must happen if it falls short. Binance’s Dubai terms, on the other hand, leave a different party in control. Binance, which provides the protection, has the option to decide which losses the fund covers.

The question for customers is therefore how much money would be available after a loss—and whether they have an enforceable right to it. Lenders have long built their contracts around that question: what will be there on the day they need it?

3. Protect the terms that matter most

Some changes alter the reason a customer chose the platform in the first place—who owns the assets, whether they can be lent or pledged, what happens after a custody loss, and whether the protection fund must pay.

The leveraged-loan agreements I have reviewed for this piece and my other projects usually deal with this by reserving certain decisions for each affected lender. A majority of lenders can approve most changes. But a short list needs the written consent of every lender affected, and each item on it has a counterpart in crypto custody:

• Cutting principal or interest, or delaying a payment. The custody counterpart is changing what a customer is owed after a loss.

• Releasing all or substantially all of the collateral, or ranking the loan behind other debt. The counterparts are giving up the customer’s ownership of the assets, or letting them be lent or pledged.

• Changing how payments are shared. The counterpart is changing who is paid from the protection fund, and in what order.

• Changing the amendment rules themselves, or the voting thresholds. This item is easy to overlook. Without it, every other protected term could be unlocked by first rewriting the clause that protects it.

US bond law goes further for the most basic term. Under the Trust Indenture Act, a bondholder’s right to be paid on the due date “shall not be impaired or affected without the consent of such holder,” however many other holders agree.

Neither Binance agreement sets aside any term, including the amendment clause, as one that needs the customer’s consent. Crypto custody agreements could name a similar set of core rights.

Where individual consent is impractical, customers should at least have advance notice and a real chance to leave without a fee while the old terms still apply. The loan market has an answer to the holdout problem that shows what a fair exit looks like. A lender that refuses a change requiring every lender’s consent can be replaced, but only by being paid in full: its outstanding principal, accrued interest and fees. The borrower can remove a lender that will not agree, but it cannot make that lender accept less. For crypto customers, the same principle would mean that anyone who declines a change to a core term can leave with their full assets before the change applies.

That chance means little if withdrawals are suspended for the whole notice period. In the loan market, the lender being replaced is paid out in cash. A customer’s exit right needs the same assurance— open withdrawals, or the change does not take effect.

4. Set deadlines for notice and rules for freezes

Customers should not have to wait until management has calculated the final loss to learn that their assets may be affected. A first notice can say what is known, what is still uncertain and when the next update will come.

Lenders already write notice this way, and they treat it as seriously as payment. The leveraged-loan agreements, usually, require the borrower to notify its lenders of any default “promptly after a Responsible Officer obtains actual knowledge thereof.” The same clause shows the difference a materiality test makes. Notice of litigation is required only if it could have a material adverse effect, but notice of a default is required regardless. Most breaches of covenant get 30 days to fix after notice. Failing to give notice of a default does not: it is an event of default at once.

Rules outside contracts set clocks too. A technology provider to a US bank must tell the bank “as soon as possible” when a cyber incident disrupts its services for four hours or more. The bank must tell its regulator within 36 hours. In the EU, financial firms, including crypto firms authorised under MiCA, must inform clients of a major incident that affects their financial interests “without undue delay as soon as they become aware of it.”

A temporary freeze may be needed to contain an attack. The agreement should state the grounds for imposing one, when it must be reviewed and who can extend it. Here, too, finance has written limits:

• A hard cap. US mutual funds may not delay payment for more than seven days, except when the New York Stock Exchange is closed or restricted, during an emergency, or with the SEC’s permission. The exceptions are named, and the last one needs a regulator’s order, not the fund’s own judgment.

• A cap per period. Under the SEC’s 2014 rule, until it was removed in 2023, US money market funds could gate redemptions for no more than ten business days in any 90-day period.

• A written way back. In the revolving credit agreement reviewed here, the lenders’ control of the borrower’s cash ends once the borrower’s spare capacity has stayed above a fixed floor for 20 days in a row. The restriction comes with its own exit.

• Notice before restricting. Even the protected party gives warning. The lenders’ agent must give five business days’ notice before it adds a new reserve, except during a default.

The Binance agreements contain none of these. There is no deadline for telling customers about an incident. Withdrawals may be suspended “at such time deemed appropriate by Binance,” and will resume once the incident is resolved, with no time limit. The Abu Dhabi terms also exclude liability for a customer’s inability to withdraw during a suspension.

No single deadline suits every incident, and announcing that a freeze is possible can itself trigger withdrawals. The SEC saw this in money market funds. In 2014 it set the gate limit at ten days rather than the 30 it had first proposed, because a longer gate could make investors run sooner. In 2023 it removed the gates altogether. So limits on freezes are a design problem, not a cure. Even so, customers should know who can restrict access to their assets, for how long, and what checks apply to that decision.

5. Arrange a replacement before one is needed

If the venue cannot safely continue custody, someone else needs the authority and the practical means to take over. Securitisations show how. In the Ford auto-loan trust reviewed for the purpose of this piece, the company collecting payments can be removed after defined failures:

• failing to hand over collections, if the failure lasts five business days (with exceptions);

• any other material breach left unremedied for 90 days after notice;

• insolvency.

The servicer must report any of these events within five business days of learning of it. The trustee may then remove it, and must do so if holders of a majority of the senior notes direct it. The trustee must engage a qualified successor: an established servicer with a net worth of at least $50 million. If no successor has accepted when the old servicer stops, the trustee becomes the servicer automatically. The old servicer must cooperate in “an orderly transition,” and hand over the cash, files and records. The original operator can fail while the assets continue to be administered.

A right to take over only works if three things are in place before it is needed. Each one already exists in the agreements reviewed here.

• Control arranged in advance. In the same Ford deal, the bank holding the trust’s accounts signs a separate control agreement. It agrees to follow the trustee’s orders on deposits, withdrawals and transfers “without the need for consent by the Grantor or any other Person.” If it receives conflicting instructions, it follows the trustee’s. And it gives up any right of set-off against the money in the accounts. The handover does not depend on the cooperation of the party being replaced.

• Records the protected party can check before trouble. The leveraged-loan agreement reviewed here lets the lenders’ agent inspect the borrower’s records and question its officers and auditors. It can do so twice a year in normal times, and at any time once a default has occurred. The revolving credit agreement adds field examinations and appraisals of the collateral: one a year, and a second one once the borrower’s spare capacity has stayed below a set floor for five business days.

• A trigger written in numbers. The same revolving credit agreement moves the borrower’s cash into an account controlled by the lenders’ agent when a defined default occurs, or when its spare capacity stays below a fixed floor for five business days. The trigger is measurable, and so is the way back.

For crypto custody, these protections would mean three concrete requirements:

• An agreement with the wallet operator or whoever controls the signing keys, giving a designated replacement custodian the authority and technical ability to transfer the remaining client assets without the exchange’s cooperation.

• Independent checks that the assets held in custody match what the exchange owes its clients, with more frequent checks when signs of trouble appear.

• Clear conditions that activate the handover—such as a verified asset shortfall or withdrawals remaining unpaid beyond a specified deadline—without waiting for the exchange to declare itself in trouble.

A clause promising independent oversight achieves little if the replacement cannot reach the wallets. Neither Binance agreement provides any of this. Instead, customers waive “any claim in rem, injunction, and/or specific performance,” the remedies they would need to reach the assets themselves.

What protection costs, and why it has to be priced

Each of these protections is, in effect, an option the venue writes for its customers. It pays out only in bad states: after a theft, a shortfall or a failure. Like any option, it has a cost. The venue needs capital, funded reserves, preparation and limits on management’s freedom to act. Leaving the protection out does not make that cost disappear. It moves the cost to customers, who carry it without being paid for it and may discover it only after a loss.

Finance has long treated guarantees this way. In 1977, Robert Merton showed that deposit insurance can be valued as a put option on a bank’s assets. Its value rises with the risk the bank takes, not only with the amount it covers. Public backstops are priced too, though usually late. Money market funds paid the US Treasury a fee for its 2008 guarantee, and US banks paid for the FDIC’s 2023 loss through a special assessment. In both cases the price was set during or after the crisis.

Crypto customers today see no price at all. A venue with strong written protections and a venue with none can charge the same fees, and a customer cannot tell them apart until something goes wrong. A venue that adds protections bears the cost but has no way to show what it is worth. So it has little reason to add them, and competition runs on fees and convenience instead.

Pricing these protections in a way that means something requires measuring three things:

1. What each clause pays, and to whom. A duty to replace, a claim on a fund and a right to replace the operator pay out in different events and in different amounts.

2. How its value changes with the venue’s condition. A duty to replace lost assets is worth little from a firm that cannot pay. A promise that looks strong while the venue is healthy can be worth far less when it is stressed, which is exactly when it is called on.

3. What happens when many venues are hit at once. A fund that can cover one theft may not cover a wave of them. A fund that holds assets which fall in the same crisis is weakest when it is needed most. Priced one venue at a time, these risks look smaller than they are.

Regulation has started to set floors. In the EU, MiCA makes crypto custodians liable for losses attributable to them, and DORA requires them to tell clients about major incidents. Those rules apply to firms authorised in the EU, and the agreements reviewed here are with Binance’s Dubai and Abu Dhabi entities. A floor is not a price, though. It says what must be written down, not what the protection is worth. For any customer, the questions remain specific: what does my agreement entitle me to, what resources back that promise, and what is that promise worth if the platform is in trouble?

Who should bear the next loss?

After the next major heist, the industry and policymakers will face a choice about how to protect customers.

One response is a backstop: lend to the venue, guarantee customer balances or otherwise provide support to stop withdrawals from spreading. These measures can be effective in a panic, but they do not all solve the same problem. Lending provides liquidity. A guarantee decides who will bear a loss if the assets available are not enough.

That distinction matters because a heist is not a run, even if it can cause one. In a liquidity crisis, a firm may have enough assets to meet its obligations over time but not enough cash—or immediately available assets—to meet everyone’s demands today. A loan against good collateral can give it time. When assets have been stolen, however, there is a loss to absorb. Lending may help the venue meet withdrawals, but the loan must eventually be repaid; it does not, by itself, replace the missing assets.

In practice, the two problems can arrive together. A theft creates a hole in the balance sheet, and fear about that hole can lead customers to withdraw what remains. By the time authorities are deciding whether to intervene, it may be difficult to know how much of the pressure is temporary and how much reflects a loss the venue cannot cover.

A public guarantee can prevent that uncertainty from becoming a wider panic. But it also raises a distributional question: if customers do not bear the loss, who does? The answer is not always taxpayers directly, and the protection is not always free. In 2008, the US Treasury created a temporary guarantee for participating money market funds, which paid a fee for coverage. After the failures of Silicon Valley Bank and Signature Bank in 2023, the FDIC protected uninsured depositors and imposed a special assessment on banks to recover the resulting loss to its insurance fund. In both cases, the terms of that extraordinary support were set during or after the crisis. US Treasury guarantee; Joint statement on SVB and Signature; FDIC special assessment.

The other door is to settle more of the question before the event. A venue could commit capital to a defined loss order, give customers an enforceable claim on a properly structured protection fund, and agree in advance on notice, withdrawal restrictions and independent control. Customers would have a clearer basis for judging the protection they receive, and venues could compete on its strength as well as on price and convenience.

That does not eliminate the need for public policy. A protection fund that can absorb a loss at one venue may not be large enough if the same shock hits many venues at once. Nor is a contractual promise valuable simply because it is written down: the resources behind it must exist, the claim must be enforceable, and the mechanism must work when the venue is under pressure.

This is where measurement matters. A dollar of protection is not worth the same amount in every state of the world. A promise backed by a healthy, well-capitalized firm may look strong today; the same promise may be worth much less if the firm is stressed and the customer has no control over how the remaining resources are used. A separately held fund, an early notice right or an independent custodian may each change that outcome in a different way.

The task is to measure those protections before a crisis: what each one costs, what it pays or enables, and how its value changes when the firm is healthy, stressed or distressed. That gives customers and venues a basis for pricing protection before they discover they need it.

That is the work we do at CovenantLab.

— Elham Saeidinezhad, Co-founder, CovenantLab

Sources

The incident and who paid

1. Elliptic, “The largest theft in history: North Korea behind the $6.4 billion Binance hack”, October 30, 2026 (final update November 1) and Elliptic, “How North Korea is laundering the $6.4 billion stolen from Binance”. Source for the $6.43 billion count, the timing of the attack and the number of blockchains.

2. Binance, Security Incident Recovery Update #3, October 31, 2026. Source for the three resources used to cover the loss, the amount set aside and the 90% replenished.

3. Binance, Security Incident Recovery Update #4, November 1, 2026. Source for the phased reopening of withdrawals and the 95% figure.

4. Binance FZE, Security Incident Update, October 30, 2026. Binance’s statement that its global-platform hot wallets were affected and that Binance FZE’s segregated client wallets were not.

The contracts

1. Binance Global Terms of Use, effective May 7, 2026. The agreement identifying the ADGM Binance entities. Relevant provisions include custody, liability and amendments.

2. Binance FZE Terms of Use, last updated January 5, 2026. The Dubai agreement.

3. Binance, Frequently Asked Questions on SAFU. Binance’s explanation of the fund and its discretion over eligible losses and claims.

Why a firm pays, and what changes in distress

1. Klein and Leffler, “The Role of Market Forces in Assuring Contractual Performance”, Journal of Political Economy (1981). The economic framework for a firm’s incentive to preserve future business by maintaining trust.

2. Bitfinex, Interim Update, August 6, 2016. Source for the initial approximately 36% reduction in customer balances and issuance of BFX tokens.

3. Bitfinex, 2016 Security Breach Recovery FAQ. Source for the later redemption of BFX tokens or their exchange for equity.

4. US Bankruptcy Code, section 547. The basis for the discussion of certain pre-bankruptcy payments that can be challenged as preferences.

5. In re Celsius Customer Preference Actions (Bankr. S.D.N.Y., July 29, 2025). The Celsius litigation administrator’s claims concerning certain pre-filing withdrawals.

Amendments

1. In re Celsius Network LLC (Bankr. S.D.N.Y., January 4, 2023), decision on Earn account assets. The court’s analysis of the terms, amendments and ownership of Earn assets. The decision did not decide ownership of separate Custody-program assets.

2. EU Payment Services Directive, Article 54. Advance notice and the right to terminate before certain changes to payment-services framework contracts take effect.

Timing and control

1. SEC Release 33-11216, Cybersecurity Risk Management, Strategy, Governance, and Incident Disclosure (2023). The materiality determination and the four-business-day filing framework.

2. EU General Data Protection Regulation, Article 33. A personal-data-breach notification clock tied to awareness, with notification to the supervisory authority.

3. SEC investor bulletin on alternatives to financial-statement audits (2023). The distinction between proof-of-reserves reports and financial-statement audits.

4. J.Crew Group, Form 10-K for fiscal 2016 (SEC EDGAR), Ayotte and Badawi, “Loopholes in Complex Contracts” (NYU) and Orrick, “A ‘Trap Door’ Intact: Fixing the J.Crew Blocker” (2020). The transfer of intellectual property outside the original lenders’ collateral structure, and the covenant named after it.

5. European Market Infrastructure Regulation, Article 45. The clearing-house default waterfall used as a comparison for a written loss order.

6. SEC money market fund reform rule, 2014, SEC money market fund reforms, 2023 and SEC press release, 2023. The earlier limit on certain temporary gates and the later removal of that gate provision.

Backstops

1. US Treasury, Temporary Guarantee Program for Money Market Funds, 2008. A crisis-period public guarantee for which participating funds paid a fee.

2. Treasury, Federal Reserve and FDIC, Joint Statement, March 12, 2023. The decision to protect all depositors of Silicon Valley Bank and Signature Bank, with any loss recovered by a special assessment on banks.

3. FDIC, Special Assessment Pursuant to Systemic Risk Determination. The assessment imposed to recover the resulting loss to the Deposit Insurance Fund.

By Elham Saeidinezhad

A Note from the Author

This piece is longer than usual, partly as a tribute to my postdoctoral supervisor, Professor Perry Mehrling, who passed away on October 13, 2026. Perry’s Money View profoundly shaped how I understand finance, particularly his insight that financial survival depends not simply on the value of assets, but on the ability to meet payment obligations when they come due.

Looking back, my work on the Covenant Structure of Yield in Private Credit was, in many ways, an attempt to follow Perry’s intuition into the more bespoke corners of finance. This article continues that journey by asking how financing contracts shape those survival constraints— how they determine which cash flows and assets remain available to meet obligations, and how these conditions change over time. The central idea is that contracts do not merely protect lenders after default; they also shape the financial constraints that matter before default.

Perry’s intellectual influence is a light I will always carry with me. Yehi zichro baruch — May his memory be a blessing.

In October 2026, one of the companies at the centre of the AI boom was sending a very different signal to credit markets. CoreWeave, whose rapid expansion has been supported by substantial borrowing to finance AI infrastructure, saw its shares and bonds come under significant selling pressure. By October 15, one of its bonds was trading at a yield of approximately 15%. [1] The contrast is striking in a sense that how can a company operating in one of the most promising areas of the economy face such serious questions about its ability to repay its lenders? More importantly, how much of the answer can be found not in its financial statements, but in the financing contracts themselves?

This article— based on our work at CovenantLab, a proprietary financial model developed jointly by Uri Ravid of Columbia University and me — uses CoreWeave’s Delayed Draw Term Loan 5.0, DDTL 5.0, to examine how financing contracts shape credit risk before default. Building on Mehrling’s Money View, we argue that a borrower’s ability to meet its obligations depends not only on its financial condition and access to funding, but also on the contractual rules governing how cash and assets can be used or transferred. Our central contribution is to show how these contractual rights change as financial conditions evolve, and how those changes can affect liquidity, lender recovery, and ultimately the value of a loan.

1. Liquidity comes first: the survival constraint

To understand how contracts affect credit risk before default, we draw on Perry Mehrling’s Money View, which distinguishes between the value of financial assets and the ability to meet payment obligations when they fall due. A borrower may own valuable assets but still face financial distress if it cannot obtain cash at the right time. This is the survival constraint, which places the timing and availability of liquidity at the centre of financial analysis.

Our contribution is to examine the contractual dimension of this constraint. While the Money View explains how institutions obtain liquidity through financial markets, CovenantLab examines how financing contracts govern the use and movement of cash and assets. These contractual rights can change as financial conditions evolve, affecting both the borrower’s ability to meet payments and the resources available to lenders in distress. The contract therefore helps determine how liquidity pressure develops and translates into credit risk.

We can express cash headroom as:

Hₑ,ₜ = Kₑ,ₜ + Iₑ,ₜ + Fₑ,ₜ − Dₑ,ₜ

where:

• Hₑ,ₜ is the borrower’s cash headroom at payment date t.

• Kₑ,ₜ is usable cash held at the beginning of the period.

• Iₑ,ₜ is cash received during the period and usable for the payments being assessed.

• Fₑ,ₜ is additional financing that can actually be obtained and used before payment is due.

• Dₑ,ₜ is the total amount of payments due.

The subscript e identifies the relevant legal entity. All amounts cover the same payment interval. If Hₑ,ₜ ≥ 0, the borrower has enough usable resources to meet the payments included in the calculation. If Hₑ,ₜ < 0, it has a funding shortfall that must be resolved.

The central difficulty in applying the survival constraint is determining how much of the borrower’s reported cash and expected inflows will actually be available when its obligations fall due. This is where the contractual structure becomes important, because cash recorded on the balance sheet may be restricted, while expected inflows may arrive too late or be subject to contractual arrangements that limit their use.

Consider a simple illustration, unrelated to CoreWeave’s actual financial data. A borrower holds $100 million in cash, expects $40 million in customer payments, and has $120 million in obligations coming due. Based on these figures, the borrower appears to have a $20 million liquidity surplus. However, suppose $30 million of its cash cannot be used for the relevant payment because of contractual restrictions, and $20 million of its expected inflows will arrive after the payment date. Assuming no additional financing is available, its liquidity position becomes:

H = 70 + 20 − 120 = −$30 million

The borrower therefore faces a $30 million funding gap despite reporting sufficient cash and expected inflows to cover its obligations. The important point is that liquidity depends not simply on how much cash a borrower has or expects to receive, but on when that cash becomes available and what the financing contract permits the borrower to do with it.

This is where contractual analysis becomes essential. A financing agreement may determine which accounts are restricted, which receipts can fund debt service, when additional financing can be drawn, and which payments take priority. Liquidity is therefore not simply the amount of cash a company owns. It is the money the relevant borrowing entity can contractually and operationally use when payment is due.

2. How contracts change the survival constraint

Mehrling’s framework explains why access to liquidity matters for financial survival. CovenantLab extends this argument by examining how the financing contract governs that access. Although contractual provisions are not systematically incorporated into conventional credit pricing and liquidity models, they can play an important role in determining how much cash remains available to meet the borrower’s obligations. Some provisions permit cash to leave the borrower through distributions or transfers, while others restrict these payments, require cash to remain in designated accounts, or direct proceeds toward debt repayment. Importantly, these contractual permissions and restrictions can change as financial conditions evolve, making the availability of liquidity partly dependent on the state of the borrower and the rights that become exercisable under the agreement.

These provisions do not generate cash, which partly explains why their economic significance may be overlooked in models that focus primarily on cash generation, balance-sheet liquidity, and access to external financing. However, they govern how cash that has already been generated or obtained can be allocated, retained, or transferred, and therefore influence the resources available to satisfy future payment obligations. This introduces a contractual dimension to liquidity risk that is distinct from, but closely related to, the borrower’s operating performance and funding conditions.

One important mechanism is cash leakage, which we define as cash leaving the resources available to the assessed borrower through distributions and other transfer channels permitted under the financing arrangement. Such payments may be entirely lawful and contractually permitted; the term does not imply misconduct. The relevant economic question is whether these transfers reduce the cash available to meet the borrower’s obligations at a future payment date. To isolate this mechanism, we distinguish between required payments and cash leakage:

Dₑ,ₜ = D⁰ₑ,ₜ + Lᶜᵃˢʰₑ,ₜ

The survival constraint becomes:

Hₑ,ₜ(C, zₜ) = Kₑ,ₜ₋ + Iₑ,ₜ + Fₑ,ₜ − D⁰ₑ,ₜ − Lᶜᵃˢʰₑ,ₜ(C, zₜ) ≥ 0

Here:

• D⁰ includes operating expenses, debt service, and other required payments, excluding transfers counted separately as leakage.

• Lᶜᵃˢʰ is cash actually paid through the distribution and transfer channels under examination.

• C represents the contractual provisions.

• zₜ represents the circumstances in which they operate, including balances, covenant compliance, notices, and payment decisions.

Contractual provisions also affect other components of the survival constraint by determining which financial resources are available to the borrower and when payment obligations must be met. For example, restrictions on bank accounts may limit the use of existing cash, conditions on undrawn credit facilities may affect access to additional financing, and repayment provisions may accelerate or defer payment obligations following specified events. Importantly, these provisions do not necessarily affect liquidity at all times.

Their economic consequences depend on the borrower’s financial condition, the occurrence of contractual triggers, and whether the relevant rights are exercised. The contractual dimension of liquidity is therefore dynamic in which the same borrower, with the same assets and operating cash flows, may face different liquidity constraints as its contractual rights and obligations change over time. This is why measuring liquidity requires more than observing cash balances and expected inflows; it also requires understanding the contractual conditions governing their availability and use.

To formalize this, for each permitted route k, we can write:

Lᶜᵃˢʰₑ,ₜ(C, zₜ) = Σₖ gₖ(C, zₜ) × ℓₖ,ₜ

Here, gₖ indicates whether the relevant conditions permit the payment. The amount ℓₖ,ₜ is actually paid through that route, subject to applicable limits and payment priority. A payment qualifying under multiple exceptions is counted once.

Consider another illustration. A borrower begins the month with $100 million of usable cash and has $80 million of required payments. Without a distribution, it has $20 million of headroom. Suppose its agreement permits a $15 million payment to its parent, and the payment is made before those obligations fall due:

H = 100 − 80 − 15 = $5 million

The borrower can still pay. No payment default has occurred. Yet its liquidity cushion has fallen from $20 million to $5 million. Holding everything else constant:

∂H/∂Lᶜᵃˢʰ = −1

Every dollar transferred out of the borrower reduces its immediate cash headroom by one dollar, holding other cash flows constant, and may also affect liquidity in subsequent periods because cash distributed today is no longer available to meet future payment obligations. However, a reduction in liquidity does not necessarily imply an equivalent reduction in economic value or creditor recovery. For example, a borrower may use $15 million to acquire an income-producing asset, reducing its current cash balance while potentially increasing future cash flows and the value of its assets.

By contrast, a $15 million distribution to shareholders reduces available cash without necessarily creating a corresponding asset or future income stream for the borrower. Although both transactions have the same immediate effect on liquidity, their consequences for future payment capacity and creditor recovery may be very different. The contractual distinction is equally important— even when a transaction creates economic value, the resulting asset or cash flows may not remain within the lender’s collateral package or otherwise be available to satisfy its claim. CovenantLab therefore distinguishes between the immediate liquidity effect of a permitted transaction, its potential contribution to future cash generation, and the extent to which the resulting value remains accessible to creditors.

3. CoreWeave’s DDTL 5.0: where contractual permissions matter

CoreWeave’s DDTL 5.0 provides a useful example of how contractual provisions can affect liquidity within a specific borrowing structure. An important distinction is that the financial condition of CoreWeave as a publicly traded company is not necessarily the same as the liquidity position of the borrower under DDTL 5.0. The facility has its own borrowing entities, designated accounts, collateral arrangements, and payment rules, meaning that CoreWeave’s share price, market capitalization, or publicly traded bond yields cannot, on their own, establish how much cash is available to service this particular debt.

To understand the contractual dimension of liquidity, we therefore need to examine how cash is allocated within the financing structure. Sections 2.20 and 6.06 of the agreement are particularly relevant because they address the cash waterfall, distribution conditions, and restricted payments, including specified tax payments, acquisition invoice payments, and transfers among Credit Parties. [3] These provisions matter not simply because they permit or restrict certain transactions, but because their economic effects depend on the order in which payments are made, the conditions under which cash may be transferred, and the restrictions that become effective as circumstances change. For example, a cash trap may prevent certain distributions without necessarily restricting payments that have priority under the waterfall. Consequently, two credit agreements with similar covenant language may provide very different liquidity protections depending on their exceptions, payment priorities, and release conditions. This is precisely why the economic analysis of covenants requires examining how contractual provisions interact, rather than simply counting the restrictions contained in an agreement.

The economic significance of these contractual provisions depends not simply on whether a restriction exists, but on when it becomes effective, which transactions remain permitted, and how the resulting allocation of cash affects the borrower’s ability to meet its obligations. This distinction is particularly important for transfers between accounts within the same borrowing entity, since an internal movement of cash does not necessarily reduce the borrower’s total cash holdings but may change the amount available for a particular payment. For example, transferring cash from an unrestricted operating account to a restricted account may leave the borrower’s balance sheet unchanged while reducing its immediately usable liquidity. Conversely, releasing cash from a restricted account may improve payment capacity without generating any additional cash. The relevant contractual effect is therefore a change in the availability and permitted use of existing resources, rather than necessarily a change in their total amount. CovenantLab distinguishes these changes in cash availability from actual cash outflows to avoid counting the same transaction twice when assessing the borrower’s liquidity position.

After applying the contractual rules, any remaining funding gap is:

Gₑ,ₜ = max(0, −Hₑ,ₜ)

The funding gap measures the additional liquidity the borrower needs to meet its obligations at a particular payment date, but it should not be interpreted as a probability of default or an expected credit loss. A borrower facing a liquidity shortfall may still obtain additional equity, refinance its obligations, draw on available credit facilities, or access other sources of funding before payment becomes due. Similarly, a covenant breach may trigger restrictions, additional creditor rights, or an event of default under the agreement without necessarily resulting in a missed payment, while a payment default does not necessarily imply that the lender will incur a loss if sufficient value can ultimately be recovered. The purpose of the contractual analysis is to identify how these mechanisms affect the transition from liquidity pressure to potential credit losses, rather than assuming that deterioration in one measure automatically produces deterioration in the others.

4. A borrower can keep paying while lender protection deteriorates

Liquidity is only one dimension of contractual credit risk. A lender also depends on the assets and cash flows available to support its claim, particularly when the borrower enters financial distress. This introduces a second dimension— the contractual erosion of credit support, which can occur even when the borrower continues to meet its scheduled payment obligations. The distinction is important because the ability to service debt today does not necessarily tell us how much value will remain available to creditors in the future.

Three mechanisms are relevant. Cash leakage reduces the cash available within the borrower’s payment perimeter, asset leakage transfers assets or their associated income rights away from the entities supporting the lender’s claim, and collateral leakage reduces the assets over which the lender holds enforceable security. These mechanisms may overlap, but their economic consequences are not identical. Consider a borrower that owns an income-producing data center pledged as collateral. A sale of the asset at fair value, with proceeds applied to debt repayment, may improve the lender’s position, while a transfer that releases the collateral without equivalent protection may weaken its recovery prospects. In both cases, the borrower may continue making scheduled payments, even though the resources available to creditors have changed.

CoreWeave’s DDTL 5.0 illustrates why these mechanisms must be examined through the interaction of contractual provisions rather than individual covenants. Section 2.09 addresses prepayments and related proceeds, Section 6.04 governs investments, Section 6.05 addresses dispositions, and Section 9.18 concerns the release of liens and guarantees. [3] Together, these provisions establish conditions under which assets may be transferred, proceeds allocated, and creditor protections modified. Their economic significance depends on what the borrower receives in return, whether the transferred assets continue to support the debt, whether the lender’s security remains enforceable, and whether the transaction reduces outstanding obligations.

The central distinction is between the current value of the assets supporting a loan and the contractual conditions under which those assets may cease to support it. Conventional recovery analysis often begins with asset values, collateral coverage, and creditor priority, while contractual analysis adds a dynamic dimension by examining how the resources available for recovery can change before default. Importantly, a contractual permission does not establish that a transfer has occurred or that the lender has suffered a loss. It identifies a possible channel through which credit support may change, whose actual effect depends on the transaction, the applicable conditions, and the protections that remain in place.

5. From contractual provisions to credit risk

To understand how contractual provisions affect credit risk, we need to distinguish between the economic resources generated by the borrower and the contractual conditions governing their availability to creditors. The borrower’s operations generate cash flows and support asset values, while its financing arrangements determine how additional funding can be obtained. However, these resources do not automatically remain available to service a particular loan. The credit agreement governs how cash and assets may be retained, distributed, transferred, pledged, or released, and therefore influences which resources remain available to meet payment obligations and support creditor recovery.

This creates three connected dimensions of credit risk: economic resources, including operating cash flows, asset values, and financing availability; contractual allocation, which determines how these resources may move between entities and creditors; and lender outcomes, including payments received, recoveries obtained, and losses incurred. Changes in business conditions can affect all three dimensions, but the contractual structure determines how some of these effects are transmitted. For example, weaker operating cash flows may reduce liquidity directly, while also activating contractual restrictions on distributions or additional borrowing. Similarly, a decline in asset values may weaken collateral coverage, but its effect on recovery also depends on whether the lender retains enforceable claims over those assets and their proceeds.

This approach draws on the organizing principle of the Clear Street/Re7 framework, which examines financial risk through the connections between different sources of exposure rather than treating each risk independently. [4] For CovenantLab, the important implication is that contractual provisions should be understood as part of a connected financial system, in which changes in business conditions interact with contractual rights to affect payment capacity and recovery over time. The relevant question is therefore not simply whether a covenant is present or restrictive, but what economic outcome it can influence, under which conditions, and through which financial mechanism. This allows contractual analysis to complement conventional credit-risk models by examining how the distribution of potential lender outcomes changes before default occurs.

6. When does contractual risk become a lender loss?

A reduction in available liquidity or the release of collateral does not necessarily result in a credit loss, although both may change the lender’s exposure to future losses. A principal loss occurs when the amount ultimately recovered is insufficient to repay the outstanding principal claim, after accounting for payments and recoveries received. The important distinction is therefore between changes in the resources available to support repayment and the actual realization of a lender loss. A contractual provision may weaken liquidity or collateral protection without producing an immediate loss, while its economic significance depends on how it affects the amount and timing of recoveries under different future conditions.

Chapter 6 of The Handbook of Insurance-Linked Securities distinguishes between attachment, the point at which an investor begins to incur a loss, and exhaustion, the point at which the exposed principal has been fully lost. [5] Although these concepts originate in structured risk-transfer instruments, they provide a useful way to describe the distribution of potential credit losses. In the context of a loan, attachment can be understood as the point at which recoveries become insufficient to protect the lender’s full principal exposure, while exhaustion represents the loss of that entire exposure. These are analytical thresholds rather than contractual features of CoreWeave’s DDTL 5.0. The contribution of contractual analysis is to examine how changes in cash availability, collateral protection, and creditor rights can affect the likelihood and severity of losses between these two outcomes, rather than treating every deterioration in liquidity or collateral coverage as an immediate credit loss.

For lender j:

Lⱼ = max(0, EADⱼ − Rⱼ)

where:

• EADⱼ is outstanding principal exposure at resolution.

• Rⱼ is principal recovered after applying contractual rights, priority, enforcement, and relevant costs.

• Lⱼ is the resulting principal loss.

Principal loss attaches when:

Lⱼ > 0

The remaining principal exposure is exhausted when:

Lⱼ = EADⱼ

For illustration, a lender with $100 million of outstanding principal that recovers $90 million loses $10 million. Loss has attached, but the exposure is not exhausted. If it recovers nothing, the entire remaining $100 million is lost.

This separates the probability of loss from its severity:

E[Lⱼ] = P(Lⱼ > 0) × E[Lⱼ | Lⱼ > 0]

The first term measures the probability that the lender incurs a principal loss, while the second measures the expected size of that loss conditional on its occurrence. Conventional credit-risk analysis often estimates these outcomes using the borrower’s financial condition, asset values, and historical default and recovery experience. Contractual analysis adds another dimension by examining how the financing agreement can change the distribution of these outcomes over time, even when the underlying economic characteristics of the borrower remain unchanged.

The contractual effects on loss probability and severity are not necessarily the same. Restrictions on distributions and cash transfers may preserve resources needed to meet scheduled payments, while restrictions on asset transfers, additional secured borrowing, and collateral releases may preserve the assets available for creditor recovery. A provision may therefore improve expected recovery without materially changing the likelihood of a payment default, while another may support near-term liquidity without providing equivalent protection in distress. Importantly, these effects depend on when contractual restrictions become effective, which exceptions remain available, and whether the relevant rights are exercised. The economic contribution of a covenant is therefore not determined simply by its existence, but by how it changes the lender’s exposure under different financial conditions.

Finally, principal loss is only one component of the lender’s economic outcome. Delayed payments, unpaid interest, and changes in the timing of recoveries can reduce the present value of the lender’s claim even when the full principal amount is eventually recovered. A complete valuation must therefore consider both the amount and timing of payments and recoveries, rather than relying exclusively on the probability and severity of principal loss.

7. Measuring the distribution of potential losses

Expected loss is an important measure of credit risk, but it does not tell us how severe a lender’s losses could become. Two loans may have the same expected loss, even though one is more likely to experience small losses while the other carries a greater risk of a large loss. For lenders, this distinction matters because the size of a potential loss can be just as important as its probability.

This is another area where contractual protections matter. For example, we could compare CoreWeave’s existing collateral release provisions with an alternative agreement that places tighter restrictions on releasing collateral. The question is whether stronger protections would reduce the likelihood of losses, limit their severity, or both. However, the answer depends not only on the contract itself, but also on future cash flows, asset values, borrower decisions, and the ability of lenders to enforce their rights.

These estimates are necessarily uncertain, particularly when some financial information is unavailable or future transactions are unknown. As Chapter 6 emphasizes, it is important to examine how different assumptions affect the results rather than rely on a single estimate. [5] The purpose of contractual analysis is therefore not to predict losses with certainty, but to understand how different contractual protections may change the risks lenders face.

8. What Is Contractual Protection Worth? The CoreWeave Case

The economic value of contractual protection ultimately depends on whether it changes the amount or timing of cash flows and recoveries received by lenders. Two loans may have the same borrower, maturity, and promised interest payments, but different values because their financing agreements provide different protections. For example, one agreement may allow collateral to be released under relatively broad conditions, while another may impose tighter restrictions. If those restrictions preserve additional value for lenders during financial distress, the second loan may have a higher present value and a lower required yield, holding other factors constant. The important point is that the value of a covenant comes from its economic consequences, rather than simply from the presence of restrictive language in the agreement.

This is also where contractual analysis differs from conventional approaches to credit valuation. Standard models generally estimate the value of a loan using expected payments, default risk, recovery assumptions, and the compensation investors require for bearing risk. CovenantLab examines how contractual provisions may change these underlying financial outcomes as the borrower’s circumstances evolve. A restriction on cash distributions may help preserve liquidity for scheduled payments, while a restriction on collateral transfers may protect the value available for recovery. These protections need not have the same effect on credit risk, and their value depends on the conditions under which they operate. By comparing otherwise similar financing arrangements with different contractual protections, we can examine how those differences affect the lender’s expected receipts and the present value of its claim.

Returning to CoreWeave, the October selloff provides an important motivation for this analysis, but market prices alone cannot tell us which contractual risks have actually materialized. A high bond yield may reflect concerns about leverage, refinancing, asset values, liquidity, or broader market conditions. It does not establish that cash has been transferred out of the DDTL 5.0 borrower, that collateral has been released, or that any contractual provision has been breached. Establishing these outcomes requires evidence about actual cash movements, collateral arrangements, debt repayments, and compliance with the financing agreement. It also requires distinguishing the financial position of CoreWeave as a corporate group from that of the specific borrowing entities and secured creditors under DDTL 5.0.

The broader lesson is that market prices reflect investors’ assessment of credit risk, while contractual analysis helps explain how some of that risk can develop and change over time. The contract does not replace conventional credit analysis, but it provides an additional economic dimension by identifying how the rights governing cash, assets, and creditor claims can influence liquidity, recovery, and ultimately the value of a loan.

Conclusion: the contract is part of the financial mechanism

CoreWeave illustrates why credit risk depends not only on leverage, cash flows, and asset values, but also on the contractual rules governing them. A borrower may own valuable assets yet lack cash when payments are due, or continue servicing its debt while the collateral protecting lenders changes.

Building on Mehrling’s Money View, which emphasizes the importance of meeting payment obligations on time, CovenantLab examines how financing contracts shape liquidity, recovery, and loan value over time. CoreWeave’s DDTL 5.0 provides a useful case study for understanding these mechanisms without confusing market concerns with actual contractual events.

Ultimately, the question is not simply what covenants a loan contains, but what those covenants protect and how much that protection is worth.

Sources

[1] Reuters, coverage of CoreWeave’s October 15, 2026 selloff and the debt financing behind AI infrastructure expansion.

[2] Perry Mehrling, Liquidity Changes Everything.

[3] CoreWeave, DDTL 5.0 Credit Agreement, particularly Sections 2.09, 2.20, 6.04, 6.05, 6.06, and 9.18.

[4] Re7 Capital and Clear Street Digital, A Framework for Decomposing DeFi Risk in the Age of AI (2026), particularly Sections 1.2–1.4 and 5.1–5.2.

[5] Pauline Barrieu and Luca Albertini, eds., The Handbook of Insurance-Linked Securities, Chapter 6, “Rating Methodology,” by Cameron Heath, particularly Sections 6.1.5, 6.2.3–6.2.5, and 6.4.2–6.4.3.

Aug 19, 2026

On August 18, the SEC proposed Regulation Crypto Assets (File No. S7-2026-27). It would create two new offering exemptions for crypto assets and a disclosure regime designed around them. It would also create a safe harbor under which an investment contract is treated as having ceased to exist once the issuer has completed the essential managerial efforts on which investors were asked to rely.

The proposal is not yet law, but the boundary it draws is already analytically useful because it separates native crypto arrangements from arrangements in which a token refers to an asset outside the network. In our recent paper, called “Significant Risk Transfer (SRT) as a Framework for Tokenized Real-World Assets”, Uri Ravid and I examine a common tokenized real-world-asset structure. The sponsor first places a building, loan, or receivable inside an SPV, which becomes the legal owner of the asset in the relevant property register, credit agreement, or other authoritative record. The SPV then issues interests to investors, and tokens are minted to record who holds those interests. Legal title to the underlying asset does not pass to the token holders. What they receive instead is a contractually defined claim on some part of the asset’s cash flows, gains, and losses.

This separation between ownership and economic exposure is the reason we describe the structure as a synthetic risk transfer. In a synthetic transfer, including the ones that happen through derivatives, the originator keeps the underlying asset and conveys some of its risk and return through a separate contract. A bank may retain a portfolio of loans while transferring a defined layer of credit losses to an investor; the tokenized RWA structure performs the same basic operation when the SPV retains the building, loan, or receivable while investors receive a contractual share of its economics. The analogy is not based on the use of a blockchain, and the token itself does not create the transfer. The transfer is created by the governing documents, which define the cash flows investors receive, their position in the capital structure, the conditions under which they can redeem or enforce their claim, and the powers the sponsor retains after issuance. The token records the holder of that claim, but the contract determines how much exposure has actually moved.

Once the instrument is classified in this way, the relevant apparatus comes from the bank risk-transfer market, where it is not enough for an originator to state that risk has been transferred, because the transfer must also be shown to be significant. The question is how much of the stated exposure genuinely reached the investor, and how much the originator can recover, redirect, or subordinate through rights that remain inside the contract.

The SEC proposal provides a useful first application of this apparatus. At the perimeter of the new regime, it recognizes that a crypto asset connected only to its native network and a crypto instrument connected to an external asset are not the same object. It leaves unresolved the harder question of how the second object should be assessed.

The question is how much of the stated exposure genuinely reached the investor, and how much the originator can recover, redirect, or subordinate through rights that remain inside the contract.

The proposed perimeter: native crypto in, external assets out

The new regime would apply only to a “covered investment contract.” To qualify, three conditions must be met: the investment contract must involve a crypto asset; that crypto asset must not itself be a security; and the same contract must not involve any other asset, whether or not that other asset is a security.

The third condition is especially important. An investment contract involving a non-security token and the development of the token’s network may qualify for the new regime. It means the new rules may apply when:

• A company creates a crypto token for its own network or app.

• The token is not itself a security, such as a share or bond.

• Investors buy the token partly because they expect the company to develop the network.

For example, imagine a company creates a token that people will use inside a new online platform. The company sells the token while promising to build and improve that platform. This arrangement may qualify for the SEC’s new regime because the investment involves only the token and its network—not a share, loan, building, or other separate asset.

As a result, a tokenized share, note, or fund interest would not qualify because the token represents a security. A contract involving a building, loan, receivable, or another real-world asset would also be excluded, but for a different reason: the contract involves an additional asset. According to the SEC, these offerings should instead use an existing path, such as registration, Regulation A, or Regulation D. However, it may still be difficult in some borderline cases to determine whether another asset is “subject to” the contract.

This does not mean that the SEC treats every RWA token in the same way. “Tokenized RWA” only means that a connection to a real-world asset is recorded on a blockchain. It does not describe one specific type of legal investment. For example, a token might represent a share in the company that owns the asset, a loan issued by an SPV, an interest in a partnership or fund, a contractual right to receive payments, or another type of investment contract involving a non-security crypto asset. Two tokens may be linked to the same building, loan, or receivable and may use the same blockchain technology, but they can still give investors very different rights, payment priorities, protections, and risks. The law therefore looks at the rights created by the contracts and other legal documents—not simply at the fact that those rights are recorded through a token.

The proposal, therefore, makes a narrower but important distinction. Its new regime is designed for an investment contract whose only relevant asset is a non-security crypto asset connected to its native network or application. Once the arrangement also includes a share, note, fund interest, building, loan, receivable, or other external asset, it no longer fits that defined category. A real-world asset cannot be treated as a minor part of a crypto arrangement. It provides its own source of value and comes with its own legal rights and obligations. As a result, the investor’s return may depend on more than the success or use of the crypto network. It may also depend on who owns the real-world asset, how much income it generates, the debts of the company or SPV that holds it, the order in which investors are paid, and how much control the sponsor continues to have over the asset and its income. The proposal therefore does not say that every RWA token is the same type of security. It simply says that a claim linked to a real-world asset cannot qualify for the new crypto regime merely because it is recorded on a blockchain.

This distinction is similar to the economic distinction we make in our paper. A native token is expected to gain value from the growth, use, and success of its own network. By contrast, an investor in the RWA structure we study owns an interest in a legal entity or has a contractual claim linked to a real-world asset. That asset is still controlled by property law and by legal documents such as the credit agreement and the SPV agreement, which also determine the order in which investors are paid. The blockchain may show who holds the investment, but it does not determine who legally owns the asset, who gets paid first, whether the investor’s rights are legally protected, how those rights can be enforced, or how much control the sponsor keeps over the asset and the SPV.

The new regime applies only to non-security tokens tied to their own networks. If the arrangement includes a security or real-world asset, its separate value and legal rights place it outside the regime.

Managerial effort and retained exposure are different tests

The proposed Rule 400 safe harbor is built around managerial effort. A covered investment contract would be deemed to have ceased if the issuer has completed or permanently stopped all the essential managerial efforts it promised to perform. The issuer must also intend to make no new promises of such efforts and must file Form TR with a certification and supporting analysis.

This is how the proposal applies the logic of Howey1. Under Howey, the arrangement may be an investment contract because investors rely on the issuer to perform important work. Rule 400 says that once the issuer finishes or permanently stops that work, the investment contract may end—if all the rule’s other requirements are also met. This legal framework is the foundation of the SEC’s proposed rule.

Our apparatus, on the other hand, has a different object and asks a different question: how much control does the sponsor retain over the investor’s economic exposure? To understand this question more formally, let the gross value of the exposure that the contract appears to convey be Vᵍ. Let Λⱼ measure the value of each right the sponsor keeps after selling the investment. For example, the sponsor may keep the right to sell the asset without investor approval. That right is valuable to the sponsor because it gives the sponsor control and may weaken the investor’s claim. The symbol j simply identifies each retained right: right 1, right 2, right 3, and so on.

These permissions may include the right to incur debt ahead of holders, transfer the asset to an affiliate, sell without holder consent, take fees ahead of distributions, or make redemption discretionary. Significance can then be written, as a conceptual decomposition, as:

When σ is close to one, most of the exposure described in the offering has been conveyed on terms the sponsor cannot later reverse. A value closer to zero describes a largely nominal transfer. In that case, the sponsor has retained valuable ways to redirect cash flows, change the holder’s priority, remove the asset from beneath the claim, or terminate the claim on terms it controls. It is important to note that this equation is an organizing device rather than a claim that these permissions are easy to price. Their values may interact, and the difficult analytical work lies in identifying them, estimating their effect, and avoiding double counting.

This distinction is fundamental. Rule 400 asks whether the promoter is still performing the essential work it promised. The SRT framework asks a different question: what legal power does the sponsor retain over the investor’s exposure once that work finishes? Rule 400 focuses on the promoter’s conduct while SRT focuses on the sponsor’s contractual control. That control matters from the moment it is granted because it can weaken the investor’s position even if the sponsor never exercises it.

Rule 400 may therefore be coherent for the native-network instruments it is designed to cover without answering the separate risk-transfer question raised by an asset-backed structure. Managerial effort and retained exposure are not competing descriptions of the same thing; they are complementary tests applied to different dimensions of the arrangement.

Rule 400 asks what the promoter must still do. SRT asks what the sponsor can still do to the investor’s claim—even if it never exercises that power.

Certification is not the same as assessment

Form TR would require the issuer to identify the covered investment contract and crypto asset. The issuer must certify that the promised essential managerial efforts have been completed or permanently ceased and provide an analysis supporting that certification. Rule 103 would separately require mainly principles-based narrative disclosure adapted to the issuer’s facts and circumstances.

The bank SRT apparatus is more demanding in one specific respect. It begins with a defined question—how much risk genuinely left the originator—and combines mechanical thresholds with a structural reading of the contract for routes through which that exposure can return. Supervisors examine call options, excess spread, replenishment rights, amortization, early-termination provisions, and servicing discretion. They then decide whether regulatory capital relief should be available and can reconsider the conclusion if the transfer changes over time. The prudential purpose of this exercise differs from the investor-protection purpose of securities law, but the method of identifying and evaluating retained rights can be carried across.

This is the practical difference between disclosure, certification, and assessment. Disclosure tells the investor that a contractual permission exists, certification records the issuer’s conclusion about a stated condition, and assessment asks what the permission does to the quantity and stability of the exposure that was transferred.

Consider two property tokens that each advertise an 8 percent yield and a 60 percent loan-to-value ratio. In the first structure, holders must approve a sale, the SPV cannot incur debt ahead of them, affiliate transfers are prohibited, and redemption is a direct claim against the SPV. In the second, the sponsor may sell the asset, add senior debt, transfer the asset to an affiliate, and decide whether or when redemption occurs. The asset class, blockchain, yield, and stated leverage are the same. The contracts nevertheless convey different quantities of exposure because the sponsor retains a much larger set of options in the second structure.

What the SEC’s line leaves unresolved

Regulation Crypto Assets would leave tokenized securities and asset-linked investment contracts under existing securities-law frameworks. Those frameworks govern registration, investor eligibility, disclosure, resale, and liability for misleading statements. But they do not measure how much of the underlying asset’s economic exposure has actually passed to investors. Regulation D makes this gap especially clear: it provides an exemption from registration, not a standard set of investor rights. Two Regulation D offerings may reference similar assets and promise the same yield while giving investors very different economic and legal positions.

The investor’s true position is defined by the SPV agreement, payment waterfall, covenants, priority rules, enforcement rights, and powers retained by the sponsor. The analysis must therefore go beyond asking what asset “backs” the token. It must determine what exposure each investor class receives, where it ranks for payment, and what rights it has to payment and redemption. It must also ask whether debt or fees can be placed ahead of investors, whether the asset can be replaced, sold, or transferred to an affiliate, and who controls amendments and restructurings. These questions must be revisited when the sponsor exercises its rights or the governing documents change.

The SEC’s proposal separates native crypto from asset-linked instruments, but classification is only the first step. For an asset-linked token, neither the token’s legal status nor the promoter’s continuing work answers the key economic question: how much exposure has genuinely and irreversibly passed from the sponsor to investors?

That is the question the SRT framework is designed to answer.

By Elham Saeidinezhad, Ph.D. | CovenantLab | July 28, 2026

There is roughly $345 billion of real-world assets recorded on blockchains today, and only about 8% of that value ever actually changes hands. The rest sits there — issued, settled, custodied, and completely still. The explanations usually offered are technology, regulation, and liquidity, but the infrastructure works, the regulatory perimeter gets clearer every quarter, and liquidity is a symptom rather than a cause.

What is actually missing is a measurement. When you buy a tokenized building, you are not buying the building; you are buying a contract that hands you some portion of the building’s economic exposure while the sponsor keeps legal ownership. The only question that determines what that contract is worth is how much of the exposure genuinely moved to you, and how much the sponsor quietly kept the ability to take back. That quantity has a name — the significance of the risk transfer — and in another asset class it is measured formally, by supervisors, every single day, because a bank’s capital relief depends on the answer. In tokenized real-world assets, nobody measures it at all. That, and not the plumbing, is what keeps this market frozen.

Start with what does not happen

When a building in Dubai is tokenized, legal title to that building does not move onto a blockchain. It stays exactly where it was, in the register maintained by the Dubai Land Department. When a private loan is tokenized, the loan stays inside its credit agreement, administered by its agent, governed by the law the parties chose. No public chain has the authority to move either of those things, and no token issued by a sponsor conveys them.

What the sponsor does instead is place the asset inside a vehicle it controls, and then sell you a contract. That contract promises to pass through the asset’s economics — its cash flows — together with a specific, negotiated list of rights over what the sponsor may and may not do with the asset while you hold the claim. The token is simply the record of who holds that contract. The economics move to you; the ownership stays with the originator.

That structure is not new, and it is not a crypto invention. It has a precise name in financial engineering, and an entire institutional market built around it.

What a synthetic risk transfer actually is

Imagine a bank holding a portfolio of corporate loans. The bank wants to reduce the risk it is carrying, because that risk consumes regulatory capital and capital is the scarcest thing on its balance sheet. The obvious solution is to sell the loans, but selling them is often impossible or undesirable: the borrowers are relationship clients who would notice and object, the loan documents may require borrower consent to assign, and the legal mechanics of transferring hundreds of separate credit agreements are slow and expensive. The bank wants to shed the risk without moving the asset.

So it writes a contract instead. The bank keeps every loan on its balance sheet and continues to service them, and separately agrees with an investor that the investor will absorb the first slice of credit losses on that identified portfolio — say the first 8% — in exchange for a substantial ongoing fee. Usually the investor funds the position up front by buying a credit-linked note, so the cash sits with the bank as collateral and is drawn down as losses arrive. The loans never move. Title never moves. What moves is economic exposure, delivered entirely by contract. That is a synthetic risk transfer, and banks use it constantly to manage capital. It is not the same thing as a synthetic CDO, and the distinction matters: an SRT references a real, identified portfolio that the seller actually holds.

Now notice what the investor in that structure owns. It is not the loans. It is a defined position in a contractually defined loss waterfall, and every word of that definition is load-bearing.

The key idea: significance is a measurable quantity, and someone is required to measure it

Here is the part the tokenization market has never absorbed.

A bank does not get capital relief simply because it signed a document titled “risk transfer.” It gets relief only if a supervisor is satisfied that the risk transfer is significant — that a real, substantial share of the portfolio’s economic exposure genuinely left the bank and will not find its way back. Under the European rules this sits in Articles 244 and 245 of the Capital Requirements Regulation, and the bank has to notify the supervisor months before the transaction closes so the assessment can be made. The label on the document buys nothing. The measurement is the thing.

The measurement has two layers, and the second is the interesting one.

The first layer is arithmetic. Broadly, the originator must not hold on to more than about half the risk in the mezzanine tranches, or, where there is no mezzanine, must part with the large majority of the first-loss piece. These thresholds are easy to check and easy to satisfy, and they are not where the real work is.

The second layer is a reading of the contract. Supervisors examine the transaction’s structural features — call options, excess spread, replenishment mechanics, amortisation, early-termination events, granularity — and they ask a single question of each one: does this feature give the originator a route to claw the exposure back? A call option that lets the bank repurchase the pool cheaply on a chosen date can hollow out a transfer that looked complete on paper. An excess-spread mechanism can absorb the losses the investor was supposedly buying, so that the investor’s slice never actually attaches. Replenishment rights let the bank swap fresh assets into the pool and change the risk the investor agreed to take. None of these features is a default, none of them is fraudulent, and none of them appears anywhere in the headline economics of the trade. They are simply terms — and terms are what determine how much exposure actually moved.

And because terms can be exercised over time, the assessment does not stop at closing. Supervisors review outstanding transactions on an ongoing basis, and where the transfer has degraded, the bank is asked to fix it. Significance is not a fact established once at issuance; it is a live quantity that decays as the seller uses the permissions the document gave it.

So in the bank market there exists a complete apparatus: a defined question (how much exposure moved?), a two-layer method for answering it (arithmetic plus a structural reading of the contract), a party obliged to perform the analysis, and a consequence attached to the answer. That apparatus is precisely what makes a bespoke, non-standard instrument financeable.

The tokenization market has rebuilt the instrument and left the apparatus behind

A tokenized real-world asset has exactly the structure described above. The reference asset stays where it is. The originator retains title and control. The investor receives economic exposure and a list of rights, by contract. The value therefore lives in the document — and yet the market prices these claims off the asset, using yield and loan-to-value, as though they were ownership securities like a bond or a REIT unit.

That substitution works for ownership securities precisely because their rights are standardized. Two bonds are comparable because bondholders get the same package; the issuer is the variable. Standardize the wrapper, add custody and settlement, and the thing trades. That is plumbing, and the tokenization industry built the plumbing very competently. But there is no standard synthetic risk transfer, no screen can show you the terms, and nobody in the pipeline is performing the significance assessment.

Take two tokenized buildings, both offered at an 8% yield and 60% loan-to-value on comparable real estate. On any interface you will ever look at, they are the same instrument.

In the first vehicle, the sponsor needs the consent of a majority of holders before it can sell the building, cannot incur new debt ranking ahead of you, cannot move the asset out to an affiliate, and your redemption right is a hard claim against the vehicle itself. Read as a risk transfer, this document moves a great deal of the building’s economic exposure to you and leaves the sponsor with very little ability to take it back. The transfer is significant.

In the second vehicle, the sponsor may borrow ahead of you at will, may transfer the building to an affiliate, may sell without asking you, and your redemption is a promise from the sponsor rather than a claim on anything. Every one of those permissions is a route by which the sponsor recaptures economics it appeared to have handed over — the exact analogue of the call options and excess-spread leakages a supervisor would flag in a bank deal. The headline says you own 8% on a 60% LTV building. The contract says you hold a thin, revocable strip of exposure that the sponsor can dilute at any time without ever committing a default. The transfer is insignificant, and the instrument is worth materially less.

Two loans can have the same borrower, the same maturity, the same leverage, and trade at the same yield, yet transfer fundamentally different amounts of economic risk because of the contractual rights embedded within them. Remarkably, while the market carefully measures credit risk, duration risk, and liquidity risk, it still has no framework for measuring the quantity of contractual risk that is actually being transferred.

What the absence costs

The gap closes both sides of the market at once. In the secondary market, a holder cannot mark, compare, or borrow against a claim whose transferred exposure nobody has measured, so it does not move. In the primary market, a sponsor offering genuinely significant transfer has no credible way to distinguish itself from a sponsor offering a hollow one, so investors apply the same discount to everybody, and the good sponsor withdraws rather than sell at a price set by the worst document in the market. That is a lemons market in textbook form, and it is self-reinforcing: as the strong issuers exit, the average quality of what remains falls, which justifies a deeper discount.

The same gap disqualifies these claims as collateral. Lending against an asset requires five conditions to hold together — a reliable mark, a defensible haircut, a perfected security interest, a clear liquidation path, and enforceable close-out netting. Without a measured transfer, the first two fail immediately, and nothing downstream can be built on top.

What the work actually is

Tokenization made a record settle in seconds. It did not make one bespoke contract comparable to another, and no additional plumbing will, because the missing function was never plumbing.

The missing function is the significance assessment, performed for this asset class the way supervisors perform it for bank portfolios — but systematically, and by an independent party rather than a regulator. It means reading the operating agreement, identifying every mechanism by which the sponsor can recapture exposure it appears to have transferred, valuing the rights that stop it from doing so, and converting all of that into standardized inputs: a comparable grade, a spread in basis points, and a recovery estimate. And because significance decays as permissions get used, it has to be recomputed as the deal ages, not fixed at issuance.

That is what we are building at CovenantLab. We started in private credit, which is itself a synthetic risk transfer negotiated covenant by covenant and priced by the market off its headline rather than its terms. Tokenized real-world assets are the same analytical object with a different asset on the input side, and the gap there is wider, because in private credit at least the lenders read the documents.

Name the instrument correctly and the path is obvious. A tokenized real-world asset is not a share of a building. It is a claim on however much of that building’s risk the contract genuinely handed you — and until somebody measures that, the market cannot price anything at all.

Sources:

• Securitisations: meeting significant risk transfer criteria — ECB Banking Supervision

• EBA Report on Significant Risk Transfer in Securitisation

• Supervisory Statement 9/13, Securitisation: Significant Risk Transfer — Bank of England PRA

By Elham Saeidinezhad

The headlines were straightforward: Blue Owl permanently paused redemptions in one of its retail-focused private credit vehicles. To many observers, that signals fragility. But that conclusion overlooks the mechanics.

At roughly the same time the retail vehicle gated redemptions, Blue Owl sold approximately $1.4 billion of assets from OBDC 2 — an institutional vehicle on the same platform — at stable NAV, returning roughly 30% of capital to investors there.

In other words, one vehicle activated its redemption gate, while another converted loans into cash without a material price concession. The point to emphasize is that those are not conflicting outcomes. They are part of the same mechanism: engineering liquidity under stress within a single platform.

Taken together, they offer a clean case study in how liquidity is engineered in private credit — not through continuous market trading, but through portfolio construction, contractual flexibility, and controlled thresholds under stress.

Start With the Correct Lens

Private credit does not trade on a continuous market. Loans are typically bilateral or club instruments. Transfers require documentation, borrower consent, and negotiated settlement. Even when secondary markets exist, they are episodic and depth varies by credit.

The mistake many observers make is to analyze private credit liquidity events through a public-market lens. In public markets, liquidity is synonymous with immediacy. There is continuous price discovery, visible bid–ask spreads, and the ability to sell at a quoted price within seconds. Under stress, liquidity deteriorates through widening spreads and price gaps — but the core mechanism remains the same: assets clear through open trading. That framework shapes how risk is interpreted. When prices gap or trading slows, it signals fragility.

But private credit does not operate through that mechanism. There is no standing order book. No continuous two-sided market absorbing flow. No guarantee of instantaneous execution. What exists instead is negotiated transfer capacity, constrained by documentation and market depth.

Applying a public-market liquidity lens to private credit leads to a category error. A redemption gate in private credit is not equivalent to a market freeze in equities. It is the activation of a contractual threshold in a system that was never designed for continuous trading in the first place.

Investors are compensated for bearing that structural illiquidity through a spread premium. That premium is priced in yield. What is often misunderstood is that liquidity in private credit is not something that disappears under stress — it is something that must be exercised within predefined structural limits.

When a fund offers quarterly redemptions against inherently illiquid assets, it cannot rely on continuous secondary trading to meet withdrawals. Liquidity must be anticipated, structured, and internally generated. That deliberate construction of exit capacity — within legal, portfolio, and market constraints — is engineered liquidity.

The Two Pillars

Engineered liquidity rests on two pillars. Both must hold simultaneously.

Pillar 1: Portfolio Construction

Figure 01- Portfolio construction in a nutshell

A manager must deliberately construct a portfolio that embeds exit capacity. That means diversification across borrowers and sectors, staggered maturities that generate predictable cash inflows, limited exposure to structurally hard-to-transfer credits, and meaningful allocation to loans with active institutional secondary demand. Position sizing must also reflect expected market depth: a $200 million position behaves very differently from a $25 million one when liquidity is needed.

Technically, this requires balancing two variables against modeled redemption scenarios: expected cash inflows and realistic asset-sale capacity. Managers run internal stress tests — assume X% quarterly redemptions, estimate which assets can be monetized within Y days, and model whether execution can occur without exceeding a Z% price concession. The goal is not perfect liquidity; it is controlled liquidation capacity under stress.

When OBDC 2 sold $1.4 billion at stable NAV, it confirmed that such a buffer had been built. The selected assets likely had clean transfer documentation, credible secondary demand, and no embedded structural constraints. No material concession was required. That outcome reflects design and preparation — not luck.

This is the pillar the market sees, measures, and rewards.

Pillar 2: Covenant Design

Figure 02- Covenant design decisions in a nutshell

The second pillar is contractual — and far less visible. Even a well-constructed portfolio can become illiquid if the underlying agreements restrict movement.

Liquidity in private credit is governed by documentation — and each clause shapes how movable an asset truly is. For example:

• Assignment provisions may require borrower consent, which means a sale cannot occur unilaterally. If consent is slow, conditional, or strategic, timing risk is introduced exactly when liquidity is needed most.

• Minimum hold thresholds can prevent partial disposals. A manager may wish to trim a position to meet redemptions, but the agreement may require maintaining a minimum exposure, forcing either a full exit or no sale at all.

• Fund-level asset coverage or diversification tests may tighten after a disposition. In this case, selling one asset can mechanically worsen concentration ratios or regulatory tests, limiting how much capital can be raised without breaching structural limits.

• Cross-default or cross-collateralization clauses create linkage across positions. As a result, exiting one loan may alter protections or triggers elsewhere in the portfolio, introducing second-order consequences that constrain flexibility.

• Amendment thresholds determine how easily terms can be modified. If a transfer requires lender approval at a high voting threshold, speed and certainty decline under stress.

• Call protection and prepayment mechanics can determine economic viability. So, even if a loan can be transferred, penalties or make-whole provisions may render an exit unattractive or value-destructive.

Each of these features affects not whether an asset has value, but whether it can be converted into cash cleanly, quickly, and without destabilizing the broader structure

These clauses determine whether assets can be transferred cleanly, whether sale proceeds can be redeployed, and whether a transaction triggers structural constraints elsewhere in the portfolio. Selling one loan may tighten concentration limits on another. An inter-vehicle transfer may be prohibited. Asset coverage ratios may cap how much can be sold without breaching tests.

If these constraints bind, portfolio skill becomes secondary. A manager may identify a willing buyer and acceptable price — yet still be unable to execute without triggering violations, penalties, or cascading structural effects that undermine the objective.

That Blue Owl executed a $1.4 billion sale without visible disruption suggests that the documentation permitted genuine maneuverability. The contracts allowed assets to move without destabilizing the broader structure.

This second pillar is rarely priced explicitly. It does not show up in headline yield or marketing materials. It becomes visible only when tested — and under stress, it often determines whether engineered liquidity functions or stalls.

Figure 03 — Pillars 1 & 2 combined

The Mechanics Under Stress

Mechanically, when redemption requests rise:

1. The fund first evaluates available cash and predictable near-term inflows — scheduled interest, amortization, and maturities.

2. If those inflows are insufficient to meet withdrawals, the manager identifies candidate assets that could realistically be monetized.

3. Each candidate asset is then screened along three dimensions:

• Transferability under loan documentation (consent requirements, assignment restrictions).

• Impact on fund-level constraints (asset coverage tests, diversification limits, concentration thresholds).

• Expected execution price relative to carrying value.

4. The manager then assesses actual secondary depth — who the natural buyers are, how much size they can absorb, and what price impact is likely under current conditions.

5. If pricing is within tolerance and no structural breach is triggered, the asset is sold and liquidity is generated.

6. If pricing becomes punitive or structural constraints bind — meaning the sale would destroy NAV or trigger contractual violations — redemption gates activate.

The retail vehicle’s redemption pause reflects that final step: a predefined protection threshold being reached. The institutional sale reflects successful clearance of every step before it.

Figure 04- Gate Decisions on A: Contractual Clearance & B: Pricing Acceptability

In the Blue Owl’s case, both outcomes operated as intended.

Figure 05 - Terminal Outcomes

What the Market Prices, Praises, and Ignores

The market prices illiquidity. Spread premiums exist precisely because loans are not continuously tradable and cannot be exited on demand.

The market also rewards portfolio skill. Managers who execute asset sales near carrying value, preserve NAV, and navigate stress without visible disruption earn credibility — and capital.

What is neither cleanly priced nor systematically analyzed is contractual design: the structural elasticity embedded in loan agreements and fund documents that determines how flexible the architecture truly is when pressure rises. It does not show up in headline yield. It is not disclosed in marketing materials. It rarely appears in quarterly reports.

It becomes visible only when tested. And under stress, it is often the variable that decides the outcome.

The Broader Implication

Private credit is illiquid. That is known. That is priced. The real question is not whether liquidity exists — it is how much of it can be manufactured when pressure arrives, and at what economic cost.

Engineered liquidity only works when both pillars operate simultaneously: portfolio construction must embed genuine exit capacity, and covenant design must allow that capacity to be exercised without triggering structural damage. Portfolio skill without contractual flexibility stalls. Contractual flexibility without portfolio depth is useless.

Blue Owl’s actions illustrate this in real time. The sale at stable NAV reflects disciplined portfolio construction. The absence of structural disruption reflects contractual maneuverability. The redemption pause reflects the activation of protective thresholds designed to prevent forced liquidation from eroding value.

This is not a revelation that private credit is illiquid. It is a reminder that liquidity in private markets is not discovered — it is designed. And the least visible input in that design, contractual elasticity, may ultimately be the most decisive variable in stress.

Figure 06- How a private credit manager navigates redemption pressure — step by step

A Note from the Author:

The idea for this analysis stems from my role as moderator for “Panel 3: Patient Capital in Impatient Times: Private Debt & Private Equity” at Fordham University Gabelli School of Business in December 2025. This discussion directly informs and is grounded in the extensive research I am undertaking with my co-author, Uri Ravid, focusing on private credit valuation and liquidity measurements. Select publications detailing our work in this area are available here:

• Saeidinezhad, Elham, and Ravid, Uri, The Covenant Liquidity Index: A Framework for Rating Private Credit Liquidity (October 19, 2025). Available at SSRN: https://ssrn.com/abstract=5621330 or http://dx.doi.org/10.2139/ssrn.5621330

• Saeidinezhad, Elham, The Covenant Structure of Yield in Private Credit, (August 01, 2025). Available at SSRN: https://ssrn.com/abstract=5375808 or http://dx.doi.org/10.2139/ssrn.5375808

I. Introduction: The Wrapper Problem in Private Credit

The rapid expansion of private credit has outpaced the financial infrastructure designed to intermediate it. As private lending has moved beyond traditional bank balance sheets into asset managers, institutional portfolios, and retail-facing vehicles, increasing attention has been paid to a central structural question: how can inherently illiquid credit assets be housed within public or semi-public investment structures without destabilizing either the assets or the vehicle itself?

Over the past decade, multiple attempts have been made to “wrap” private credit in structures originally designed for liquid markets. Some of these efforts—notably Business Development Companies (BDCs)—have proven durable and scalable. Others—most prominently exchange-traded fund (ETF)–based approaches to private credit—have struggled to gain traction or have failed outright. This divergence is often attributed to liquidity mismatch, regulatory constraints, or operational frictions. While these factors are relevant, they do not fully explain why one structure has adapted successfully to private credit while the other has not.

This paper argues that the decisive difference lies deeper, at the level of financial engineering: in the valuation regime embedded at the core of each structure.

ETFs are engineered to eliminate discrepancies between market price and net asset value. NAV is calculated daily and treated as a clearing price, while creation and redemption mechanisms allow authorized participants to arbitrage deviations. Price–NAV gaps are therefore interpreted as transient distortions to be eliminated. This architecture presupposes observable prices, tradable collateral, and continuous liquidity in the underlying assets, since liquidity must exist at the asset level in order for convergence to be enforced. In this sense, the ETF valuation regime is built around what may be termed NAV parity. The central function of the ETF market maker is precisely to compress price–NAV deviations and maintain this regime.

Business Development Companies adopt the opposite logic. Net asset value is disclosed periodically and functions as an accounting anchor rather than a clearing price, while market prices are permitted to trade at persistent discounts or premiums. From the perspective of BDC managers, such deviations are not treated as distortions, but as informative outcomes that absorb uncertainty related to exit timing, valuation discretion, covenant structure, and credit-cycle positioning. Liquidity is not imposed on the underlying assets; it is deliberately relocated to the equity layer and priced through discounting. The BDC valuation regime is therefore based on what may be termed NAV realism. The role of the BDC manager is not to eliminate valuation dispersion, but to manage the portfolio through it and price liquidity accordingly.

The core claim of this paper is simple but consequential. BDCs succeeded because they institutionalize NAV realism—persistent valuation dispersion aligned with private-market cash flows and exit timing—while private credit ETFs failed because they attempt to enforce NAV parity through public-market mechanisms incompatible with illiquid, model-valued assets. By integrating practitioner-facing valuation metrics—NAV construction, net investment income, distributions, portfolio quality, non-accruals, and interest-rate sensitivity—into this conceptual framework, the paper explains why BDCs have functioned as a durable liquidity wrapper for private credit. The analysis then extends beyond wrapper design to the future of private credit market making, where the same choice between public-market parity and private-market realism will determine which intermediation models ultimately endure.

In other words, in private credit, the market maker is not an arbitrageur of prices, but an arbiter of contracts—continuously judging how differences in covenant strength, control rights, collateral, and exit paths should be priced. That judgment, exercised across time rather than through instantaneous trading, is the market maker’s contribution. This is what it means to engineer liquidity rather than attempt to manufacture it.

Table 1- ETF Arbitrage Market Making vs. BDC-Style Balance-Sheet Intermediation

II. Two Valuation Regimes: Public-Market Parity vs. Private-Market Realism

The divergence between Business Development Companies and private credit ETFs reflects not differences in asset quality or investor demand, but a fundamental difference in valuation regime. Each structure embeds a distinct conception of how prices, net asset value, and liquidity are meant to interact. In public markets, valuation is organized around continuous price discovery, with market makers tasked with enforcing arbitrage-based convergence. In private markets, by contrast, valuation is episodic and—even in the presence of market makers—contractually mediated and judgment-based. In such an environment, liquidity wrappers must be designed to tolerate persistent disagreement rather than eliminate it. Whether a wrapper succeeds or fails in private credit therefore follows directly from the valuation regime it attempts to impose.

Crucially, private-market valuation is not merely less frequent than public-market valuation; it is structurally different. Value is shaped not only by expected cash flows, but by the design of contractual protections—seniority, covenants, call protections, and exit rights—that govern downside risk and determine the timing and form of realizations. These features do not clear continuously through markets; they operate discretely through legal and contractual triggers. Any valuation regime intended to intermediate private credit must therefore accommodate this contractual architecture rather than override it.

A. ETFs and the Logic of Continuous NAV Parity

Exchange-traded funds are engineered to maintain continuous parity between market price and net asset value. This objective is implemented through a tightly coupled system of daily NAV calculation, creation and redemption mechanisms, and authorized participant arbitrage. When an ETF trades at a premium or discount to NAV, authorized participants are incentivized to create or redeem shares using the underlying assets, thereby enforcing convergence. Deviations from NAV are therefore treated as transient inefficiencies to be eliminated through trading.

This design presupposes a specific set of conditions. First, the underlying assets must have observable prices that can be incorporated reliably into frequent NAV calculations. Second, those assets must be sufficiently liquid to be assembled, delivered, or liquidated in size as part of the creation and redemption process. Third, arbitrage must be economically and operationally feasible on short horizons. When these conditions hold—as they generally do in public equity and government bond markets—NAV parity stabilizes prices and enhances liquidity.

In private credit, however, these conditions do not hold. Net asset value is inherently model-based, reflecting discounted cash flows, credit assessments, and long-horizon expectations rather than executable market prices. More importantly, valuation depends critically on contractual features—covenant protections, collateral packages, amortization schedules, and negotiated exit rights—that shape cash-flow priority and loss severity but cannot be traded or arbitraged in real time. The underlying loans are not continuously tradable, cannot be delivered into an arbitrage mechanism, and frequently lack a meaningful secondary market. As a result, the creation and redemption process cannot operate as designed. When price and NAV diverge, there is no effective mechanism through which arbitrage can enforce convergence.

Crucially, this breakdown is mechanical rather than operational. The failure of NAV parity in private credit ETFs does not stem from poor execution, insufficient market making, or temporary illiquidity. It reflects a deeper incompatibility between a public-market valuation regime and assets whose value is contractually defined and realized through discrete legal and economic events. The ETF structure treats persistent deviation from NAV as a malfunction to be corrected. In private credit, such deviation is not a malfunction but an unavoidable consequence of covenant-based valuation. The result is not temporary dislocation, but structural fragility.

B. BDCs and the Logic of NAV Realism

Business Development Companies adopt a fundamentally different valuation logic. Rather than engineering continuous convergence between market price and net asset value, BDCs are explicitly designed to tolerate—and indeed institutionalize—persistent valuation dispersion. Net asset value is calculated and disclosed periodically, typically on a quarterly basis, and functions as an accounting reference point rather than a clearing price. Market prices are permitted to trade at sustained discounts or premiums without triggering mechanical correction.

This tolerance is not incidental; it is central to the structure. The price–NAV gap in a BDC absorbs precisely those dimensions of private credit that cannot be arbitraged away: uncertainty surrounding borrower outcomes, discretion in valuation methodologies, exit timing risk, and exposure to the credit cycle. Crucially, it also reflects differences in contractual and covenant design—such as seniority, collateral coverage, call protections, and renegotiation rights—that shape downside risk and recovery but cannot be continuously priced or traded. Rather than forcing liquidity onto the underlying assets, BDCs relocate liquidity to the equity layer and allow it to be priced explicitly through discounting.

Under this regime, NAV serves as an anchor rather than an equilibrium. It conveys information about the estimated fundamental value of the portfolio while acknowledging that such estimates are necessarily incomplete and contestable. Market prices therefore reflect not only expected cash flows, but also confidence in management, conservatism of valuation marks, portfolio construction, and the credibility of exit engineering embedded in covenant structures. Persistent discounts are not interpreted as failures of arbitrage, but as expressions of valuation realism.

This distinction has direct implications for how liquidity is provided. In the ETF regime, liquidity is presumed to arise from continuous trading and arbitrage at the asset level. In the BDC regime, liquidity is episodic and indirect. Investors obtain liquidity by trading claims on the portfolio, not by trading the portfolio itself. Valuation dispersion is thus not a defect of the structure, but the mechanism through which contractually defined private-market uncertainty is priced in a public setting.

The contrast between these two regimes clarifies why BDCs have adapted successfully to private credit while ETFs have not. ETFs attempt to impose public-market discipline on assets whose value is legally and contractually structured, treating valuation disagreement as an error to be eliminated. BDCs embed private-market valuation logic within a public wrapper, allowing disagreement to persist and be priced. The difference is not one of regulatory detail or investor sophistication, but of financial engineering at the most fundamental level.

III. How BDC Valuation Actually Works (And Why It Fits Private Credit)

Having distinguished between public-market valuation based on enforced NAV parity and private-market valuation based on tolerated dispersion, this section examines how Business Development Companies are valued in practice. The central point is not simply that BDCs rely on a different set of metrics, but that each metric operates coherently within a regime of NAV realism.

Under this regime, valuation is anchored by four interlocking elements, with NAV serving only as one of these pillars:

1. Net Asset Value,

2. Net Investment Income,

3. Distributions, and

4. Portfolio Management Quality.

Each anchor plays a distinct role, but none is designed to function in isolation or to clear markets continuously.

Importantly, these anchors do not operate as independent signals. They function as a system, jointly anchoring expectations in an environment where valuation is discrete, judgment-based, and contractually structured. Together, they form a valuation architecture suited to private credit—one that accommodates uncertainty, respects time, and prices liquidity rather than presuming it.

A. Net Asset Value (NAV) as an Accounting Anchor

Net asset value per share remains the primary reference point in BDC valuation. It represents the fair value of the investment portfolio net of liabilities and serves as the accounting benchmark against which market prices are interpreted. Recent disclosures illustrate this anchoring role. Golub Capital, for example, reported a NAV per share of $14.97 as of September 30, 2025, while other large BDCs—such as Blackstone Secured Lending Fund—report NAV on a quarterly basis using independent valuation inputs.

Crucially, NAV in a BDC is not intended to function as a continuously tradable price. Portfolio assets are valued using fair-value methodologies that incorporate discounted cash flow analysis, credit assessments, and long-horizon assumptions about borrower performance and exit outcomes. These valuations are necessarily model-based and updated episodically. NAV is therefore informative but incomplete by design: it provides an internally consistent estimate of portfolio value while acknowledging that such estimates cannot be instantaneously validated through market transactions.

Within a regime of NAV realism, this incompleteness is not a weakness but a defining feature. It creates space for market prices to reflect disagreement over valuation assumptions, including differences in contractual terms and covenant structures. Persistent discounts or premiums to NAV are not treated as valuation errors, but as the market’s assessment of uncertainty, exit risk, and the credibility of the valuation process itself.

B. Net Investment Income (NII) as an Earnings Anchor

If NAV anchors valuation, net investment income anchors confidence. NII measures the recurring income generated by the portfolio after operating expenses and financing costs. Unlike mark-to-market changes in asset values, NII reflects contractual cash flows—interest payments, fees, and amortization—embedded in private credit instruments and governed by legally enforceable agreements.

Recent performance highlights the centrality of this metric. In the third quarter of 2025, Blackstone Secured Lending Fund generated net investment income of $0.82 per share, corresponding to an annualized return on equity of approximately 12 percent. Golub Capital reported adjusted NII of $0.39 per share, translating into an adjusted NII ROE of 10.4 percent, while Ares Capital Corporation reported core earnings of $0.50 per share.

NII provides a counterweight to valuation discretion. Even when NAV is discounted by the market, sustained net investment income demonstrates the portfolio’s ability to generate cash flows under existing contractual terms. In private credit—where exits are episodic and asset sales are infrequent—this income-based validation stabilizes expectations by tying valuation to realized performance rather than hypothetical liquidity.

C. Distributions as the Liquidity-to-Investor Anchor

Distributions translate accounting measures and income generation into investor experience. To maintain their tax-advantaged status, BDCs are required to distribute at least 90 percent of taxable income, making distribution yield and dividend coverage central components of valuation rather than residual considerations.

Recent data underscore this role. Blackstone Secured Lending Fund maintained a quarterly dividend of $0.77 per share in the third quarter of 2025, corresponding to an annualized yield of approximately 11.3 percent. Golub Capital distributed $1.65 per share during fiscal year 2025, representing roughly 10.9 percent of its end-of-year NAV per share. In both cases, distributions were supported by net investment income rather than asset sales.

Within a NAV realism regime, elevated yields are not signals of hidden volatility. They are compensation for structured illiquidity. Capital is committed over long horizons, exits are discretionary rather than market-driven, and liquidity is obtained through secondary trading of BDC equity rather than liquidation of underlying loans. Dividend coverage therefore serves as a credibility test of the valuation system as a whole.

D. Portfolio Management and Quality as the Disciplining Anchor

The final anchor of the NAV realism regime is portfolio management quality. The ability to incorporate private assets—through structures such as BDCs—into the portfolios of patient investors depends critically on disciplined underwriting, covenant design, and active portfolio management rather than continuous price validation.

Tolerance for persistent price–NAV dispersion does not imply indifference to asset quality or valuation rigor. On the contrary, NAV realism requires discipline precisely because valuation is not continuously validated through market trading. Credibility must be earned over time through observable portfolio outcomes. Portfolio quality metrics—non-accruals, covenant performance, recoveries, and realized exits—perform this disciplining function.

Rather than discovering prices in real time, these metrics constrain valuation discretion ex post. Managers that underwrite conservatively and preserve covenant protection are rewarded with tighter discounts or sustained premiums. Those that rely on aggressive assumptions face widening discounts even when reported NAV remains stable. Valuation realism, in this sense, is not permissive; it is reputational.

E. Valuation Metrics as a System, Not a Checklist

Taken together, net asset value, net investment income, distributions, and portfolio quality form a coherent valuation system rather than a checklist of independent indicators. Net asset value provides a long-horizon estimate of portfolio worth; net investment income validates that estimate through realized, contractually generated cash flows; distributions translate income into observable investor returns; and portfolio quality disciplines valuation over time through realized outcomes.

This architecture fits private credit because it aligns with the asset class’s defining characteristics: contractual cash flows, covenant-based risk control, episodic exits, and limited secondary liquidity. BDC valuation works not by suppressing uncertainty, but by structuring how uncertainty is absorbed, priced, and ultimately resolved across time.

F. Preparing the Ground for Market Making

The implications of this valuation architecture extend beyond investment vehicles to the provision of liquidity itself. If valuation evolves discretely and income adjusts contractually, market making in private credit cannot be organized around continuous clearing prices. Liquidity provision must instead operate across valuation horizons, with inventory held through macro cycles and priced accordingly.

BDCs demonstrate that such an approach is viable. They show that liquidity can be provided without requiring assets to trade and that valuation credibility can be sustained without continuous market validation. As private credit markets deepen and secondary activity expands, the central question is whether new intermediaries will respect this temporal structure—or attempt to override it.

This question becomes particularly salient as large private credit managers increasingly seek to intermediate secondary transactions and act as market makers. Whether these efforts succeed will depend on whether they adopt a valuation regime consistent with commitment pacing, or revert to public-market assumptions ill suited to private assets.

IV. From Wrapper Design to Market Making: The Apollo Question

The distinction between NAV parity and NAV realism does not end with wrapper design. As private credit markets mature and secondary activity expands, the same valuation logic will determine the success or failure of market-making efforts. The analytical shift—from designing investment vehicles to intermediating transactions—places large private credit managers at the center of the next phase of financial engineering. Among them, Apollo serves as a natural stress test.

A. Apollo as a Natural Stress Test

Apollo occupies a distinctive position within private credit markets. It has been a leading architect of private credit as an asset class, a sponsor of retail-facing structures, and an increasingly active participant in secondary transactions and balance-sheet intermediation. Importantly, Apollo has operated on both sides of the wrapper debate: it has explored ETF-style vehicles for private credit exposure while simultaneously expanding its role as a long-horizon capital provider and secondary buyer.

This dual positioning makes Apollo analytically useful. The question is not whether Apollo will participate in private credit market making—its scale, origination footprint, and balance-sheet capacity make that outcome highly likely. The relevant question is how it will intermediate liquidity. That distinction turns not on technology, distribution, or branding, but on valuation philosophy.

B. Two Paradigms for Private Credit Market Making

There are two conceptually distinct approaches to market making in private credit, each corresponding directly to the valuation regimes developed earlier.

Under the first approach, market making is modeled on public-market arbitrage. Prices are expected to update continuously, bid–ask spreads are compressed, and inventory is held briefly. Liquidity provision is organized around enforcing convergence to a reference value—implicitly, a continuously updated NAV. Valuation dispersion is treated as a problem to be eliminated, and liquidity is equated with trading frequency. Applied to private credit, this approach reproduces the assumptions underlying ETF-based wrappers and inherits their structural fragilities.

Under the second approach, market making is understood as balance-sheet intermediation across valuation horizons. Inventory is held for extended periods, pricing reflects exit optionality and uncertainty, and liquidity is provided episodically rather than continuously. Valuation dispersion is not eliminated; it is monetized. Prices clear not because they converge to NAV, but because counterparties are willing to transact at discounts or premiums that compensate for timing, uncertainty, and contractual rigidity. This paradigm extends the BDC logic from wrapper design into secondary intermediation.

The distinction between these paradigms is not cosmetic. In the first, market making attempts to impose public-market discipline on private assets. In the second, market making is adapted to the economic reality of private credit, where valuation is discrete, exits are negotiated, and liquidity is contractually constrained.

C. Market Making as Liquidity Engineering

Seen through this lens, successful private credit market making is less about price discovery than about liquidity engineering. The role of the intermediary is not to reveal a single, continuously valid price, but to warehouse risk across time, structure exits, and intermediate between agents with heterogeneous liquidity preferences. Balance-sheet capacity, underwriting expertise, and tolerance for valuation uncertainty become central capabilities.

BDCs provide a proof of concept for this model. They demonstrate that liquidity can be provided without forcing assets to trade and that valuation credibility can be sustained without continuous convergence. Extending this logic to market making implies accepting wider spreads, longer inventory horizons, and persistent valuation disagreement as structural features rather than inefficiencies.

Apollo’s market-making ambitions therefore hinge on a regime choice. If private credit is treated as a public market with insufficient plumbing, market making will gravitate toward arbitrage and parity enforcement, reproducing the ETF problem in a different institutional form. If private credit is treated as a private market requiring engineered liquidity, market making can evolve as an extension of balance-sheet intermediation rather than a substitute for it.

D. Implications for the Next Phase of Private Credit

This distinction matters not only for individual firms, but for the architecture of private credit markets more broadly. A parity-based approach to market making would compress spreads, increase correlation, and amplify stress during downturns. A realism-based approach would accept slower adjustment, wider dispersion, and differentiated pricing as the cost of stability.

The evolution of private credit market making will therefore reveal whether the industry has internalized the lessons of wrapper design. The same valuation regime that allowed BDCs to succeed—tolerance for disagreement, discipline through realized outcomes, and liquidity that is priced rather than imposed—will determine whether secondary intermediation deepens private credit markets or destabilizes them.

Table 2:What Apollo Risks Under a Parity-Based Regime

V. Conclusion: Even in the Presence of Market Makers, Liquidity Is Engineered, Not Discovered

The contrast between Business Development Companies and private credit ETFs ultimately reflects a choice about valuation architecture, not liquidity provision. BDCs succeeded because they institutionalize NAV realism: they permit valuation disagreement to persist, price liquidity explicitly through discounts, and align public claims with private-market cash flows and negotiated exit timing. Private credit ETFs failed because they attempted to enforce NAV parity through public-market arbitrage mechanisms that cannot operate on illiquid, model-valued assets.

As private credit markets now turn toward secondary trading and market making, the same distinction will determine their evolution. Liquidity in private credit cannot be discovered through continuous pricing or enforced convergence. It must be engineered—through balance-sheet intermediation, tolerance for valuation dispersion, and commitment pacing across time.

Whether large intermediaries such as Apollo extend the BDC logic into market making—or instead reproduce the ETF mistake under a different institutional form—will shape the next generation of private credit infrastructure.

Appendix:

How to read figure 1?

A Business Development Company (BDC) functions as a crucial liquidity-relocation and valuation-realism mechanism that bridges the retail public market and the illiquid private credit market. As illustrated in Figure 1, the BDC structure is composed of three interconnected layers. At the foundation is the Private Credit Portfolio, which consists of illiquid assets valued using model-based contractual cash flows, where liquidity is inherently scarce and exits are episodic. This portfolio is intermediated by the BDC Vehicle layer, which provides periodic (typically quarterly) Net Asset Value (NAV) calculation using fair-value accounting, applying discipline through covenant and seniority structures. Finally, the BDC issues Publicly Traded BDC Shares at the top layer, providing investors with a continuously tradable entry and exit point via a secondary market. This structure effectively relocates liquidity from the illiquid private assets up to the public equity shares, allowing the market to impose valuation realism through the persistent discount or premium of the share price to the accounting NAV.

October 19, 2025

Next week, as we convene for the Market Microstructure Symposium 2.0 at NYU Stern, I want to share my latest paper:

“Rating Private Credit Liquidity and Secondary Market Discounts.”

Read the full paper on SSRN

The paper introduces the Covenant Liquidity Index (CLI) — a framework for scoring how liquidity in private credit is designed, structured, and ultimately priced.

It’s the first step toward building a systematic measure of how time itself is engineered through contract.

As with all my research, this project is driven by dialogue with practitioners — by observing how liquidity actually behaves in markets, and how theory might translate that behavior into structure.

I look forward to continuing that conversation next week, as we explore how liquidity in private credit is no longer found in markets, but written directly into covenants.

Liquidity as a Designed Property

Private credit has grown larger than the leveraged loan and high-yield bond markets combined, yet it still lacks a common language for one of its most essential properties: liquidity.

Unlike public markets — where liquidity can be observed through spreads, volumes, or market depth — private credit trades episodically and is priced by appraisal.

Even so, investors and secondary buyers routinely assign different values to comparable assets, implicitly discounting for the time and certainty of repayment.

The question is not whether liquidity matters, but how it can be measured when markets are missing.

This new paper develops a structural answer.

It builds on my earlier work, The Covenant Structure of Private Credit Yield (2025), which argued that liquidity in private markets is not discovered through trading — it is manufactured through contract design.

Here, that insight becomes measurable through the Covenant Liquidity Index (CLI): a standardized scoring system that translates covenant architecture into a liquidity rating.

The CLI links the structure of contractual control — how liquidity is designed — to observed secondary-market discounts, showing how value evolves not by price discovery but by rule design.

The Architecture of Liquidity

Each covenant in a financing agreement — call protection, make-whole provisions, PIK toggles, NAV triggers, or distribution lock-ups — acts as a timing valve that determines when and how capital can move.

Together, these clauses form a contractually defined liquidity structure, a legal and temporal design that governs the flow of value across time.

The CLI quantifies this structure along three interacting dimensions:

• Exit Engineering sets the boundary of liquidity realization — when and under what conditions capital can be returned or recycled.

• Cash-Flow Pacing governs the velocity of liquidity, translating performance into repayment through rules such as mandatory sweeps and restricted-payment tests.

• Optionality Structuring defines the curvature of liquidity — the extent to which timing can flex without breaking pacing discipline.

When these elements are aligned, liquidity behaves predictably and moves through time as a designed process.

When they conflict, liquidity becomes path-dependent, driven by negotiation rather than structure.

From Covenant Design to Pricing Logic

The CLI aggregates these structural dimensions into a standardized score from 1 to 5 — mapped into an intuitive letter scale from CCC to AAA.

High-CLI instruments — with deterministic exits, automatic pacing, and bounded flexibility — show shorter, more stable time-to-liquidity and lower required discount rates.

Low-CLI structures rely on discretion and opportunism, producing deeper discounts and greater valuation dispersion.

In this framing, the secondary-market discount rate is the price of engineered time, and the CLI measures how much of that time is already designed ex ante.

This relationship gives rise to what I call the Contractual Liquidity Curve — a private-market analogue to the yield curve.

In public debt, the yield curve maps the price of time through rates; in private credit, the covenant curve maps the price of time through design.

High-CLI assets cluster near par, reflecting rule-based liquidity.

Covenant-lite loans occupy the far end of the curve, where timing is uncertain and discount rates steep.

Why It Matters

As private-credit markets deepen, liquidity resides less in trading infrastructure and more in contract structure.

The CLI provides a framework for scoring, comparing, and ultimately pricing that structure — turning covenant analysis from a narrative exercise into a quantifiable architecture of liquidity.

It gives investors a standardized benchmark for assessing liquidity strength and provides economists a structural rationale for secondary-market pricing.

By linking the liquidity designed ex ante to the liquidity priced ex post, the CLI unifies two perspectives that have long remained separate.

It formalizes a simple but powerful idea: in modern private credit, liquidity is not discovered — it is engineered.

This theme will continue next week at the Market Microstructure Symposium 2.0 at NYU Stern, where our private-credit panels will explore how liquidity, once a market property, has become a contractual one.

I look forward to learning from the practitioners and researchers who continue to shape how markets now design their own time.

Key Takeaway: Market structure may still look dealer-centric, but the industrial organization of “dealers” has fragmented so much that market making now means radically different things across firms, with profound implications for research and regulation.

As we approach the Market Microstructure Symposium 2.0 on October 24 at NYU Stern, I’ll be using this space to highlight different panels from the program.

This week, we held the pre-panel call for Panel III — Market Making Without the Label: The Rise of Prop Traders, Transformation of Bank Dealers, and Hedge Fund Ambiguity. The discussion reminded me of a simple yet profound point:

Market microstructure cannot be understood without industrial organization — and the industrial organization of market makers has fragmented beyond recognition.

This panel will feature:

Sam Berberian (Citadel Securities)
Matthew Schrager (TD Securities)
Andreas Koukorinis (Optiver)
Lawrence Rascio (Olden Lane) — Moderator

Full program: https://lnkd.in/di2BAjUa

From Uniformity to Fragmentation

For decades, the market structure of many asset classes remained dealer-centric. Liquidity was provided by a relatively homogeneous group of bank dealers. That uniformity made stylized models such as Treynor’s classic framework(Treynor, 1987) and Glosten–Milgrom’s asymmetric information model (1985) at least semi-useful for describing how dealers absorbed order flow, warehoused risk, and set spreads to balance informed versus uninformed trading. Those days are long gone.

Even though markets are still formally organized around “dealers,” the industrial organization of liquidity providers looks nothing like it once did. Liquidity provision now spans systematic proprietary trading firms such as Citadel Securities or Optiver, which rely on predictive analytics, low-latency execution, and balance-sheet agility (Menkveld, 2013). Bank dealers like TD Securities remain central, but they are constrained by capital rules such as Basel III and the Volcker Rule, while supported by client relationships and integration with other business lines (Duffie, 2012). Hedge funds add another layer of complexity: strategically ambiguous, they step into dealer-like roles when spreads and volatility reward them, but retreat when obligations would be binding (Adrian & Shin, 2010). Sometimes, even within the same asset class, one firm’s notion of “market making” looks radically different from another’s.

Why This Matters

This fragmentation carries important implications for both research and regulation.

• For researchers, it raises a precision problem. When we study “the economics of market making,” which kind are we analyzing? A model that explains systematic liquidity provision may be meaningless for relationship-driven bank intermediation — and vice versa.

• For regulators, the stakes are even higher. The designation of “dealer” is no longer straightforward. Rules crafted for a homogeneous class of bank dealers now apply unevenly across actors who look nothing alike, yet all provide liquidity in some form.

This is the essence of the IO–microstructure interplay:

• Market structure remains formally dealer-centric.

• But the industrial organization of “dealers” is now heterogeneous enough that the microstructure itself operates differently depending on who is providing liquidity.

Rethinking the Dealer Concept

The old comfort of a single, uniform model of market making has disappeared. What we have instead is an ecosystem of liquidity practices: diverse, overlapping, and sometimes contradictory. To call all of them “dealers” obscures more than it clarifies, yet ignoring the differences leaves us blind to how volatility is absorbed, how risk is warehoused, and how market efficiency evolves.

This is exactly why Panel III excites me: it forces us to put these differences side by side, and to ask how they add up to a functioning — or fragile — market architecture.

Even though organizing this symposium is a lot of work, I love it precisely because it lets me learn as a student. Each panel is a window into how practitioners see their craft, and each call reshapes my own understanding of market plumbing. I am grateful that what began as a single event has now become an ongoing dialogue. With Symposium 3 already confirmed, we are building not just a program, but a sustained forum for examining how industrial organization and market microstructure together are reshaping financial stability.

To attend, please contact me at es5799@stern.nyu.edu, or reach out to co-organizers Kai Owen (kdo2116@columbia.edu) or Itai Dreifuss (id2349@columbia.edu).

Special thanks to our sponsors, the Department of Economics at NYU Stern and Nasdaq Exchange, especially Elizabeth Coogan and Nicholas Smith, for their support in making this event possible.

References

Adrian, T., & Shin, H. S. (2010). Liquidity and leverage. Journal of Financial Intermediation.

Duffie, D. (2012). Market Making Under the Proposed Volcker Rule. Rock Center for Corporate Governance.

Glosten, L., & Milgrom, P. (1985). Bid, ask and transaction prices in a specialist market with heterogeneously informed traders. Journal of Financial Economics.

Menkveld, A. J. (2013). High frequency trading and the new market makers. Journal of Financial Markets.

Treynor, J. (1987). The Economics of the Dealer Function. Financial Analysts Journal.

On October 24, 2025, we’ll gather at NYU Stern for the Market Microstructure Symposium 2.0.

This symposium is not a typical conference. It is an invitation-only, off-the-record forum (under the Chatham House Rule) where leading practitioners, policymakers, and academics will engage in candid discussions about how market plumbing and financial infrastructure shape financial stability.

Our aim is simple: to open up the “black box” of market structure, to look closely at the rules, institutions, and technical infrastructures that quietly determine how markets function—and how resilient they are in the face of stress.

We are grateful to the Economics Department at NYU Stern and to Nasdaq, especially Elizabeth Coogan and Nicholas Smith, for their sponsorship and support in making this possible.

Spotlight: Opening Panel on Private Market Secondaries

Yesterday we held our first pre-panel call for:

Panel I — Beyond Liquidity, Before Maturity: Rethinking Market Design in Private Secondaries

Panelists

• Brian A. Ruane (BNY)

• Lars Pace (Hamilton Lane)

• Gerald Cooper (Campbell Lutyens)

• Christoffer Tinsley, CFA, CAIA (Nasdaq)

• Dan Sullivan (PwC) — Moderator

Private secondaries have grown from a niche market to a central tool for institutional investors. Once thought of as a distressed-liquidity option, secondaries are now used by both LPs and GPs as strategic portfolio management instruments. Yet the basic infrastructure has not kept pace: transactions remain highly bespoke, documentation is slow, and settlement processes are manual.

As volumes grow and deal types expand into GP-led transactions, NAV-linked deals, and real asset exposures, the question becomes: can this market scale without building something closer to the standardized, infrastructure-driven model of public markets?

Topics We Will Explore

This panel will surface some of the most pressing challenges—and opportunities—facing the secondaries ecosystem:

• Can clearing and settlement processes evolve to handle growth, or will bottlenecks continue to drag on efficiency?

• Are manual processes and bespoke documentation the key bottleneck—making standardization the missing link in private market infrastructure?

• How do regulatory and tax frictions constrain who can transact and how quickly, and how do these differ from public markets?

• What do evergreen structures really offer: broader access for investors, or simply a shifting of liquidity pressures further downstream?

• How do covenant terms drive pricing and valuation, especially in real asset and infrastructure transactions, where NAVs are harder to mark and maturities are longer?

• A fundamental design question: is liquidity always desirable, or can illiquidity itself be a feature—creating the conditions for alpha?

These discussions connect directly to my recent paper, The Covenant Structure of Private Credit Yield, which develops a framework for understanding how liquidity in private markets is not infrastructural—it is contractual, embedded in the covenants of deals themselves. You can read the paper here:

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5375808

Why This Matters

What makes secondaries such a fascinating case study is that they sit at the intersection of private deal design and market infrastructure. Unlike public equities or ETFs, where liquidity is underwritten by exchanges, clearinghouses, and market makers, secondaries rely on contracts, covenants, and bespoke negotiation.

This raises fundamental questions:

• Can a covenant-driven market really scale to meet the needs of institutional investors?

• Or will the market need to adopt elements of public market infrastructure—centralized clearing, standardized documentation, or shared pricing benchmarks—to deliver efficiency at scale?

• And if private markets remain contract-driven, what risks does this create for stability, especially when real assets and longer-dated cash flow profiles are involved?

Why Attend

The Market Microstructure Symposium 2.0 is not a place for generic panels or high-level soundbites. It is a forum where:

• Experts can speak freely, without attribution.

• Practitioners and academics engage directly on technical questions.

• The architecture of markets—from private secondaries to credit platforms to ETFs—is treated as the real foundation of financial stability.

If you want to understand not just where markets are, but where they are going—and if you care about the infrastructure that underpins modern finance—this is the room to be in.

Full program:

https://smallpdf.com/file#s=b88f7583-f9e2-4d97-8c5d-f005b956aa13

If you are interested in attending, please email es5799@stern.nyu.edu. Space is limited, and we want the right people in the room.

— Market Microstructure Symposium 2.0 Organizers

Elham Saeidinezhad <es5799@stern.nyu.edu>

Itai Dreifuss <id2349@columbia.edu>

Kai Donovan Owen <kdo2116@columbia.edu>

As the Market Microstructure Symposium 2.0 approaches on October 24 at NYU Stern, I will be sharing regular spotlights on the panels that make up this year’s program. The symposium theme — How Infrastructure Shapes Financial Stability — runs through both the private and public market sessions. Each spotlight will outline the questions we aim to raise, the perspectives represented, and the linkages across markets that the discussions will surface.

Today’s spotlight is on Panel IV: From Options to Outcomes — Buffered ETFs and the New Architecture of Public Markets.

Why Buffered ETFs?

Buffered ETFs are often presented as straightforward, investor-friendly products: defined outcomes, partial downside protection, and capped upside. Their appeal lies in this apparent simplicity. Yet beneath the surface is a multi-layered process that connects investor demand, product structuring, and market execution.

Issuers translate investor requests into product concepts. Structurers then design payoff profiles that can be implemented within the constraints of listed options and clearinghouse capacity. Finally, trading desks ensure that these products remain liquid and functional under live market conditions.

Themes We Will Explore

At our pre-panel call, we outlined several themes that merit closer examination at the symposium. One is the coordination — and sometimes friction — among issuers, structurers, and trading desks as they translate investor demand into viable products. Another is how constraints in clearing, liquidity, and execution flow back upstream to shape what can credibly be offered in the first place.

We will also consider what the growth of buffered ETFs suggests about the evolving industrial organization of public markets. These products may be reshaping the role of infrastructure itself — not only as neutral plumbing, but as part of the design process.

Panelists

This discussion will bring together perspectives from across the intermediation chain:

• James Maund (Krane Funds Advisors) — Issuance

• Samuel Moore (Vest Financial) — Structuring

• Josh Khalif (Flow Traders) — Trading

• Elham Saeidinezhad (NYU Stern) — Moderator

Learn More

The full symposium program is available here: link.

If you are interested in joining us in the room, please contact me directly at es5799@stern.nyu.edu, or reach out to co-organizers Kai Owen Own (kdo2116@columbia.edu) or Itai Dreifuss (id2349@columbia.edu).

Today I’m happy to share the first draft of a paper that has been both intellectually challenging and deeply rewarding to write: The Covenant Structure of Private Credit Yield: A Contractual Theory of Valuation.

Read the full paper on SSRN

Introduction

In public markets, liquidity is infrastructure-centric. In private credit, it’s contract-derived.

This distinction isn’t cosmetic—it’s foundational. It shapes how capital moves, how yield is generated, and how valuation must be understood. My new paper, The Covenant Structure of Private Credit Yield: A Contractual Theory of Valuation, starts from this basic asymmetry and builds toward a new model of pricing: one in which yield is not discovered through markets, but structured through law.

The paper makes a simple claim with far-reaching consequences: in private credit, yield is not a market-clearing price—it is a contractual output. It is manufactured through a set of covenants that govern when capital may be deployed, recycled, or retrieved. The credit agreement does not merely protect return. It produces it.

This framing matters now because the private credit ecosystem is rapidly evolving beyond origination and hold-to-maturity. A new generation of secondary markets is emerging—GP-leds, LP-leds, credit secondaries—all aiming to trade exposures that were never meant to be liquid. The common approach has been to import the pricing heuristics of public markets: cash flow modeling, spread curves, peer comparables.

But this misses the core fact: liquidity in private credit doesn’t come from pricing infrastructure. It comes from contractual engineering. Exit isn’t found—it is authored. And if secondaries are to scale responsibly, we must first standardize the legal architecture that defines how—and whether—capital can move.

The Covenant-Centric Model of Private Credit Yield

That’s where this paper intervenes. It introduces a fourth valuation model—covenant-centric valuation—to stand alongside the Market, Income, and Replacement Cost approaches outlined in the IPEV Guidelines (2018). Traditional models rely on the assumption that liquidity either exists or can be proxied; they treat covenants as marginal frictions on cash flow. In contrast, the covenant-centric approach begins from the opposite premise: in private credit, covenants are the infrastructure. They determine whether capital can exit, under what conditions, and on whose terms. Yield, then, is not simply a spread over risk—it is the structured output of a liquidity schedule embedded in contract.

While this framework locates the source of value within the so-called “fundamentals” of the asset, it should not be mistaken for a traditional fundamentalist approach. Here, the fundamental is not the estimated cash flows or their timing, but the contractual fundamentals that govern shiftability—that is, the asset’s ability to move through a market structure. In this sense, the covenant-centric model is a hybrid valuation framework: it draws on the specificity of fundamental inputs (contract terms, exposure duration, pacing logic), but reinterprets those inputs through the lens of liquidity engineering and legal design. It treats market structure not as a background condition, but as a design problem to be solved contractually. Value is not discovered through pricing dynamics; it is authored through contract and revealed through structured liquidity.

To formalize this view, the paper introduces the contractual liquidity hierarchy: a three-layered model that classifies covenant mechanisms based on their liquidity function. Exit engineering provides legal pathways for capital withdrawal; optionality structuring governs who holds the decision rights over timing; and cash flow pacing sequences repayment across time. These covenant layers collectively shape the investor’s exposure path—not as an incidental constraint, but as the engineered architecture through which private yield is created.

Figure 1- The Contractual Liquidity Hierarchy

The Contractual Liquidity Hierarchy

1. Exit Engineering

At the apex of the pyramid, Exit Engineering contains mechanisms dictating how and when capital can be retrieved. Examples include:

• NAV-based resets, where exit is triggered if portfolio NAV falls below a threshold

• Take-out rights, where sponsors agree to refinance the lender under specified conditions

• Change-of-control clauses, where ownership shifts trigger repayment

• Mandatory cash sweep provisions, which accelerate principal return when excess cash exists

These instruments function as synthetic exits. They substitute for secondary sales or refinancing, creating engineered pathways for capital return.

2. Optionality Structuring

The middle layer governs who controls exit timing and on what terms. For example:

• Call protections prevent borrowers from repaying early, preserving contractual duration

• Prepayment premiums ensure compensation if borrowers terminate early

• Equity kickers give lenders upside in exchange for bearing illiquidity

• Performance ratchets shift pricing based on borrower milestones

These terms embed liquidity asymmetry—only one party controls the exit, and the other must price in that constraint.

3. Cash Flow Pacing

Forming the base, Cash Flow Pacing gathers instruments that control the sequences of repayment in the absence of tradability. It includes:

• PIK toggles, which let borrowers defer interest into principal

• Amortization schedules, which pre-structure capital return over time

• Step-up/step-down coupons, which align return timing with performance or time horizon

• Reinvestment restrictions, which control how returned cash can be redeployed

These covenants define the liquidity timeline, not in response to markets, but as a legal rhythm embedded in the deal.

Together, these three covenant layers are not just risk mitigants—they are functional liquidity modules. They form a hierarchy of covenant liquidity that defines the investor’s exposure path. They determine the capital timeline that, in public markets, would be shaped by hedging, trading, or refinancing. Here, it’s done by design.

From Fund Pacing to Deal Microarchitecture

This contractual hierarchy does not emerge in a vacuum. To understand how these covenant layers take shape, the paper extends Thomas Meyer’s theory of commitment pacing (Wiley Finance, 2024)—a fund-level framework through which limited partners (LPs) manage illiquid exposure over time by strategically sequencing capital commitments, distributions, and reinvestments across vintages. At that level, commitment pacing is a design solution for managing liquidity constraints, achieving target NAV exposure, and smoothing return volatility.

What this paper contributes is a bridge from that macro-level architecture to the micro-level mechanics of the deal. The central insight is that covenants are not merely negotiated protections against borrower behavior—they are instruments that translate portfolio-level pacing preferences into contract form. Underwriters and lenders do not simply react to borrower fundamentals; they actively shape capital flow through clauses that reflect their own exposure timing preferences, liquidity requirements, and reinvestment logic.

In this sense, the credit agreement becomes a microarchitecture of capital pacing. Where the fund-level model governs how exposure is layered across time, the covenant structure determines how it is sequenced within a single deal. Tools like delayed funding clauses, PIK toggles, reinvestment restrictions, and take-out rights are not just technicalities—they are engineered modules designed to map portfolio strategy onto deal structure. The contract becomes the site where liquidity is not only constrained—but constructed.

Figure 2- Fund and Deal-Level Finance Features

If a lender requires liquidity protection, the credit agreement may include delayed funding provisions or NAV-based reset clauses that trigger early repayment under defined stress conditions. If they aim to extend exposure, call protections or PIK toggles preserve duration. If they’re managing reinvestment pacing, cash sweep redeployment rights or reinvestment restrictions are used to time the recycling of capital.

Each of these covenants operates as a temporal module—an instrument for sequencing capital. They do not merely manage risk; they choreograph return. When assembled into a deal, they form a structured exposure path tailored to the investor’s liquidity profile. The contract does not simply mirror the liquidity environment of public markets—it replaces it with its own internal scaffolding: contractual liquidity.

To illustrating how covenant design encodes liquidity pacing—drawn from common structures in mid-market direct lending deals, consider a $150 million unitranche loan provided by a direct lending fund to a sponsor-backed healthcare services company. The lender—an insurance-backed private credit platform—has known reinvestment pacing constraints: it prefers a 5–7 year exposure, needs semi-annual cash flow for liability matching, and cannot tolerate uncertain capital return near the end of the term.

To align the deal with these pacing needs, the credit agreement includes the following covenants:

1. Staged Drawdowns: Instead of deploying the full $150 million upfront, the contract specifies a $75 million initial draw, with the remaining capital available in two delayed tranches based on revenue milestones. This gives the lender control over capital deployment timing and limits cash drag.

2. NAV-Based Reset Clause: If the borrower’s reported enterprise value (via an agreed-upon valuation agent) drops below a 7.5× EBITDA threshold, the lender gains the right to accelerate repayment or force a refinancing. This serves as a synthetic secondary: a contractual exit trigger in lieu of a liquid market.

3. Call Protection (Soft Call Schedule): For the first two years, the borrower must pay a premium to exit early—3% in year one, 2% in year two. This preserves the lender’s planned duration and protects against premature recycling.

4. PIK Toggle: The borrower can defer up to 50% of interest payments for four quarters, converting them into principal. This creates temporal flexibility, allowing the lender to remain exposed without needing interim cash—useful if they are managing quarterly liability peaks.

5. Cash Sweep with Reinvestment Restriction: Any excess free cash flow (above a negotiated capex threshold) must be swept to repay the loan. However, if the lender is still within their reinvestment window (defined as years 2–4), they may re-deploy swept capital into a related acquisition facility. After year 4, the reinvestment right expires.

In this example, the investor’s exposure timeline has been codified in contract. There is no assumption of refinancing, no reliance on an active secondary market, and no pricing of embedded options via observable volatility. Instead, the deal engineers its own liquidity through the careful design of covenants—each chosen to reflect the pacing logic of the capital provider.

This is contractual liquidity, and this is where yield is formed—not in a clearing price, but in a legal structure. As private credit secondaries expand, these pacing-linked covenants will become the key unit of analysis. They must be standardized, disclosed, and eventually priced as modular liquidity instruments.

Why This Matters for Secondaries: Either We Build Synthetic Standards—or Accept That Underwriters Rule

The covenant-centric model has immediate implications for the design and functioning of secondary markets in private credit. As private credit evolves beyond its origin-and-hold foundations, a new generation of secondary transactions—GP-led continuation vehicles, LP-led sales, and NAV-based lending—seeks to introduce liquidity into instruments that were never meant to trade. Yet the dominant pricing approaches on these platforms still rely on heuristics borrowed from public markets: peer comparables, discounted cash flow models, and backward-looking IRRs.

This presents a structural problem. In private credit, liquidity is not an emergent feature of market depth or trading volume. It is not revealed through pricing. Instead, it is encoded directly into the legal structure of the deal. The contract determines whether, when, and how capital can move. This means that yield is not just a reflection of risk—it is a function of liquidity rights designed ex ante.

In this context, underwriters—those who negotiate and structure the deal—are not peripheral actors. They are the core architects of liquidity. Their choices around call protections, NAV triggers, reinvestment restrictions, and amortization schedules define the mobility of capital. As such, the stability or fragility of the secondary market depends not only on pricing infrastructure, but on the cumulative effect of these contractual decisions.

When platforms fail to account for this, they risk treating structurally distinct assets as if they were fungible. Consider the earlier example: a $150 million unitranche loan designed around an investor’s pacing needs. Its NAV-based reset clause, staged draw schedule, and mandatory cash sweep provision effectively create a built-in exit path. The loan can be refinanced or bought out under conditions specified in advance. It functions more like a callable, amortizing bond with contractual liquidity engineered from the start.

Now compare this with a mezzanine loan of the same size and yield that lacks prepayment flexibility, NAV triggers, or reinvestment rights. Though the IRR may be similar on paper, the exposure path is fundamentally different. One offers structured flexibility; the other locks capital in place. Yet without covenant-level analysis, secondary buyers may price them as equivalents.

To address this, secondaries infrastructure must evolve beyond simple price matching and become covenant-aware. This requires:

• Mapping exit engineering: identifying the legal mechanisms that permit capital retrieval (e.g., take-out rights, NAV-based resets, change-of-control clauses).

• Scoring optionality asymmetry: understanding who holds control over exit and timing (e.g., call protections, performance ratchets).

• Analyzing pacing constraints: evaluating how repayment is timed and sequenced (e.g., PIK toggles, amortization schedules, reinvestment restrictions).

The goal is not contractual uniformity. Rather, it is synthetic standardization: grouping assets into comparable liquidity categories based on their legal architecture, not just their pricing profile. In this view, valuation begins not with cash flow projections, but with legal form.

This is not merely a technical improvement—it is foundational to building scalable, credible secondary markets. Without covenant-level transparency and structure-aware underwriting, platforms risk mispricing risk, allocating capital inefficiently, and undermining the legitimacy of private credit as an investable asset class. In markets where liquidity is governed by law, the covenant is the infrastructure.

This is a first draft, and I expect to revise and expand it—especially with empirical validation. But I wanted to share it early, in the spirit of dialogue. Many of the ideas were sharpened through discussion—especially with my phenomenal RA Kai Owen, whose insight and dedication made this process a pleasure.

Read the full draft here

I’m especially excited to continue exploring:

• How synthetic secondaries can be priced via exit engineering,

• How pacing metrics can link fund-level exposure targets to deal-level covenant design.

In private credit, the deal is the market—and contractual design is the first principle of value. If secondaries are to scale responsibly, they must be grounded not just in price discovery, but in the covenant logic that governs liquidity itself.

From the Author

Over the past month, three projects have quietly reshaped my thinking on market structure and valuation.

The first is a speculative proposal for an index fund built around my dealer-based, market microstructure–driven model of valuation—an attempt to translate theory into product design. The second is a working paper titled The Covenant Structure of Private Yield, which argues that in private credit, value formation depends less on pricing models and more on contractual architecture.

The third project is the upcoming Market Microstructure Symposium 2.0 at NYU Stern [link to details], where the current version of the program [link here] places substantial focus on infrastructure development in private market secondaries.

Together, these projects have forced me to confront a central question in today’s financial engineering landscape: Can the infrastructure that creates liquidity in public markets be meaningfully applied to private markets?

This post is an attempt to answer that question—and to explain why importing the logic of public-market liquidity into private credit risks not just a mismatch, but a conceptual failure.

Introduction: Public and Private Asset Liquidity Have Different Origins

The financial world often treats liquidity as a single, universal concept—something that can be manufactured, transferred, or wrapped into a product. But this assumption breaks down the moment we cross the boundary between public and private markets. Liquidity may be the goal in both domains, but the means of achieving it—and the architecture that makes it possible—are fundamentally different.

In public markets, liquidity is a function of trading infrastructure. It’s engineered through systems like dealer intermediation, continuous pricing, arbitrage mechanisms, and exchange connectivity. This “market plumbing” doesn’t just enable transactions—it produces prices and enables valuation through the act of trading itself.

In private markets, however, liquidity is not engineered but granted—and only under specific legal conditions. Here, value and tradability emerge from the structure of contracts, not from the existence of a secondary market or a central dealer. The scaffolding that makes movement possible consists of covenants, exit provisions, waterfall rights, and control clauses—not bid-ask spreads or clearinghouses.

Treating these two domains as interchangeable doesn’t just create a liquidity mismatch. It reflects a deeper conceptual confusion about what makes assets move—and what makes them valuable in the first place.

Public Markets: Liquidity as Engineered Infrastructure

In public markets, liquidity is not an intrinsic property of the asset. It is an outcome of a highly developed intermediation system: dealer inventories, bid-ask spreads, margin networks, continuous price feeds, arbitrage loops, and exchange-based trading protocols. What I call market plumbing is the infrastructure that allows an asset to circulate, be repriced, and be hedged.

This infrastructure doesn’t just support liquidity—it produces it. In fact, prices in public markets do not exist independently of this structure; they are generated through it. The valuation of a security is shaped as much by its tradeability as by its fundamentals. Trading is not merely a reflection of value—it is part of how value is constructed.

This is why ETFs can transform assets that are otherwise illiquid in their native markets. A thinly traded bond, when added to an ETF, often becomes more liquid—not because the bond itself has changed, but because it has been recontextualized. The ETF connects the asset to a pipeline of market makers, authorized participants, and arbitrageurs who create a reliable mechanism for trading and valuation.

You’re not creating liquidity out of thin air—you’re giving the asset access to a system that knows how to carry it. The ETF doesn’t change the bond. It changes the environment around the bond.

In public markets, prices do not exist outside the system. They are generated through it.

This shiftability is key. ETFs give illiquid assets access to the broader ecosystem of pricing, hedging, and trading. The wrapper connects them to pipes they wouldn’t otherwise reach. In this way, public-market liquidity is less about the characteristics of the asset itself and more about the environment into which it is inserted. When that environment is dense with intermediaries and structured for flow, even the thinly traded can circulate.

Private Markets: Liquidity as Legal Possibility

In private markets, liquidity does not flow through engineered infrastructure—it is written into existence. Private credit assets don’t become tradable because they’re plugged into an exchange or routed through an ETF. They become tradable—if at all—because the legal structure of the contract allows it.

There is no centralized order book. No standing market maker quoting both sides. No shared infrastructure channeling trades. Instead, liquidity is contingent on what the deal documents authorize: the covenant package, transfer restrictions, consent rights, waterfall mechanisms, and control provisions. These are not mere legal technicalities—they are the very architecture of tradability in private credit.

This is why I refer to private market liquidity as a function of legal scaffolding. It is not engineered from the outside—it is granted from within. A private credit instrument can only move if the contract says it can. Some deals allow flexible exits, secondary transfers, or partial restructurings. Others lock down exposure so tightly that even a motivated seller with a willing buyer cannot transact.

In private markets, liquidity is not engineered. It is authorized by the covenant structure of the contracts.

In public markets, liquidity is created by plugging the asset into an ecosystem that knows how to price, hedge, and circulate it. In private markets, no such ecosystem exists independently of the contractual covenants of the private deals. Every contract defines its own market.

And this creates a critical asymmetry. Financial engineering in public markets—ETFs, syndication platforms, algorithmic routing—works because assets are designed to respond to infrastructure. But in private markets, infrastructure cannot override legal restrictions. A wrapper can’t make an asset liquid if the asset’s legal form prohibits movement.

Put simply: in public markets, liquidity is a feature of the system. In private markets, it’s a condition of the contract. This is not just a difference in degree—it’s a difference in kind. Confusing the two leads to failed products, distorted valuation logic, and a misunderstanding of where liquidity actually comes from.

And yet, many private credit ETFs rely on precisely this false logic: the idea that by placing an illiquid private loan into a public wrapper, one can unlock liquidity.

Why Private Credit ETFs Misunderstand Liquidity

This is why many private credit ETFs rest on a flawed foundation. They borrow the engineering logic of public markets—the idea that liquidity can be designed around the asset by embedding it in a wrapper and plugging it into a functioning trading system. That logic holds in public markets because liquidity is activated by infrastructure. But in private credit, the precondition for liquidity isn’t infrastructure—it’s permission.

Even if you build an ETF with efficient redemption mechanics and a polished market-making apparatus, it cannot force liquidity where the legal terms of the asset forbid it. You can’t transfer what you don’t have the right to sell. You can’t price what you’re not allowed to circulate. You can’t hedge what isn’t even shiftable.

This makes the private credit ETF not just a mismatch of asset and vehicle—but a conceptual error. It assumes that liquidity can be ported from one market system to another without accounting for the source of that liquidity. But in private credit, liquidity is not derived from circulation—it is rooted in legal design.

In public markets, liquidity follows structure. In private markets, structure defines whether liquidity is even possible.

So when these ETFs promise public-market liquidity for inherently private-market instruments, the problem is not simply a “liquidity mismatch.” That framing implies a continuum, a spectrum of liquidity that can be stretched or compressed. But what’s actually at stake is a categorical difference: between assets that derive liquidity from market structure, and assets that derive liquidity from contract law.

Public-market logic cannot be superimposed on private-market realities without structural consequences. When financial engineering is used to simulate liquidity where none is permitted, it doesn’t just distort tradability—it distorts valuation itself. Investors are led to believe that continuous pricing is meaningful, when in fact it may be mechanically generated by a wrapper with no underlying ability to move the asset.

In this way, the ETF becomes a shell of liquidity—not its substance. And without clear attention to the legal scaffolding underneath, the result is not market innovation but valuation confusion.

Why Standardization Matters

Why Standardization Matters

If private credit is to develop functional secondary markets—and eventually support public-facing structures like ETFs—then the foundation must be legal, not just financial. That foundation is standardization.

In public markets, standardization is what enables scale. Bonds can be traded across dealers because their terms are broadly similar; equities can be priced continuously because their rights and cash flows are governed by uniform rules. This uniformity allows liquidity to emerge through market infrastructure: if assets are interchangeable in form, they can be circulated, priced, and hedged interchangeably in function.

In private markets, that logic breaks down. Every loan agreement is bespoke. The ability to transfer, restructure, or resell an asset is determined not by market conditions, but by the specific language of each contract. There is no generalized infrastructure because there is no generalized asset.

The route to functional secondary markets in private credit doesn’t begin with ETFs—it begins with enforceable, standardized contracts.

Standardization changes that. By converging around repeatable covenant structures—transfer rights, consent mechanics, callability, waterfall provisions, and default resolution protocols—private credit instruments can become legible to infrastructure. Only then can systems like continuous pricing, active secondary trading, and ETF-based liquidity start to function meaningfully.

Put differently: before private markets can absorb public-market engineering tools, they must first adopt public-market legibility. The path to real liquidity runs not through innovation in wrappers, but through alignment in contractual terms.

Without standardization, liquidity must be granted. With it, liquidity can be engineered. This is the pivot point for private credit’s next phase of development—and the critical precondition for any effort to translate private yield into tradable public exposure.

Conclusion: Infrastructure Only Works on Assets Designed for It

The public-private bridge must be built with attention to what holds each side up.

In public markets, liquidity is created through pipes and infrastructure. It is a function of intermediation. In private markets, liquidity is created through covenants and contracts. It is a function of law.

Treating these domains as interchangeable is not simply imprecise—it risks undermining the valuation integrity of both.

Market infrastructure can circulate, price, and trade—but only if the asset is built to move. In private asset, that mobility cannot assumed: you can only plug a private asset into financial plumbing if the legal scaffolding allows it to flow.

---

Real liquidity in private markets requires more than an ETF. It requires a redesign of the legal architecture that governs the asset. Only then can infrastructure be meaningfully applied.

Until that happens, financial engineering must learn to respect the constraints of its foundations. Otherwise, it doesn’t create liquidity—it builds the illusion of it.

References

• Tuckman, B., & Serrat, A. (2011). Fixed Income Securities: Tools for Today’s Markets. Wiley Finance.

• Meyer, T. (2022). The Art of Commitment Pacing: Investment Management in Illiquid Markets. Wiley Finance.

• Hill, J. M., Nadig, D., & Hougan, M. (2015). A Comprehensive Guide to Exchange-Traded Funds (ETFs). Wiley Finance.

• SEC Rule 6c-11 and public filings on ETF structures for illiquid assets.

• Saeidinezhad, E. (2025, forthcoming). The Covenant Structure of Private Yield: Synthetic Standardization and a Theory of Yield Formation in Private Credit.

• Date/ Time/ Location: Friday, October 24, 2025 | 9:00 AM to 4:30 PM | NYU Stern

• Format: In-person | Invitation-only | Chatham House Rule

We are pleased to announce the second annual Market Microstructure Symposium, hosted at NYU Stern and co-sponsored by Nasdaq. This year’s program explores how the architecture of modern markets is being quietly—but profoundly—reshaped by innovations in infrastructure, product design, and institutional roles.

The symposium brings together leading voices from asset management, trading, banking, and policy to examine how modern plumbing—from secondaries and NAV financing to buffered ETFs and synthetic market makers—is shifting the boundaries of risk and liquidity.

The full draft program is attached here.

Attendance is by invitation only. To request an invitation, please contact:

• Elham Saeidinezhad (NYU Stern / CCNY) — es5799@stern.nyu.edu

• Itai Dreifuss (Columbia University) — id2349@columbia.edu

• Kai Donovan Owen (Columbia University) — kd3581@columbia.edu

Symposium Themes

This year’s panels center on two critical domains—private markets and public markets—and explore one unifying question:

As financial infrastructure evolves, who is underwriting these systems—and who ultimately bears the risk?

Each panel addresses a specific facet of this question, surfacing the deeper linkages between innovation, intermediation, and systemic stability.

Panel I

Beyond Liquidity, Before Maturity: Rethinking Market Design in Private Secondaries

9:30 – 10:30 AM

As secondaries shift toward GP-leds, NAV-linked deals, and real asset structures, deal design becomes infrastructure. This panel explores how workflows, counterparties, and market conventions are adapting to the evolving complexity—and whether scale is outpacing standardization.

Panelists:

• Brian A. Ruane (BNY Mellon)

• Lars Pace (Hamilton Lane)

• Gerald Cooper (Campbell Lutyens)

• Christoffer Tinsley (Nasdaq)

• Moderator: Daniel Sullivan (PwC)

Key Questions:

• Are secondaries today too bespoke to scale?

• What market design improvements are needed to support growth in GP-leds, real assets, and NAV-linked deals?

• Can private markets develop shared infrastructure without sacrificing flexibility?

• What role should intermediaries and allocators play in shaping next-generation market standards?

Panel II

Silent Partners: Following Banks Through the Architecture of Private Credit

10:40 – 11:40 AM

NAV loans. Subscription lines. Hybrid leverage. As banks become deeper partners in private credit infrastructure, are they accumulating unseen exposures? This panel unpacks how bank-driven platforms shape capital flows—and where fragility might quietly build.

Panelists:

• Jannis Meindl (Goldman Sachs)

• John Golden (Apollo Global Management)

• Adam Reinmann or another Partner (Wendel Group)

• Moderator: Samim Ghamami (SEC)

Key Questions:

• What tools—NAV facilities, subscription lines, hybrid leverage—are drawing banks into private credit markets?

• How visible are these exposures on balance sheets and regulatory filings?

• What are the tradeoffs of bilateral infrastructure versus neutral platforms?

• Are banks becoming the ultimate risk holders for underperforming portfolios and opaque deals?

Panel III

Market Making Without the Label: The Rise of Prop Traders, Transformation of Bank Dealers, and Hedge Fund Ambiguity

Time: 1:00 – 2:00 PM EST

Panelists:

• Sam Berberian (Citadel)

• Matthew Schrager (TD Securities)

• Andreas Koukorinis (Optiver)

• Larry Rascio (Olden Lane) — Moderator

This panel explores the shifting architecture of modern liquidity provision. Traditionally the domain of designated dealers—typically large banks—market making has evolved as proprietary trading firms increasingly operate as de facto liquidity providers, using speed, data, and balance sheet agility. At the same time, hedge funds now straddle both sides of the market: in some venues, they compete with dealers for flow and pricing; in others, they maintain the privileges of clients. This strategic ambiguity places them in a structural gray zone. The panel will examine how banks, prop shops, and hedge funds are adapting their strategies as traditional roles and labels lose their clarity.

Key Questions:

• Are prop trading firms now the dominant liquidity providers in key markets?

• How does the dealer–client boundary function in practice, and where does it break down?

• What trade-offs do hedge funds face in straddling both sides of the liquidity relationship?

• How should we understand the future of market making when labels no longer match roles?

From Options to Outcomes: Buffered ETFs and the New Architecture of Public Markets

2:10 – 3:10 PM

Buffered ETFs promise defined returns—but require tight coordination across listed options, clearinghouses, and ETF flows. This panel explores how infrastructure is being used to engineer retail outcomes—and whether it signals a deeper transformation in how public markets handle risk.

Panelists:

• Josh Khalif (Flow Traders)

• James Maund (Krane Funds Advisors)

• Samuel Moore (Vest Financial)

• Moderator: Elham Saeidinezhad (NYU Stern)

Key Questions:

• How are buffered ETFs linking ETF, options, and clearing markets—and what does that mean for risk transmission?

• Are market makers and clearinghouses becoming active participants in outcome engineering?

• Does the rise of structured ETFs signal a deeper transformation in public market infrastructure?

Keynote Conversation

Infrastructure, Intermediation, and the Next Frontier of Financial Stability

3:20 – 4:30 PM

This closing conversation connects market microstructure to broader questions of systemic resilience. What does today’s fragmented and bespoke infrastructure mean for transparency, risk aggregation, and financial stability?

Speakers:

• Steven Kelly (Advisor to Chairman, Guggenheim Partners; formerly Yale SOM)

• Chuck Mack (SVP, Strategic Operations & Public Policy, Nasdaq)

Key Themes:

• How are private markets redefining liquidity—and what infrastructures underpin that transformation?

• Can bilateral systems scale without hidden fragility?

• What role should regulation play in shaping next-gen market plumbing?

• Where do transparency gaps emerge between public and private oversight?

Logistics

• Location: NYU Stern

• Date: Friday, October 24, 2025

• Time: 9:00 AM to 4:30 PM (Breakfast & registration begin at 9:00 AM)

• Format: In-person, invitation-only

• Chatham House Rule: All discussions are off-the-record to foster open dialogue

➤ [Link to the Symposium’s Program]

If you would like to attend, please email one of the organizers listed above. Due to space constraints, attendance is limited and strictly by invitation.

June 2025

“It does not matter how good your trading and investment idea is—what matters is the execution of it.”

PART I- Introduction

Economist Elham Saeidinezhad- Market Microstructure Project- speaks with Robert Almgren—co-creator of the Almgren–Chriss (A/C) model and a founding figure in modern execution theory—as part of the Market Microstructure Project. This interview examines how evolving market structure—fragmented liquidity, dark pools, and algorithmic execution—shapes portfolio management today. Almgren argues that in modern markets, execution quality is not a back-office concern; it is the core of investment strategy.

Risk-Adjusted Return vs. Market Structure

The conversation revisits the central insight behind the A/C model: every trade involves a fundamental trade-off between market impact and execution risk. Unlike the standard Sharpe ratio framework, which emphasizes optimizing portfolio returns based on fundamentals and risk-adjusted performance, the Almgren–Chriss approach focuses on the market structure itself—how liquidity, volatility, and execution conditions determine the feasibility and cost of implementing any strategy. We explore how this execution-centric framework has evolved over two decades of transformation in equity markets.

The interview also engages with a modular portfolio framework proposed by Giuseppe Paleologo, who incorporates the A/C model as the final step in a simplified investment workflow. While Almgren does not endorse this framework, it provides a useful reference point for exploring how simplicity, alpha uncertainty, and execution constraints intersect in real-world portfolio construction.

Elham has also written an analytical essay based on this interview, available here:

Beyond Sharpe: Reimagining Portfolio Management Around the Almgren-Chriss Ratio

For those interested in learning more from Robert, you can also read the first part of our conversation, which focuses on futures market structure:

Inside the BlackBox of Futures Market Trading Structure: An Interview with Robert Almgren

PART II- An Interview by Elham Saeidinezhad with Robert Almgren

ES: Robert, thank you for joining the Market Microstructure Project. Today’s discussion bridges your foundational work on optimal execution with Giuseppe Paleologo’s reinterpretation of that work in his recent Columbia talk. Giuseppe frames the Almgren-Chriss (A/C) model as the final step in a modular, spreadsheet-based approach to portfolio construction. My role is to moderate a conversation between your theoretical foundation and its evolving application.

Let’s start with the original assumptions. Your model is widely recognized for providing a formal framework to manage the trade-off between market impact—the price slippage caused by executing trades—and execution risk, typically modeled as return variance during execution. But it was developed before smart order routing, fragmented order books (CLOBs), and modern quant infrastructure. What assumptions about market structure—such as order book mechanics or venue access—did the original A/C model embed?

RA: Very few. Some assume we were modeling market versus limit orders, but in reality, we abstracted away from specific order types. The model divides time into discrete intervals—say, 5 or 10 minutes—and assumes an underlying “micro-trader” executes the trade within each interval. What we modeled is the aggregate effect of this execution process: how trading intensity affects price over time.

“We didn’t model alpha in the original paper, but the trade-off still holds: trade slowly to reduce market impact, trade quickly to reduce risk.” — Robert Almgren

ES: So aside from the time slicing, the model assumes stable market conditions. Given how much execution environments have evolved—dark pools, smart routing, fragmented liquidity—how reliable are those assumptions today?

RA: Correct. The model’s only real assumptions are that you can estimate market impact parameters and that they remain stable over the execution window. Whether that holds today is another matter. Since we published the model 25 years ago, others—like Jim Gatheral—have introduced refinements, including time decay and nonlinear impact. We chose simplicity: trade a fraction of daily volume, incur a proportional cost in daily volatility.

ES: Giuseppe describes your model as capturing 80% of what’s needed in portfolio construction with just 20% of the complexity. He keeps your core trade-off but adds alpha as another layer. How do you think about incorporating alpha while preserving that simplicity?

RA: Start with the economic trade-offs. If you trade slowly, you reduce market impact. If you trade quickly, you reduce risk from price volatility over time. That’s the core. We didn’t include alpha—expected return forecasts—but it’s easy to add. If you have alpha, then speed also reflects urgency: you want to complete a trade before prices move. But that’s a layer on top of the volatility–impact trade-off.

“The Almgren-Chriss model is more than an optimizer—it’s a framework for thinking about how to convert a portfolio allocation into an execution path.” — Elham Saeidinezhad

ES: Giuseppe starts his framework by “tokenizing” alpha—removing overlap with market beta to isolate pure signals. Since your original model didn’t include alpha, do you see this as a natural extension, or does it change the model’s purpose?

RA: It’s thoughtful. Around the same time we published our work, Bertsimas and Lo proposed a model for optimal execution with short-term alpha signals. Their paper often gets cited alongside ours, which can be misleading. Their model incorporates alpha but omits risk aversion, which was central to ours.

ES: Could you explain why it’s misleading to group your model with theirs? What’s the key structural difference?

RA: Structurally, their framework is algebraically separable into two components: a baseline execution trajectory and an overlay of alpha-driven adjustments. I suspect Giuseppe’s interpretation builds on this: use the A/C model as a base, then layer in short-term signals.

ES: Since both models solve for optimal execution, what do you think causes the confusion? Is it a matter of appearance versus motivation?

RA: They appear similar—they both solve for optimal liquidation under market impact—but the underlying motivations differ. We focus on minimizing execution cost under risk aversion. They focus on exploiting alpha signals. If you plug in realistic parameters—small alphas, modest autocorrelation, typical volatilities—you find that the alpha signal adds very little. It just doesn’t generate meaningful profit in most realistic scenarios.

“Covariance matrices are well-understood and measurable. Simplifying them too far means giving up on hedging and diversification.” — Robert Almgren

ES: Let’s talk about how changes in market structure affect the A/C model. On one hand, we’ve seen structural shifts—like Reg NMS and the rise of central limit order books (CLOBs)—that alter how markets function. On the other, there are transitory but impactful dynamics, such as discovering hidden liquidity mid-trade. How do these changes—both structural and informational—affect the model’s core assumptions?

RA: Not much, actually—at least in terms of structural changes. Developments like Reg NMS and the growth of CLOBs have increased market transparency and regularity. The A/C model doesn’t rely on forecasting prices; it requires that you can estimate stable characteristics like volatility and market impact. If the market environment is reasonably time-homogeneous, the model’s calibration holds.

That said, in practice, volatility and impact can shift. I addressed this in later work: if those inputs evolve, you can adjust the solution dynamically—assuming you have a model for how they change. But what really complicates things are transitory informational shocks—like discovering unexpected liquidity mid-trade.

Dark pools are a good example. If hidden volume becomes available over a period longer than your typical execution window, it violates the stationarity assumption—that conditions remain stable across the trade. When that happens, the model has to be recalibrated on the fly. Still, decisions like whether to route to dark or lit venues are delegated to the micro-trader—the execution algorithm operating beneath the model.

ES: That idea—that simplification requires knowing which assumptions matter—also shows up in Step 2 of Giuseppe’s framework. After isolating alpha, he assumes a diagonal covariance matrix to make the optimization separable and easier to implement. The assumption is that asset returns are uncorrelated. How do you view that trade-off?

RA: Just to confirm—he assumes diagonal covariance, not diagonal impact?

ES: Correct. Diagonal covariance.

RA: Got it. If both the cost matrix and the covariance matrix  are diagonal, then the resulting matrix is also diagonal. That certainly simplifies the optimization—especially for spreadsheet-based workflows—but it comes at the expense of realism.

In our own work, we also assumed a diagonal market impact matrix for tractability. But we kept a full covariance matrix because that’s where cross-asset relationships live. It’s what lets the model detect hedging opportunities and optimize along low-risk directions. Ignoring that structure means discarding a key part of portfolio behavior.

“Dark pools and secondary LP markets both violate the assumption of stationarity. Once that breaks, you have to re-optimize.” — Robert Almgren

ES: So by simplifying the covariance matrix, we lose the ability to model how assets interact. Can you explain what types of trading strategies or risk management techniques depend on keeping those relationships intact?

RA: Suppose two assets are highly correlated. If you go long one and short the other in proportion, you form a near-zero-risk position. That’s a core strategy in relative value trading. If your model assumes no correlation, you miss it.

The optimization is still solvable—even with a full covariance matrix—by using eigenvector decomposition. But diagonalizing it discards valuable information on diversification and hedge efficiency.

ES: Given that correlation enables strategies like relative value and improves risk control, do you think assuming a diagonal covariance matrix—just for simplicity—overlooks too much? 

RA: Yes. Covariance matrices are well-understood and relatively easy to estimate. Impact matrices are harder to calibrate, so simplifying them makes more sense. But assuming zero correlation between assets is too aggressive.

ES: Can you give a concrete example where preserving correlations changes the execution strategy—something a diagonal covariance matrix would completely overlook?

RA: Take a portfolio of equities and a futures position. Futures provide a liquid hedge for broad market exposure. A model with full covariance sees this and recommends hedging with the future first—then slowly adjusting the rest. A diagonal covariance matrix would miss that logic entirely.

ES: So in that case, the model wouldn’t even recognize common execution strategies like index futures hedging or basis trades—because it fails to see the correlation structure?

RA: Exactly. The model becomes blind to any strategy that relies on asset correlation. That may be computationally convenient, but it comes at a cost.

“Giuseppe’s motto—‘Simple is not easy’—captures what good modeling should strive for.” — Robert Almgren

ES: Let’s move to the third step in Giuseppe’s framework: accounting for alpha uncertainty. He adjusts position sizes based on the confidence level in each alpha signal, typically using the Information Coefficient (IC) as a proxy for forecast accuracy. But how distinct is alpha uncertainty from volatility in practice? Aren’t they often capturing the same thing?

RA: It’s an interesting idea, but I’d frame it differently. In most cases, uncertainty in alpha is indistinguishable from price volatility. For example, suppose a stock has zero alpha and 10% volatility—that’s straightforward. Now suppose it has an alpha of 10% with 5% uncertainty, and an additional 5% volatility. You still end up with 10% overall uncertainty. It’s not clear what portion of that comes from alpha uncertainty versus price movement.

ES: So unless the alpha signal is stable over time, there’s no real modeling benefit to treating its uncertainty separately from volatility?

RA: Exactly. Conceptually, alpha is the expected value of future returns. Volatility captures the uncertainty around that outcome. Unless there’s time persistence in the alpha signal—meaning the signal holds consistently over a period—there’s little reason to separate the two.

ES: Can you give a practical example where it actually makes sense to separate alpha uncertainty from volatility?

RA: In practice, I’ve explored this in our trading systems and in a lesser-known paper. There, we model asset managers forming alpha forecasts overnight. Those signals remain fixed throughout the day. At the open, estimates are noisy, but as the day progresses and drift becomes clear, the alpha becomes observable. In that context, uncertainty in alpha is distinct from volatility because it has a time structure.

“Giuseppe’s modular framework is built for clarity and speed—but without a full covariance matrix, it risks missing the very structure that powers basis trades and relative value strategies: the correlations that turn portfolios into hedges, not just positions.”— Elham Saeidinezhad

ES: So in a case like a clinical trial, where information unfolds gradually, the alpha becomes more predictable as time passes?

RA: Precisely. That’s different from saying, “I expect a return of 2% with a ±1% margin,” and then adding 5% volatility. In that case, you’re just dealing with total uncertainty—around 6%. You don’t need two separate terms. Without persistence, the alpha’s uncertainty is effectively absorbed into volatility.

ES: So in Step 3 of Giuseppe’s framework—where he adjusts position sizing based on alpha uncertainty—do you think that adds unnecessary complexity unless the alpha has real time persistence?

RA: Correct. If the signal is persistent—like in an overnight model or in slowly unfolding news—then distinguishing alpha uncertainty makes sense. Otherwise, it’s just volatility.

ES: Are there any market patterns or behaviors that can help you spot when alpha might be persistent?

RA: Not explicitly. But you might infer it from behavior. If buy or sell pressure emerges in consistent bursts—perhaps because asset managers are updating positions in waves—you might detect short-term persistence in price moves. That could serve as a proxy for alpha. But again, it’s a question of autocorrelation, not separate uncertainty.

ES: In Step 5, Giuseppe feeds all inputs—alpha, uncertainty, covariance, and costs—into what he frames as the A/C model to generate an execution schedule. Do you think that’s a fair application of your framework?

RA: He’s generous in calling it the A/C model—it includes more than we originally proposed—but yes, I see what he means. Once you’ve specified the input parameters, the model computes the optimal trade schedule. In some cases, you can solve it analytically. In others, you need numerical optimization. But it’s a well-posed problem once the structure is defined.

ES: One thing that makes your model so distinctive is how clearly it captures the trade-off between market impact and volatility. It feels less like a pure optimizer and more like a framework for thinking about execution under risk.

RA: That was intentional. We designed the model to be analytically tractable. Linear market impact leads to a quadratic cost function, and if you use variance as the risk measure, the objective remains quadratic. That symmetry means scaling the trade size increases both cost and risk proportionally—by a factor of four if size doubles.

That symmetry means scaling the trade size increases both cost and risk proportionally—by a factor of four if size doubles. An important implication is that optimal execution time doesn’t depend on portfolio size. That’s counterintuitive—most people assume larger trades should take longer. But in our framework, the time scale remains constant because cost and risk grow at the same rate.

ES: That’s subtle but powerful. So the constant execution time isn’t something you built in—it’s a direct consequence of the model’s core assumptions?

RA: Exactly. It follows from using arithmetic Brownian motion for prices and a mean-variance objective. If we’d used geometric Brownian motion, the results would be more complex without much added insight.

ES: That’s really helpful for practitioners applying the model today. Taking a step back, do you feel Giuseppe’s framework stays true to the core principles behind your original model?

RA: I do. The sequence makes sense: define the market impact model, specify your objective and constraints, then solve. That’s what we did. We assumed linear impact, arithmetic Brownian motion, and mean-variance optimization. From there, the math follows.

“A lesson for practitioners: If your alpha doesn’t persist over time, its uncertainty isn’t special—it’s just volatility in disguise.”— Elham Saeidinezhad

ES: Before we wrap up, I’d like to touch on two final topics. First, how does your model connect with broader risk management tools, like hedging and synthetic risk transfer? And second, could the framework be adapted for private markets—particularly secondaries in private credit or private equity?

RA: The A/C model doesn’t translate easily to private markets. One of its core assumptions is market smoothness—the idea that trades can be executed continuously and that execution cost is a deterministic function of trading speed. This assumption fits public equities, where order books update many times per second. But in private markets, execution is discrete, lumpy, and highly dependent on buyer availability. There’s no continuous price discovery mechanism.

ES: So it’s not just about speed—it’s that the basic assumptions behind the A/C model don’t hold in private markets?

RA: Exactly. In private markets, you can’t continuously vary your trading rate. There’s no guarantee of finding counterparties, and pricing is often opaque. That said, some ideas do carry over—particularly the trade-off between price impact (e.g., discounting to NAV) and holding risk. If faster liquidation means accepting a larger NAV discount, you could reinterpret that relationship as a kind of market impact function—but it wouldn’t be continuous. It would look more like a step function or regime-switching model.

ES: So in principle, could you adapt the A/C model to private markets by treating NAV discounts as a form of market impact tied to exit speed?

RA: Yes, but you’d need a new cost function that reflects buyer behavior. Instead of a smooth curve, the model might reflect discrete pricing regimes: slow exits with minimal discount, faster exits with steeper discounts. You’d also need to model constraints around buyer availability, fund structure, or timing—perhaps waiting for a block buyer or accepting near-term liquidity at a cost. It’s possible, but it takes you well beyond the original scope of the A/C framework.

ES: What’s striking is that the core trade-off—between execution cost and exposure risk—still applies. And your point about liquidity discovery seems to carry over too: in private markets, finding a block buyer mid-trade feels a lot like uncovering hidden volume in a dark pool. Would you say the same modeling challenges apply in both cases?

RA: Absolutely. In both cases, discovering liquidity mid-execution violates the assumption of stationarity—that market conditions remain constant during the trade. Once that assumption breaks, the entire solution must be recalculated. Whether it’s a dark pool or a secondary LP market, the principle is the same.

ES: One final concept I want to revisit is time decay and impact persistence. Giuseppe touches on this in his presentation, which shows that market impact tends to decay over time in a power-law fashion. How does this align with the A/C model?

RA: It’s a critical extension, but it wasn’t part of our original model. At the time, there wasn’t much empirical literature on decay dynamics. So we simplified: we modeled market impact as either permanent (a lasting price shift) or temporary(an effect that disappears instantly). Nothing in between.

But we now know that impact decay exists. Each trade has a lasting influence, and the total effect is cumulative. This leads to an integral formulation, where every execution decision echoes through future prices. Gatheral formalized this idea, and Obizhaeva and Wang built models that explicitly solve for optimal execution under decay assumptions—whether exponential or power-law.

These models are more realistic but significantly harder to solve. The optimal strategy often involves front-loading, waiting for impact to dissipate, and then finishing the trade. It’s elegant, but computationally intensive. 

ES: What stands out in both your original paper and this conversation is how deliberately you manage complexity—modeling what matters and bracketing the rest. “Simple is not easy” motto really seems to echo that mindset.

RA: Exactly. The goal is not to ignore complexity but to distill it. You want a model that captures the essential trade-offs while staying tractable. That was our philosophy.

ES: That’s exactly what we hoped to highlight through this dialogue.

References

1. Almgren, R., & Chriss, N. (2000). Optimal execution of portfolio transactions. Journal of Risk, 3(2), 5–39.

2. Bertsimas, D., & Lo, A. W. (1998). Optimal control of execution costs. Journal of Financial Markets, 1(1), 1–50.

3. Gatheral, J. (2010). No-dynamic-arbitrage and market impact. Quantitative Finance, 10(7), 749–759.

4. Obizhaeva, A., & Wang, J. (2013). Optimal trading strategy and supply/demand dynamics. Journal of Financial Markets, 16(1), 1–32.

5. Paleologo, G. A. (2025). The Elements of Quantitative Investing. Wiley Finance.

6. Paleologo, G. (2025). Simple (Portfolio) Heuristics That Make Us Smart(er). Talk presented at Columbia University. [Unpublished lecture slides].

7. Regulation National Market System (Reg NMS). (2005). U.S. Securities and Exchange Commission. https://www.sec.gov/rules/final/34-51808.pdf

This piece emerges from an ongoing research dialogue with my RA, Kai Owen, as we explore the structural foundations of private credit markets. It is part of a broader academic project on the architecture of private capital, and also serves as intellectual groundwork for the HFA–Market Microstructure Symposium, which we are co-organizing at NYU Stern on October 24, 2025.

The symposium will feature two panels dedicated to private capital:

• Panel I — Beyond Liquidity, Before Maturity: Closing the Infrastructure Gap in Private Secondaries

• Panel II — Silent Partners: Following Banks Through the Architecture of Private Credit

Together, these panels examine how risk, return, and intermediation are being reshaped in private markets—not just through capital flows, but through the evolving microstructure of contracts, infrastructure, and institutional roles.

Intro: A Shift in Perspective

When execution is no longer just a step. It’s “the” portfolio management strategy.

This analytical essay accompanies a forthcoming interview for the Market Microstructure Interview Series (Forthcoming, June 2025) with Robert Almgren, co-creator of the Almgren-Chriss model. In that conversation, we explore Giuseppe Paleologo’s recent book, The Elements of Quantitative Investing, using it as a structured lens to revisit the Almgren-Chriss framework and reimagine its relevance in today's investment environment.

Paleologo’s five-step portfolio construction pipeline provides a pragmatic scaffold for thinking through portfolio design under modern execution constraints. In this piece, I use his pipeline to reinterpret the Almgren-Chriss model not just as an execution overlay, but as a foundational theory of portfolio construction—one that displaces the classical Sharpe ratio (expected return per unit of volatility) with what I call the Almgren-Chriss ratio: the trade-off between market impact and volatility risk.

Along the way, I explore an idea Almgren himself is skeptical of: whether the logic of the A/C model—originally built for public markets with smooth, continuous trading—can be extended to the world of private capital. In particular, I look at private credit and secondaries, where liquidity is episodic, pricing is negotiated, and execution is deeply intertwined with the structure of covenants and deal pacing. These features make the Almgren-Chriss perspective unexpectedly illuminating for understanding how portfolios in private markets are assembled, held, and unwound.

Each section of this piece corresponds to one of Paleologo’s five steps. Together, they build a parallel vision of portfolio management—one grounded in execution reality rather than forecasted returns. The final section provides a summary table of five key implications for managing private portfolios, especially in covenant-intensive environments like private credit.

Step 1: Tokenizing, Not Chasing, Alpha

Alpha is only useful if it survives execution

The Almgren-Chriss model introduces a radical rethinking of portfolio construction. Whereas traditional frameworks like the Sharpe ratio optimize expected return relative to volatility, Almgren-Chriss pivots away from return altogether and instead centers execution quality. The key question becomes: how do we minimize the combined cost of market impact and volatility risk over the life of a trade?

In this framework, alpha—defined as the expected excess return—plays no explicit role. This omission is intentional. Including alpha would require making assumptions about the strength and stability of predictive signals. As Robert Almgren explains in his interview, unless an alpha signal persists over the duration of execution, its uncertainty is effectively indistinguishable from volatility. Rather than mixing forecasting with execution, the model constructs a "clean-room" foundation: one that isolates the mechanics of trading from the uncertain art of return prediction.

In practical terms, this means the model omits expected returns entirely. Execution strategy is based solely on market mechanics—how much cost and risk one incurs by trading more or less aggressively. This framing is well suited to public markets, where liquidity is continuous and transaction costs are estimable. But its real contribution lies in reshaping the axis of portfolio decision-making: it shifts us from optimizing return per unit of risk to managing execution quality per unit of exposure.

This shift also offers a powerful lens for understanding private capital markets—even though the A/C model was not designed for them. In private credit, for example, alpha is not a forecasted return series but is embedded in covenant structures: in how enforceable they are, how quickly capital can be deployed or withdrawn, and under what terms exits can occur. The value of a position depends less on its yield and more on its liquidity path—on the contractual and structural channels through which that yield can be realized.

Thus, while the A/C model is not directly applicable to private markets—where liquidity is episodic and buyer availability uncertain—its core principle still resonates. By putting execution feasibility at the center, it allows us to reinterpret private capital not as a yield-maximizing game, but as a sequence of transactional choices shaped by structure, timing, and constraint.

This perspective also aligns with Giuseppe Paleologo’s five-step portfolio construction pipeline. His first step, “tokenizing alpha,” involves isolating actionable signals only after structural exposures have been accounted for—echoing the A/C ethos. In Paleologo’s view, alpha is not the starting point but the final adjustment, introduced only once the machinery of execution has been fully specified.

Ultimately, the Almgren-Chriss model formalizes a different trade-off: not risk-adjusted return, but impact-adjusted volatility. It defines a new axis for portfolio thinking—one that foregrounds cost, feasibility, and timing. This creates a safer, more stable investment universe, but one that also demands more from alpha. To earn its place, alpha must not only be statistically significant, but persistent, resilient, and most of all—executable.

In execution-driven models, alpha is guilty until proven tradeable.

Step 2: Correlation as Execution Strategy, Not Just Risk Transfer

When Contracts Interact, Execution Must Coordinate

In traditional portfolio theory, correlation is all about risk smoothing. Step 2 is typically where diversification happens: combine uncorrelated assets to reduce volatility. This is the Sharpe ratio worldview—optimize risk-adjusted return by spreading bets. Even in Giuseppe Paleologo’s streamlined pipeline, this step simplifies the covariance structure, often by assuming assets are uncorrelated, so alpha can be isolated and optimized cleanly.

But both of these frameworks are fundamentally return-centric. The Almgren-Chriss model flips the axis. It treats correlation not as a risk-spreading tool, but as an execution constraint—or advantage. In this world, correlation matters if it affects how—and when—positions can be traded. It’s not about co-movement of prices; it’s about co-dependence of liquidity. This is an execution-centric view.

This shift is especially critical in private capital markets, where traditional diversification tools are often unavailable—and, in many cases, irrelevant. You can’t hedge sector exposure with ETFs, and there’s no liquid secondary market to offload concentrated risk. Yet correlation is everywhere—not in the form of co-moving prices, but as contractual interdependencies. These emerge when multiple loans share a sponsor, rely on similar covenant structures, or are tied to the same funders, legal frameworks, or refinancing timelines. These aren’t statistical correlations that smooth returns; they’re structural linkages that constrain—or enable—execution. If ignored, they can cause liquidity bottlenecks. But when properly understood and mapped, they allow managers to sequence trades intelligently, unlocking pockets of execution flexibility where none seemed to exist.

Here’s the distinction: in traditional theory, uncorrelated assets are good because they reduce portfolio risk. In private markets, uncorrelated contractual structures are good because they allow you to move capital independently. That’s execution feasibility. If two positions aren’t contractually linked, you can stage, price, and negotiate them separately. If they are tightly linked, they must be sequenced carefully to avoid triggering negative signaling or pricing spirals.

Take private credit secondaries as an example. In a GP-led continuation fund, multiple loans might share the same borrower or covenant framework. These positions can’t be sold independently without risk—selling one too fast may depress the perceived value of the others. But if you can separate positions by contract structure—different borrowers, fund vintages, or cash-flow rules—you gain flexibility. That’s the Almgren-Chriss version of correlation: it’s not about smoothing returns—it’s about enabling trade choreography.

This reframing is vital for private markets. When returns depend not just on holding assets but on exiting them under constrained conditions, understanding contractual correlation becomes key. The Almgren-Chriss model, though designed for public markets, helps clarify this point: portfolios are not just optimized for return—they’re sequenced for exit.

In the A/C world, correlation becomes choreography. It doesn’t just tell you what to hold—it tells you how and when to move.

Step 3: When Alpha Becomes Urgency, Not Weight

Fleeting signals don’t shape portfolios—they complicate execution

In traditional portfolio theory, alpha represents the forecasted return above a benchmark. The uncertainty around that forecast is measured separately—typically using tools like confidence intervals or standard errors—and portfolio sizes are adjusted accordingly. The more reliable the alpha, the larger the position; the less reliable, the smaller the bet. Giuseppe Paleologo’s framework follows the same logic: alpha uncertainty is used to adjust position sizing, typically informed by the Information Coefficient (IC). The more confident you are in a forecast, the more capital you commit.

But in the Almgren-Chriss framework, alpha is only meaningful if it translates into a change in execution speed—that is, if it accelerates trading. The model is designed not to evaluate the long-term merits of an asset, but to determine how quickly and cost-efficiently a portfolio transition can occur. So if a signal implies that an asset will outperform in the near term, it should lead to faster execution. Conversely, if the alpha doesn’t create urgency—if acting on it tomorrow is no worse than today—then from an execution standpoint, it doesn’t matter.

This is where the Almgren-Chriss logic diverges sharply from traditional portfolio theory. In classic mean-variance optimization, alpha and volatility are the inputs to a utility function that determines how much of an asset to hold. But in A/C, alpha becomes a temporal signal, not a sizing signal. Its only role is to influence how fast a position is taken—because cost and risk increase the longer you wait.

In this framework, a signal that can’t survive the execution process—because it fades too quickly or is too noisy—isn’t treated as useful information. It’s treated as execution risk. The model has no use for an alpha that disappears before the trade completes, because it only adds volatility without improving outcomes. This is a fundamental reframing: not all information is valuable unless it can be acted on within the real constraints of market microstructure.

In other words, while Sharpe-era thinking builds portfolios around alpha and scales risk exposure accordingly, A/C-era thinking asks whether that alpha merits any execution at all. In this world, alpha only matters if it changes the speed at which you must act. In other words, alpha must earn its place by generating execution urgency—not just incremental return.

This distinction also sheds light on how risk is managed in covenant-heavy asset classes like private credit. In private markets, alpha is rarely a forecasted return stream; it’s embedded in contractual structures and timing pathways. If an investment’s value can’t be monetized swiftly—due to lockups, sponsor delays, or missing triggers—then its informational advantage is moot. As in the A/C model, value arises only when alpha aligns with feasible execution.

Seen this way, alpha uncertainty in private capital isn’t about volatility; it’s about enforceability. A sponsor covenant might signal strong expected performance, but if the exit mechanisms are weak or delayed, the alpha cannot be realized on time. What A/C offers is a different kind of filter: if the “signal” cannot be traded against—now, not in theory—then it’s not useful for sizing or scheduling.

Practitioner takeaway: If your alpha can’t survive execution, it’s not conviction—it’s just volatility in disguise.

Step 4: Cost-Centeric Optimization

When execution cost replaces volatility as the true measure of risk

In traditional portfolio theory, portfolio optimization revolves around the risk-return trade-off. Expected returns are weighed against volatility, and portfolios are constructed to sit on an “efficient frontier” where no further return can be gained without taking on more risk. This model works well in theory—but often falters in markets where the true constraint is not volatility, but liquidity.

The Almgren-Chriss framework redefines this frontier. Here, the key trade-off is not between return and volatility, but between execution speed and cost. The faster you trade, the greater the market impact; the slower you trade, the higher your exposure to price volatility. This creates a new type of optimization problem—one that balances two forms of execution risk rather than balancing performance expectations.

This is where the Almgren-Chriss ratio comes into full focus. Instead of asking “How much risk must I take for this return?” the model reframes the question as “How much impact must I incur to contain volatility?” In this framework, cost becomes a shadow form of risk—not abstract risk, but real, transaction-driven slippage that affects realized performance.

Giuseppe Paleologo’s fourth step aligns closely with this view. After isolating alpha, simplifying covariance, and adjusting for confidence, his model uses execution cost as the governing constraint on trade speed. It’s not just a technical final step—it’s the operational reality check. No matter how elegant your signal or how diversified your holdings, if your trades can’t be implemented without excessive cost, the strategy breaks down.

In private markets, this perspective is invaluable. Unlike public markets, where liquidity is assumed and cost can be modeled continuously, private capital operates in bursts—driven by deal flow, negotiated exits, and timing constraints. Execution feasibility is not a detail; it’s the central design feature. The Almgren-Chriss model, though not directly applicable in form, offers the right lens: portfolios must be built and unwound with an eye toward pacing, not just payoff.

In the world of private credit, this means understanding execution cost not as a bid-ask spread or slippage percentage, but as the discount required to offload exposure, the concessions needed to accelerate an exit, or the structural friction embedded in legal covenants. Cost-aware optimization here is not about choosing the best trading algorithm; it’s about knowing when a covenant can be triggered, how quickly an asset can be syndicated, and what price a liquidity-constrained buyer will pay.

In the Almgren-Chriss world, portfolios are built to execute first and earn later. Liquidity constraints aren’t post-trade headaches—they’re pre-trade inputs, often embedded in the contract itself.

Step 5: Execution and Time Decay

When patience isn’t passive—it’s a dynamic execution strategy

In traditional portfolio theory, execution is treated as the final step—mechanical, after-the-fact, and detached from allocation or optimization. But in both the Almgren-Chriss model and Giuseppe Paleologo’s reinterpretation, execution isn’t a final detail—it’s the convergence point where every part of the portfolio design becomes real. Execution is where alpha gets tested for urgency, where correlation takes the shape of trade sequencing, and where cost isn’t an afterthought—it becomes part of the portfolio’s architecture.

The original Almgren-Chriss model advanced this insight by modeling the execution path itself—not just the end-state portfolio, but the sequence and timing of trades required to build it. They shifted the focus from “what to own” to “how to get there.” Execution wasn’t treated as trivial—it was something to optimize alongside risk and return, not around them. Their core innovation was this: execution isn’t just the last mile, it’s the terrain the portfolio must navigate.

Later research extended this concept. Jim Gatheral and the Obizhaeva-Wang model showed that market impact isn’t instant and done—it decays over time. A large trade doesn’t just move the price now; it affects future prices too. This gives rise to more dynamic execution strategies: trade early when urgency is high, pause when pressure builds, and resume when impact dissipates. Execution becomes time-sensitive, adaptive, and strategic.

Giuseppe Paleologo builds on this view in the final step of his pipeline. For him, execution is where all earlier inputs—alpha, confidence, correlation, and cost—feed into a real-world, time-aware plan. It’s not the caboose of portfolio construction—it’s the control tower. Execution is where abstract signals meet market mechanics.

This insight is especially valuable in private capital markets, where trades don’t happen continuously and pricing isn’t always transparent. In private credit and secondaries, for example, when you sell is often as important as what you sell. Move too soon, and you might take a deep NAV discount. Move too late, and liquidity can dry up. The trade-off isn’t just price—it’s deal readiness, buyer timing, and covenant triggers.

In these environments, execution isn’t about how quickly, cheaply, or discreetly you can trade—it’s about how deliberately you coordinate your exits. Portfolio managers must decide which tranches to offload first, how to time transactions to match buyer constraints, and how to stage secondary sales in a way that preserves contractual terms and avoids unintended ripple effects across the rest of the portfolio. Some situations call for strategic patience; others demand urgency. In this context, execution becomes the governing mechanism of portfolio management—not the final step, but the process that determines how and when value is ultimately realized.

That’s where Almgren-Chriss—even with its public market origins—offers real value. Its emphasis on execution feasibility over theoretical return speaks directly to how private capital operates. Its framework for time-sensitive execution gives private credit managers a way to think about staggering liquidity, designing better exit windows, and managing contractual timing constraints across an entire portfolio: trade early when urgency is high, pause when pressure builds, and resume when impact dissipates.

In short: execution is no longer just a step. It’s “the” portfolio management strategy.

When impact fades over time, execution becomes where the portfolio is made—not just how it’s implemented. Waiting isn’t delay—it’s strategy. In both public and private markets, patience becomes part of the portfolio design itself.

Summary: Rethinking Private Credit Through Execution-Centric Portfolio Design

This piece repositions the Almgren-Chriss model not just as a tool for public market execution but as a lens for interpreting risk and value in private credit—especially in covenant-intensive strategies and secondaries. In traditional models, execution is the final step, a technical matter of trade scheduling. But through the Almgren-Chriss framework—interpreted via Giuseppe Paleologo’s pipeline—execution becomes the starting point: the stage where market impact, volatility, liquidity constraints, and timing all converge to shape not just how, but whether a portfolio can be constructed.

In private credit, where exit timing is uncertain, buyer availability is episodic, and covenants define economic rights, this shift is crucial. The value of a portfolio is no longer solely about yield—it’s about how executable that yield is under real-world constraints. Covenant structures that allow for reliable enforcement, negotiated exit pathways, or control over collateral are not just risk mitigants—they are execution facilitators. Their design affects not just portfolio durability, but whether and when value can be realized.

The Almgren-Chriss model reminds us that in these markets, execution feasibility isn’t a constraint on portfolio theory. It is portfolio theory.

References

• Almgren, R., & Chriss, N. (2000). Optimal execution of portfolio transactions. Journal of Risk, 3(2), 5–39.

• Almgren, R., & Chriss, N. (1997). Optimal Liquidation. Available at SSRN: https://ssrn.com/abstract=53501 or http://dx.doi.org/10.2139/ssrn.53501

• Gatheral, J. (2010). No-dynamic-arbitrage and market impact. Quantitative Finance, 10(7), 749–759.

• Obizhaeva, A., & Wang, J. (2013). Optimal trading strategy and supply/demand dynamics. Journal of Financial Markets, 16(1), 1–32.

• Paleologo, G. (2025). Simpler (Portfolio) Heuristics That Make Us Smart(er). Columbia University Talk.

• Paleologo, G. (2025). The Elements of Quantitative Investing. Wiley Finance Series.

• Staff, P. (2023). The covenant economy: How contract structures shape outcomes in private credit. Private Credit Review, 12(4), 14–25.

• Metrick, A., & Yasuda, A. (2010). The economics of private equity funds. Review of Financial Studies, 23(6), 2303–2341.

• DeMarzo, P., & Duffie, D. (1999). A liquidity-based model of security design. Econometrica, 67(1), 65–99.

• Townsend, R. R. (2022). Contract enforceability and deal execution in private markets. Journal of Alternative Investments, 24(2), 77–91.

This piece emerges from an ongoing research dialogue with my RA, Kai Owen, as we explore the structural foundations of private credit markets. It is part of a broader academic project on the architecture of private capital, and also serves as intellectual groundwork for the HFA–Market Microstructure Symposium, which we are co-organizing at NYU Stern on October 24, 2025.

The symposium will feature two panels dedicated to private capital:

• Panel I — Beyond Liquidity, Before Maturity: Closing the Infrastructure Gap in Private Secondaries

• Panel II — Silent Partners: Following Banks Through the Architecture of Private Credit

Together, these panels examine how risk, return, and intermediation are being reshaped in private markets—not just through capital flows, but through the evolving microstructure of contracts, infrastructure, and institutional roles.

Executive Summary

Private credit yields more than high-yield bonds—not because the borrowers are riskier, but because the deals are structurally more complex. This paper introduces the concept of the covenant risk premium: the return investors earn for navigating bespoke, non-standardized contracts where enforceability, rights, and remedies vary dramatically.

As private secondaries scale, the market is moving from a liquidity phase—solved with tools like NAV loans and continuation vehicles—into an infrastructure phase, where clearing and standardization become essential. At that point, the key challenge won’t be capital or pricing—it will be understanding what’s actually inside the deal.

The better analogy is no longer “private credit as debt,” but “private credit as derivatives.” Like swaps before them, these bespoke instruments will require shared infrastructure to scale. Without it, the covenant risk premium will remain high—and the market will remain structurally constrained.

In short, private credit doesn’t pay more because it’s riskier—it pays more because it’s harder to understand.

The Covenant Risk Premium: Rethinking Return Drivers in Private Credit

Why do private credit loans yield 200 to 400 basis points more than comparable high-yield bonds, despite exhibiting lower historical default rates and less interest rate sensitivity? The answer, we argue, lies in a hidden but critical return driver: the covenant risk premium.

This premium doesn’t reflect higher credit risk or greater exposure to macro volatility. It reflects the deal-by-deal complexity of private credit lending—the fact that each loan is bespoke, privately negotiated, and governed by covenants that vary widely in structure, enforceability, and transparency. These legal and operational features are not captured by models or ratings. They must be read, understood, and underwritten at the document level. In short, private credit investors are paid not just to hold risk—but to navigate contract complexity.

A Missing Piece: The Microstructure of Covenant Risk

Private credit deals lack the standardization that defines public markets. Unlike syndicated loans or bonds, which come with familiar covenant packages and disclosure templates, private loans are structured case by case. Default triggers, collateral rights, financial reporting terms, and enforcement mechanisms are all negotiated—and often re-negotiated—on a one-off basis.

This idiosyncrasy makes each loan structurally unique. Pricing the risk of a given deal is less about modeling expected loss, and more about understanding whether the lender has real recourse when something goes wrong. There is no off-the-shelf analytics engine for this kind of underwriting. The premium investors demand is compensation for the non-fungibility and opacity of these structures.

From Liquidity Solutions to Covenant Risk: Secondaries as a Two-Stage Market

This insight also helps illuminate the evolving role of secondaries in private markets. The growing secondaries market has often been framed as a solution to illiquidity—and in its current form, that’s exactly what it is. But the type of illiquidity it addresses is specific.

In what we might call Stage One of secondaries development, the core risks are market illiquidity (the difficulty of selling a position) and funding illiquidity (the inability to raise capital against private assets). The market has developed practical tools to solve both. Syndication, securitization, and continuation vehicles have provided pathways for transferring exposures. Meanwhile, NAV loans, subscription lines, and hybrid facilities have attracted banks and non-bank lenders eager to finance these assets. In this phase, familiar financial engineering techniques have enabled the market to mimic liquidity without requiring deep structural reform.

But Stage Two of secondaries will look very different. As deal flow expands and the market matures, secondaries will increasingly require standardized clearing and settlement infrastructure. That’s when the market will encounter a deeper and less familiar challenge: the covenant risk embedded in these assets. At this point, the relevant question is no longer “Can this asset be bought or funded?”—but rather, “Do we understand what’s inside the deal?”

This is the pivot point where covenant risk becomes central. If secondaries are to scale into a true, institutional market, they will require infrastructure capable of handling the heterogeneity and legal complexity of private loan agreements. Without such infrastructure—whether digital, legal, or administrative—these markets cannot clear efficiently, nor can they price risk transparently.

In this sense, the secondaries market is approaching the same kind of turning point that once reshaped the derivatives market. For years, swaps and other OTC derivatives operated under bespoke bilateral agreements—they were treated, contractually and operationally, as private deals. But once volume, counterparty exposure, and systemic entanglement grew, the industry pivoted toward standardization and clearing, laying the foundation for the centralized infrastructure we now take for granted.

As secondaries shift from solving liquidity to confronting clearing, the better framework is derivatives—not debt. Like swaps before them, bespoke private credit deals must move toward infrastructure that can process legal and structural complexity at scale.

The analogy matters. In the liquidity phase (Stage One), private capital can still be understood as “debt”—a credit instrument that is underwritten, traded, and financed in bilateral ways. But as secondaries grow in complexity and scale, and as we enter the infrastructure phase (Stage Two), the better framework may be that of derivatives: contracts whose value and risk cannot be priced or transferred without understanding their legal structure and institutional ecosystem.

The real bottleneck is no longer liquidity—it is contractual complexity, and the absence of infrastructure to map and manage it. That is what makes covenant risk not just a feature of private credit, but the defining obstacle to the next stage of secondaries market evolution.

Origins in Special Situations Investing

This kind of structural diligence is not new. It has deep roots in special situations investing, where teams were assembled to read complex contracts, map capital structures, and identify overlooked risks in distressed or off-the-run credit. These groups didn’t specialize in forecasting defaults—they specialized in understanding documentation.

Private credit today is, in many ways, an institutionalization of the special situations approach—but with far more capital deployed, and far fewer hours spent per deal. What was once a niche, high-touch activity has become the mainstream engine of private lending, without the institutional apparatus fully evolving alongside it.

A New Name: The Covenant Risk Premium

This reality calls for more than just a better yield model—it demands a new conceptual framework for how we understand return generation in private credit. The excess yield observed in these markets is not, as some assume, a byproduct of traditional fixed-income factors.

It is not a term premium, because most private credit loans are floating-rate and carry minimal interest rate risk. It is not primarily a credit premium in the conventional sense, because default rates—especially in the middle market—are often lower than in the public high-yield universe. And while liquidity risk does play a role, much of it is now engineered around through continuation funds, NAV facilities, and fund-level leverage.

Instead, the defining feature of private credit is its structural complexity. These are not just loans—they are contracts: dense, negotiated agreements with bespoke covenants, variable enforcement rights, uneven disclosure, and legal terms that must be interpreted in context, not just read at face value. And when stress hits, outcomes are driven not by bond math but by how these contracts actually function—who has rights, what counts as a breach, and how quickly value can be defended or recovered.

That is what gives rise to what we call the covenant risk premium: the return investors demand for absorbing structural uncertainty, contractual asymmetry, and documentation friction. It is the premium for underwriting not just financial performance, but legal viability—under stress, in default, and through restructuring. It is paid to those who can decode complexity before it becomes costly.

This premium has several defining features. It is not diversifiable, because each deal is idiosyncratic—defined by its own terms, jurisdictions, counterparties, and negotiation history. It is not easily modelable, because its drivers live in legal clauses, side letters, and operational enforcement mechanics—not in time series data. And it is not always visible, even to those inside the deal, because what matters is not just what is written, but how the pieces interact in practice, especially under downside scenarios.

In effect, private credit returns are not a product of capital market beta—they are a form of contractual alpha. The investor’s edge is not just picking good credits; it’s knowing which documents to trust, which covenants to enforce, and how to do so when the time comes. The covenant risk premium is what bridges the gap between high-yield returns and private credit yields—without relying on worse fundamentals or more market risk.

Recognizing this premium for what it is reframes the discussion not only about pricing and performance—but also about the market infrastructure required to support it. If the primary risk in private credit is embedded in legal structure, then the ecosystem must eventually evolve to track, monitor, and intermediate contractual complexity—just as public markets built tools to manage price volatility and credit risk.

Why It Matters for Microstructure

Understanding the covenant risk premium changes how we think about private market microstructure. It explains why these markets have failed to develop secondary trading platforms despite their size. It clarifies why post-trade infrastructure—from covenant monitoring to compliance tracking—is now critical for sponsors, lenders, and LPs alike. And it helps explain why banks are increasingly involved not just as lenders, but as structuring partners and infrastructure builders.

This isn’t just about capital—it’s about scaffolding. Just as early-stage secondaries used engineering to overcome illiquidity, the next stage will require a different kind of market architecture: one capable of aggregating, tracking, and enforcing covenant-level complexity at scale.

Conclusion

Private credit is often described as a high-yield alternative. But that framing misses what is most distinctive about the asset class. These are not just loans—they are engineered contracts, embedded with custom structures, non-standard covenants, and legally complex terms. The risk premium investors earn is not primarily for taking on macroeconomic or credit risk, but for underwriting structural uncertainty.

This is the covenant risk premium, and it is central to how both private credit and secondaries markets function today. As these markets mature, understanding and managing this premium will determine not only returns—but also the scalability, transparency, and systemic resilience of the ecosystem itself.

References

1. Cliffwater LLC. 2023 Annual Private Credit Fund Performance Report.

2. PitchBook LCD. Private Credit Default Rate Benchmarks, 2010–2023.

3. Preqin. Private Credit Fund Terms and Structures, Q1 2024.

4. BlackRock Alternatives. Understanding NAV Lending: A Growing Toolkit in Private Markets.

5. Oaktree Capital Management. Special Situations Strategy Overview, Investor Presentation 2021.

6. Coller Capital & ILPA. 2023 Global Private Equity Secondary Market Report.

7. Bank of America Research. Private Credit and the Role of Structuring Banks in the Secondary Market, 2024.

8. Brookings Institution (Victoria Ivashina & Josh Lerner). The Evolving Role of Private Credit in the Capital Markets, 2022.

9. SEC Division of Investment Management. Private Funds Advisor Study, 2023.

10. Duff & Phelps. Valuation Challenges in Illiquid Credit Markets, 2023.

Personal Note

My first job after grad school was at the Milken Institute, where I focused closely on capital markets and credit. That experience left a lasting impression—especially Milken’s approach to structuring and credit as tools for transformation.

Private credit has since become a central theme in my work: it’s a chapter in my upcoming book, the subject of a paper I’m writing, and the focus of two panels I’m organizing at the upcoming Market Microstructure Symposium (in collaboration with the Hedge Fund Association).

The more I’ve explored this world, the more I’ve found that Milken’s framework—his way of thinking about high-yield credit—is still one of the best tools for understanding today’s private capital markets. This piece is a sketch of that evolving framework, shaped with help from two exceptional RAs: the fantastic Kai Owen and the fascinating Itai Dreifuss.

Executive Summary

Private credit has transitioned from a niche financing option into a core pillar of modern capital markets. Yet despite its rapid growth, it remains loosely defined and analytically underdeveloped. Discussions often revolve around what it lacks—such as transparency, standardized ratings, or daily liquidity—rather than what internal logic drives its evolution.

This paper proposes that the most coherent framework for understanding private credit is rooted in the intellectual structure originally developed around high-yield bonds. Drawing on insights from Michael Milken’s reframing of credit markets, the paper introduces three structural pillars that explain how private credit operates: Dynamic Risk Sharing, Convexity Thinking, and Capital Velocity.

These pillars shed light on how private credit allocates risk, anticipates market movements, and creates liquidity not through trading, but through engineered structures. By integrating these elements, private credit emerges not as a marginal asset class but as a parallel credit system—one that is increasingly central to modern financial architecture.

Too often, private credit is defined by its absences—no ratings, no benchmarks, no daily liquidity—instead of by the structural logic that makes it work.

---

Introduction

Private credit markets have expanded rapidly over the past two decades, encompassing a range of instruments such as direct lending, asset-based finance, mezzanine loans, special-situations credit, and NAV-based lending. These markets now account for trillions in global assets under management. Yet for all this growth, private credit remains conceptually under-theorized. It is often described by what it lacks—public disclosure, ratings, exchange trading—rather than by what drives its internal logic.

Credit is not just a priceable risk—it is a designable structure.

---

To better understand this space, I turn to a foundational insight: that credit is not merely a priceable risk, but a designable structure. This principle, core to Milken’s approach to junk bonds, also underlies the architecture of today’s private credit markets. By focusing on structure rather than surface, we can identify three core mechanisms that define private credit’s logic:

1. Dynamic Risk Sharing through Structuring

2. Convexity Thinking over Yield Curve Thinking

3. Capital Velocity as a Source of Engineered Liquidity

Pillar 1: Dynamic Risk Sharing Through Structuring

Public bond markets typically rely on standardized issuance and credit ratings. Once a bond is issued, its risk profile is largely fixed, and investor focus shifts to managing exposure via trading or hedging.

Private credit turns this sequence around. Risk is not passively received; it is actively structured into the deal. In financial engineering terms, structuring refers to the process of designing a transaction’s components—such as cash flow rights, covenants, collateral arrangements, and payment hierarchies—to allocate and manage risk across participants. Features like floating-rate instruments, performance-based triggers, bespoke covenants, collateral packages, and subordinated tranches allow investors to customize how different types of risk are absorbed, transferred, or mitigated throughout the life of the deal.

This reflects a fundamental principle in modern portfolio theory: diversification and risk allocation are most effective when incorporated during the portfolio construction phase, not after the fact. Applied to private credit, this means that risk isn’t something to be managed reactively in secondary markets—it’s something to be actively structured at the outset.

Through tools such as covenants, tranching, interest rate resets, payment waterfalls, and collateral layering, lenders and borrowers co-engineer credit transactions that distribute different forms of risk—credit risk, interest rate risk, liquidity risk, and operational risk—across counterparties in a deliberate and pre-agreed manner.

For example, senior lenders might receive tighter covenants and priority claims, while subordinated tranches absorb more downside risk in exchange for higher return potential. Interest rate mechanisms might be adjusted to pass along market movements to borrowers, and collateral arrangements are crafted to secure exposure in ways that reduce the need for later renegotiation or hedging.

Where public markets price risk and manage it later, private credit designs the risk itself—right into the bones of the deal. In other words, public markets hedge risk. Private credit structures it.

This stands in contrast to public debt markets, where the risk premium is priced and fixed at issuance, and risk management typically occurs later—through trading, hedging, or secondary market adjustments. In private credit, by contrast, the structure of the debt itself—through tranching, covenants, NAV-based features, and other engineered mechanisms—acts as the de facto form of risk management.

This mirrors an important hierarchy in modern portfolio theory: risk-sharing belongs to the construction phase of the portfolio, while risk management occurs after the portfolio is formed. Private credit collapses that distinction. Because structuring dynamically shares and distributes risk upfront, the need for active, reactive risk management later is often significantly reduced. What emerges is a bespoke credit profile that reflects both the borrower’s constraints and the investor’s return and risk preferences—aligned through intentional design rather than aftermarket correction.

Pillar 2: Convexity Thinking Over Yield Curve Thinking

Traditional fixed income models rely on static risk assessment frameworks—specifically, the risk structure and the term structure of interest rates. The risk structure explains how credit risk affects interest rates across different borrowers at a single point in time, while the term structure captures how interest rates vary with the length of time to maturity. These frameworks underlie the predominance of fixed-income instruments in public debt markets, where most securities are issued with fixed coupons and fixed terms to maturity. This structure simplifies valuation, enables benchmark comparison, and supports liquid secondary trading. Together, these models assume relatively stable relationships between time, credit quality, and return, emphasizing current pricing over evolving risk dynamics.

Private credit isn’t about where the spread is—it’s about where it’s going.

---

Private credit practitioners think differently. They often focus not on the absolute level of spreads, but on the potential for spreads to compress—that is, the narrowing of yield differentials between risky and less risky credits as market conditions or issuer fundamentals improve. In private credit, this often occurs when a borrower transitions from a distressed or unrated state to a more stable and creditworthy position, either through operational improvement or strategic restructuring. Alternatively, investors may engineer spread compression by embedding credit enhancements—such as tighter covenants, improved collateral coverage, or subordinated capital layers—into the deal structure itself. Spread compression becomes a key value driver because it enables investors to earn returns not only from the contractual interest payments but also from mark-to-market gains as the borrower's perceived credit risk declines.

For example, consider a private lender who provides financing to a company undergoing a turnaround. Initially, the loan may carry a spread of 800 basis points over SOFR due to high perceived risk. As the company improves operations, secures additional equity backing, and begins to meet performance covenants, the market may reprice the loan to reflect only a 500 basis point spread. Even if the lender continues to hold the loan, the improvement in perceived creditworthiness increases the market value of the asset—generating unrealized gains that can be monetized through securitization, secondary sale, or as part of NAV-based borrowing. These repricings are often triggered by operational turnarounds, upgrades by internal or external credit assessments, shifts in macroeconomic sentiment, or capital market arbitrage opportunities. This is convexity thinking: a sensitivity to second-order effects, where small changes in underlying risk or behavior can produce outsized returns.

Milken’s original insight—that undervalued credit can reprice as the market re-evaluates risk—remains central. Investors are not just being paid for taking risk; they are betting on how risk is likely to change over the life of a deal. This requires underwriters to account for path-dependence, optionality, and embedded catalysts.

Path-dependence refers to how the future evolution of a credit position depends on the specific sequence of events or performance milestones along the way, rather than just its current status. Optionality reflects built-in features that allow borrowers or lenders to adjust terms under certain conditions—such as early repayment rights, step-up coupons, or performance-based pricing. Embedded catalysts are deal features or external events (like M&A, refinancing, or regulatory changes) that can trigger a revaluation of risk. Structuring in private credit is therefore not just about managing today's exposures but about anticipating and harnessing these future inflection points to create value.

Milken’s original insight remains central: in credit, the payoff often lies not in taking risk, but in anticipating how it will change. That’s why private credit underwriters don’t just price static exposure—they model path-dependence, build in optionality, and structure around embedded catalysts.

---

Pillar 3: Capital Velocity Engineered Liquidity Over Market Liquidity

Private credit lacks the trading-based liquidity of public markets. But that doesn’t mean it lacks liquidity altogether. Instead, liquidity is generated through capital velocity—the ability to recycle, refinance, and repackage capital across multiple transactions. In public markets, liquidity depends on the existence of continuous secondary trading, supported by standardized terms and transparent pricing. In contrast, private credit markets create liquidity by embedding exit mechanisms into deal structures themselves—such as refinancing triggers, repayment schedules, or embedded take-out provisions. The emphasis shifts from being able to sell an asset quickly on a market to designing the asset in a way that ensures capital can be turned over and re-deployed efficiently. This structural design allows institutions to maintain portfolio agility and support new deal flow, even in the absence of deep secondary markets.

Private credit doesn’t depend on trading for liquidity; it manufactures it through capital velocity.

---

This is achieved through a variety of mechanisms that enable credit risk to be redistributed (linking the third pillar with the first pillar and securitization), capital to be recycled through secondaries (linking the third pillar with the second pillar, NAV loan, etc), and liquidity to be engineered:

• Loan securitization (e.g., CLOs, ABS): Bundling and selling loan exposures as tradable securities, allowing originators to remove risk from their balance sheet and reinvest proceeds.

• Re-securitization and tranching of structured products: Creating new layered securities out of existing structured products to appeal to different risk-return profiles and optimize capital efficiency.

• Syndication and LP-led secondary sales of private loan exposures: Selling or sharing exposure to existing loans among multiple investors to free up balance sheet capacity and create liquidity.

• GP-led secondaries and NAV-based facilities in private equity-backed credit: Providing exit or recapitalization options to existing investors through asset-level refinancing or lending against fund NAV, extending the capital lifecycle and injecting liquidity into otherwise illiquid positions.

In this model, liquidity is not an external market feature—it is endogenously created. Instruments are structured with exit pathways (e.g., take-outs, refinancings, repayments) embedded into the deal itself. This reflects a shift from market-based liquidity to design-based liquidity, where financial innovation substitutes for market depth.

Conclusion: A Structural Theory for a Parallel Credit System

Together, these three pillars provide a cohesive framework for understanding private credit as a structurally distinct domain of finance.

• Dynamic structuring replaces standardized issuance.

• Convexity thinking replaces yield curve analysis.

• Capital velocity replaces market-based liquidity.

Private credit is not merely "illiquid public debt"—it is a parallel credit system optimized for customization, adaptability, and innovation. Drawing on Milken's original insight—that credit is a tool to be shaped rather than a static product—modern private credit continues to push the boundaries of how financial capital can be structured and deployed.

Understanding this system is essential for investors seeking alpha, for regulators managing systemic risk, and for scholars aiming to capture the financial architecture of the 21st century.

In this interview, Economist Elham Saeidinezhad sits down with Matthew Schrager, Managing Director and Co-Head at TD Securities Automated Trading, to explore how modern market makers navigate today’s fast-evolving fixed-income landscape.

Schrager unpacks TD Securities’ hybrid approach, which blends traditional market-making with proprietary trading elements, leveraging automation, ETFs, and derivatives to optimize liquidity provision, capital efficiency, and risk management. He highlights how market makers operate in an increasingly electronic and regulatory-constrained environment, where speed, balance sheet optimization, and strategic positioning are key to staying competitive.

One central theme of the discussion is ETFs’ role in transferring risk. Schrager explains how market makers dynamically manage risk using long and short positions, allowing one side of the book to finance the other and reducing reliance on external funding. While this principle applies broadly to market-making balance sheets, ETFs play a key role by providing liquid instruments that facilitate efficient hedging. This, combined with the blended funding cost of dealing banks, reduces dependence on any single risk-free rate benchmark while reinforcing ETFs’ role in market structure.

Another key takeaway from this discussion is that such new models of market-making should not be mistaken for proprietary trading. While the firm leverages technologies traditionally associated with prop trading—such as automation, predictive alpha, and low-latency execution—it does so in service of liquidity provision, not directional speculation. The aim is not to take risk, but to price, execute, and manage that risk more efficiently in real time.

Beyond trading mechanics, the conversation highlights broader structural transformations in market-making. Schrager discusses the shift from principal trading (dealing) to agency execution (brokerage), the expanding role of algorithmic strategies, and how firms like TD Securities, which blend traditional bank-based market-making with technology-driven high-frequency trading, are stepping in to fill the liquidity gap left by traditional bank dealers.

His insights provide a rare look into the evolving intersection of liquidity provision, risk-taking, and technological innovation, illustrating how modern market makers are redefining fixed-income trading in a post-crisis regulatory landscape.

This is their conversation:

Elham Saeidinezhad (ES): Matthew, thank you for joining us. It’s great to have you as part of the Market Microstructure Project. Before we dive into market making in fixed income, could you briefly outline the TD Securities Automated Trading core business?

Matthew Schrager (MS): We are a liquidity provider and market maker specializing in fixed-income instruments, particularly corporate and municipal bonds. Every day, thousands of firms need liquidity in these markets, and our role is to meet that demand in all its forms. This includes executing cash bond transactions, responding to hundreds of thousands of RFQs (requests for quotes), and facilitating trades both directly with clients and anonymously on trading platforms. Beyond bonds, we also provide liquidity in ETFs, ensuring clients can efficiently access fixed-income markets through multiple channels. We also plan to expand into total return swaps and other derivatives in order to provide a full-service offering across all the main forms of risk transfer in the broader market ecosystem. .

ES: Could you briefly explain what a total return swap is and why it is important for fixed-income end users?

MS: A total return swap (TRS) is a contract where one party pays another the total return on an asset without trading the asset directly. .

Widely used in fixed-income markets, TRSs replicate bond exposure without ownership, allowing investors to achieve the same risk profile but with leverage. This leverage enhances capital efficiency and, when used effectively, can generate higher risk-adjusted returns. Institutional investors utilize derivatives like credit default swaps (CDS) and TRSs for their ability to mirror bond exposures while optimizing capital deployment and potential returns.

ES: Could you explain how leverage works in credit derivatives like total return swaps?

MS: Let me answer this question through an example. Consider purchasing $1m worth of exposure to a particular bond. In a standard bond purchase, you pay the full $1 million upfront and take ownership. With a total return swap (TRS), you achieve the same exposure with leverage—you post a margin, say $200,000, while the market maker holds the bond and delivers you its total return.

The remaining $800,000 is effectively loaned to you, and you pay interest on that amount. This structure allows investors to gain bond exposure with less capital, enhancing capital efficiency. For banks with large balance sheets, offering this leverage is profitable while managing associated risks, making TRSs a mutually beneficial tool for both parties.

ES: Before moving on from this concept, how do ETFs, TRSs, and CDSs fit into your market-making strategies alongside traditional cash bond trading?

MS: Conceptually speaking, there is a hierarchy of instruments, when it comes to the risk-adjusted returns, in the fixed income market. In this market, bonds remain the gold standard for achieving direct exposure, offering investors the exact risk and return profile they seek. However, bond ownership requires capital and direct balance sheet usage, leading investors to explore alternative instruments with correlated exposures.

At the first level of alternatives, fixed income ETFs have emerged as a dominant vehicle, offering bond-like exposure with enhanced liquidity and accessibility, despite concerns over their underlying market depth. Their rapid growth—from niche products to trillions in assets—reflects this demand. Fixed income ETFs function as structured packages of bonds with built-in arbitrage mechanisms, allowing conversion between ETFs and their underlying securities. The fact that they trade as equities makes them an especially appealing option for retail investors, who are more familiar with equity trading mechanics than bonds.

Beyond ETFs, investors turn to credit derivatives such as credit default swaps (CDS) and total return swaps (TRS). These instruments provide exposure to fixed-income risk and returns without requiring direct ownership of the bonds, making them attractive for capital efficiency, leverage, and hedging. They can also provide a convenient packaging mechanism for more complicated trades, such as delivering the return of an index rather than that of a single bond.

ES: Given this hierarchy of exposures, how do market-makers provide liquidity across these instruments?

MS: Fixed-income market makers provide liquidity by ensuring seamless execution across cash bonds, ETFs, and derivatives, allowing investors to access fixed-income exposure in whichever form they prefer. While cash bonds remain the foundation, ETFs serve as both trading instruments and hedging tools, offering flexibility and liquidity. Derivatives like total return swaps (TRS) and credit default swaps (CDS) provide leveraged exposure and balance sheet efficiency, making them important tools for institutional investors. Market makers bridge these markets, ensuring liquidity flows smoothly across all levels of fixed-income trading.

ES: ETFs are the primary alternative to cash bonds. Beyond providing liquidity, do market makers use ETFs to hedge their own exposures? If so, how?

MS: ETFs play a dual role in fixed-income market-making, serving as both trading instruments and risk management tools. In a mature market-making business, the line between trading for profit and hedging is fluid—our objective is to optimize Profit and Loss (P&L) while staying within risk constraints set by capital limits, internal controls, and risk policies.

P&L represents the net result of trading activities, factoring in execution costs, hedging expenses, market fluctuations, and revenues generated from liquidity provision. Successful market-making requires balancing risk and profitability—ensuring hedging costs do not erode trading revenues while still maintaining exposure control.

ES: You emphasize “exposure control” over “risk elimination,” diverging from financial theories that equate perfect hedging with creating a synthetic risk-free portfolio. Could you elaborate on this distinction?

MS: In practice, risk management is about dynamic hedging, not full elimination. ETFs provide an efficient way to adjust exposure in real time. If we accumulate a long bond position, we can hedge by shorting a correlated ETF, allowing us to manage risk while continuing to provide liquidity.

More broadly, we maintain a balanced long-short portfolio, where offsetting positions naturally reduce risk and funding needs. As you highlighted in your recent work on ETFs as alternative risk transfer mechanisms, their liquidity and flexibility make them essential tools for managing fixed-income exposures.

ES: So ETFs allow market makers to fine-tune risk exposure dynamically. Do they serve other functions beyond hedging?

MS: Yes. If a bond is trading rich (i.e., trading at premium), we can short it and hedge with a long position in a similar ETF or an undervalued bond. ETFs offer an efficient arbitrage mechanism—if we’re long bonds and short the corresponding ETF, we can use the issuer’s creation-redemption process to exchange bonds for ETF shares and close our position. This liquidity transformation makes ETFs a powerful risk management tool.

This hedging mechanism can also be viewed through the lens of relative value. For example, if an ETF trades at a premium to its underlying bonds, we might short the ETF while going long the bonds to capture the arbitrage opportunity.  This is simultaneously a hedged position and a relative value trade. In this way, the creation-redemption mechanism corrects pricing dislocations, allowing for efficient market pricing.

ES: That’s a fantastic explanation—it clarifies how ETFs create a smart, automatic hedge through the AP mechanism and why market makers use them for both hedging and arbitrage.

ETFs clearly play a crucial role in your risk management and trading strategies. Stepping back to the broader business model, TD Securities operates at the intersection of market making and proprietary trading. How do you balance these approaches while maintaining your role as a liquidity provider?

MS: That’s a great question. Traditionally, market making focused purely on providing liquidity, while proprietary trading involved taking speculative positions. However, as markets have evolved, this distinction has become less rigid.

Today, successful market making requires an element of predictive modeling—what we call predictive alpha—the ability to forecast short-term price movements, liquidity shifts, and client order flow with high precision. By leveraging real-time data, order book dynamics, and historical trading patterns, we can anticipate how spreads will evolve and adjust pricing accordingly. This isn’t about taking outright directional bets but rather optimizing execution strategies to minimize adverse selection and improve trade profitability.

Profitability in this business hinges on two primary factors: operational efficiency and risk management. A naive perspective might suggest that narrower bid-ask spreads hurt market makers by reducing per-trade margins. However, this is only problematic for firms with inefficient cost structures.

In a competitive market, firms that streamline operations—reducing reliance on manual trading and maximizing automation—can sustain profitability even with compressed spreads. This is why algorithmic market-making is dominant today. By combining predictive analytics with automation, we can execute trades more efficiently, manage risk dynamically, and continue providing liquidity while maintaining profitability.

ES: Your firm’s cost structure allows you to compete even as spreads compress. How does automation factor into that?

MS: Exactly. We operate more like a tech company than a traditional trading firm. About 75% of our staff are software engineers and quantitative researchers—not traditional traders.

This heavy focus on technology enables us to automate and optimize as much of our business as possible. When we started, our team was just 10-15 people, yet we became the largest liquidity provider in the municipal bond market. In contrast, the cost structure of a traditional bond trading business involves hundreds of individuals across sales, trading, software development, quantitative research, operations, etc., which becomes unsustainable as spreads shrink. By leveraging automation, we can provide tighter spreads while remaining profitable.

ES: That’s fascinating. So, automation and technology allow you to scale in ways traditional firms can’t. Would you say this makes TD Securities more like a high-frequency trading (HFT) firm, but within fixed income?

MS: That’s correct, but it’s only part of the picture. In fixed income, we apply principles from low-latency, high-frequency trading (HFT) while adapting to market structures that differ significantly from equities.

The term “latency” refers to the speed with which one reacts to incoming market data. In equities, market-makers can react in nanoseconds - far faster than the blink of an eye. While such extreme latencies are not required for fixed income, we have adapted many of the high-level techniques from that world for use in our approach. This ultimately allows us to adjust quotes, tighten spreads, and improve efficiency in ways that are difficult for others to replicate, especially those who rely on more manual pricing and slower reaction times.

ES: So, in essence, automation isn’t just about cost efficiency—it also enhances execution quality and competitiveness.

MS: Exactly. The more we automate and optimize, the lower our costs and the more competitive we become. It’s about both scalability and precision—leveraging technology not just to reduce overhead but also to make better trading decisions and provide superior liquidity.

ES: You mentioned how low-latency trading and automation are transforming fixed-income market making, bringing it closer to high-frequency trading in equities. Yet, bond markets have historically been much slower to electronify. What structural or market-specific factors have contributed to this lag?

MS: Several structural factors have historically slowed the electronification of bond markets.

First, market fragmentation presents a significant challenge. Unlike equities, where there are only a few thousand stocks, the fixed-income market consists of tens of thousands of issuers, each with many unique bond offerings that trade infrequently. As an example, the municipal bond market alone has more than one million individual securities. This diversity makes standardization difficult, reducing the feasibility of centralized, automated trading.

Second, data limitations have hindered automation. Bonds have traditionally been traded over the counter (OTC), resulting in limited price transparency. Unlike centralized exchanges in equities, the decentralized nature of bond trading makes it difficult to aggregate and analyze trading data in real time, slowing the adoption of algorithmic execution.

Third, capital constraints differentiate fixed-income market makers from their equity counterparts. High-frequency trading (HFT) firms in equities can operate with minimal overnight exposure, allowing them to trade rapidly without holding large inventories. In contrast, fixed-income market makers must warehouse significant inventory, requiring substantial capital and balance sheet capacity to facilitate liquidity. These capital requirements create an additional barrier to electronification.

That said, electronification is rapidly transforming the fixed-income space. Trading platforms are aggregating liquidity and data, improving execution quality and enabling greater automation. As more fixed-income instruments become electronically tradable, we expect further efficiency gains, much like those seen in equities over the past two decades.

ES: As bond markets become more electronic and trading speeds accelerate, risk management becomes even more critical. That is why I want to return to the issue of risk management again:

 Unlike equities, where positions can be quickly adjusted, fixed-income market makers often hold significant inventory due to the capital-intensive nature of the business. Given these constraints, how do you manage inventory risk?

MS: Risk management is at the core of our business, embedded directly into our market-making operations. We don’t treat it as a separate function but as an integral part of our pricing and execution strategy. Our trading algorithms continuously assess market conditions and dynamically hedge positions in real time.

For instance, when we take on inventory in a less liquid bond, our system adjusts pricing to reflect the anticipated holding cost and liquidation risk. At the same time, we actively seek natural offsetting flows to minimize exposure efficiently. This ensures that we can provide liquidity without accumulating excessive directional risk, keeping our market-making operations both sustainable and competitive.

ES: Trade size plays a crucial role in offsetting trades that aim at controlling exposures. Do you specialize in a particular type of trade? Do you focus more on block trades or odd lots?

MS: Over time, the balance between institutional block trades and smaller trades has shifted as technology evolves.

Historically, large orders were executed as blocks via voice trading. Today, institutions increasingly break them into smaller trades and execute them algorithmically—just as happened in equities. As a result, average trade sizes have declined while the number of trades has increased. More clients now send small trade sizes through electronic platforms instead of executing as a single block.

ES: How does this trend impact execution quality?

MS: It’s had a major impact. In equities and most other asset classes, retail traders benefit from tight spreads and minimal market impact. But in fixed income, historically, it was the opposite—institutions got better execution than retail traders.

ES: Why was that the case?

MS: Before fixed-income markets became electronic, retail orders were too small to justify a human trader’s time. A $1,000 bond trade might only generate a few dollars in spread profit, making it uneconomical to handle manually. To compensate, dealers widened spreads significantly, leading to poor execution for smaller trades.

With automation, however, the cost of quoting has dropped to near zero. As algorithmic trading has taken over, execution for retail-sized trades has improved dramatically.

ES: So in some cases, retail execution is now as good as—or even better than—larger trades?

MS: Exactly. In equities, this transformation happened earlier, but we’re now seeing the same trend in fixed income. As technology advances, execution quality will continue to improve.

ES: As technology reshapes fixed-income market making, we see a shift in execution models. Automation and cost efficiency have enabled tighter spreads, but they have also contributed to a broader transition from principal trading (mostly done through dealing banks) to agency execution (brokerage function). 

This shift from bank-dominated principal trading to alternative liquidity providers has undoubtedly changed market structure. With electronic platforms and non-bank entities playing a larger role, how do you see regulation evolving to address these changes? Are there gaps that still need to be addressed?

MS: It’s an important concern. Traditionally, banks have been the dominant market makers in fixed income, but post-2008 regulations—particularly the Dodd-Frank Act and the Volcker Rule—restricted their ability to hold inventory. This opened the door for non-bank liquidity providers, such as proprietary trading firms and alternative market makers, to take on a greater role.

However, banks still maintain a stronghold in the primary issuance market, where new bond allocations are often driven by client relationships rather than purely competitive pricing. This relationship-driven approach can lead to inefficiencies, as it doesn’t always reflect real-time market demand. While the secondary market has become increasingly electronic and competitive, the primary market remains opaque, creating potential distortions in price discovery and liquidity distribution.

ES: Let’s delve into the shift from principal trading (dealers) to agency trading (brokers). In your view, what have been the most significant transformations in market-making over the past few decades?

MS: In liquid markets, the most significant shift has been the transition from principal-based market-making to agency execution. Historically, banks operated as principal dealers, actively taking on inventory risk to facilitate client trades. However, competition from alternative liquidity providers along with post-2008 regulatory changes and capital constraints have increasingly pushed them toward an agency model, where the focus is on matching buyers and sellers rather than holding positions on their own balance sheets. That said, non-bank liquidity providers, such as Jane Street and Citadel, now have large balance sheets and play a significant role in liquidity provision, reducing the reliance on banks in some areas.

ES: How has this shift affected market liquidity?

MS: It’s a trade-off. Banks used to provide liquidity by holding inventory, but stricter regulations reduced their ability to do so, shifting liquidity provision to non-bank players—especially high-frequency trading (HFT) firms. HFTs are highly efficient under normal conditions, offering fast execution and lower costs. However, assessing their role in market liquidity is more complex. Some argue that HFTs do not warehouse inventory through volatility and can withdraw liquidity in stressed conditions, while others point out that modern market-makers, such as Citadel, now operate with bank-level balance sheets, mitigating some of these concerns. This is an ongoing debate in market structure research, and empirical evidence is mixed.

ES: Beyond taking on a larger role in liquidity provision, how have non-bank players reshaped market structure?

MS: Non-bank liquidity providers, particularly HFT firms, have transformed market structure through algorithmic trading and quantitative modeling, often supported by ultra-fast execution capabilities. They dominate highly liquid, electronic markets but struggle in less liquid areas like corporate bonds, where banks’ relationship networks and underwriting capabilities remain essential.

Unlike most HFT firms, banks can use their balance sheets to warehouse risk, facilitate primary issuance, and support market liquidity over longer time horizons. Their underwriting capacity allows them to commit capital to new bond offerings and distribution—something HFT firms, which rely on rapid turnover rather than balance sheet deployment, are not equipped to do. That said, in practice, banks often do not commit much capital to underwriting, as deals are generally fully subscribed. Instead, banks dominate the underwriting space due to their relationships with issuers, the lack of true price competition, and the extensive infrastructure required to support underwriting activities, including regulatory and operational considerations.

ES: Given these changes, how have banks adapted?

MS: Banks have adapted to the evolving market structure in several key ways. First, they have invested heavily in electronic trading, leveraging automation and data analytics to compete with high-frequency trading (HFT) firms where possible. By improving execution speed, optimizing pricing models, and enhancing liquidity provision through algorithmic strategies, they have increased efficiency in more liquid markets.

Second, many banks have adopted a hybrid market-making model, blending agency execution with principal trading. While regulations limit the extent to which banks can take on inventory risk, they still selectively use their balance sheets where it is most effective—typically in markets where liquidity is fragmented or where their client relationships provide a competitive edge. Notably, while banks have shifted toward agency execution in equities, they still engage in significant principal trading for fixed-income markets. However, as bid-ask spreads continue to compress, a similar agency shift could occur over time.

Third, banks have expanded algorithmic execution services, offering clients tools to improve trade execution and reduce market impact. By incorporating predictive analytics and smart order routing, they add value beyond simple liquidity provision, helping clients navigate increasingly electronic and competitive trading environments.

Perhaps most critically, banks continue to play a vital role in underwriting and relationship-driven capital deployment. This is due in part to their ability to commit balance sheet to primary issuance when necessary, facilitating bond offerings and structured products.. However, in practice, underwriting typically does not require banks to hold significant inventory risk, as most deals are fully subscribed. The key advantages banks retain in underwriting come from their issuer relationships, the lack of robust price competition, and the technical, regulatory, and operational infrastructure required to support these activities.

Unlike HFT firms, banks absorb and distribute risk over longer horizons, making them important in less liquid markets where capital deployment is necessary to facilitate risk transfer over time. Their ability to support primary issuance, market stability, and long-term capital allocation ensures their continued relevance despite the rise of algorithmic and agency-based trading models. Importantly, the distinction between liquid electronic markets (such as equities and Treasuries) and less liquid markets (such as corporate and municipal bonds) remains crucial when evaluating market-making dynamics.

ES: That brings me to my next question—what are the advantages of integrating into a bank? Many non-bank market makers prefer independence due to regulatory constraints.

Specifically, I want to connect this discussion to an important part of your firm’s history. You and your team joined TD Securities by way of acquisition, with TD acquiring your prior firm, Headlands Tech Global Markets (HTGM), in 2021. This was a critical step in accelerating TD’s capabilities in automated trading, particularly within fixed income. This integration strengthened TD Securities’ electronic bond trading infrastructure and reinforced its commitment to data-driven innovation.

So, why is it useful to integrate a market-making business with a bank? And why did you choose to join TD Securities?

MS: The biggest advantage of integrating with a bank is balance sheet access and the scale of resources that wouldn’t be available otherwise. Proprietary trading firms generally have limited capital compared to banks, which have massive balance sheets that provide liquidity and funding at a scale most standalone firms can’t match.

At Headlands, before being acquired, our capital was relatively limiteds, which constrained our ability to scale. Now, as part of a top-tier bank, we have access to substantial funding—so long as we generate an appropriate return on capital. While banks come with regulatory constraints, the trade-off is well worth it as long as those constraints are managed thoughtfully. This deeper liquidity provision allows for more efficient market-making operations and a greater ability to support clients across various fixed-income markets.

ES: Speaking of funding, how does your firm source liquidity? Do you rely on repo markets like traditional bank market makers?

MS: Before joining TD, we partnered with a bank for financing, where we received leverage based on our capital. Now, as part of a bank, we are self-funded. The bank assigns us a blended funding rate, reflecting its cost of equity, debt issuance, and deposits.

An interesting nuance is that, depending on how a dealer structures its portfolio, its cost of funding may not be as reliant on repo markets as you might expect. If a dealer holds a portfolio comprising both long and short positions, the proceeds from selling the short side of the portfolio can fund the purchases on the long side. This self-funding structure can reduce exposure to short-term funding market rates, making cost of capital less sensitive to repo fluctuations. 

ES: Could you provide an example to illustrate how this works in practice?

MS: Sure. While we still typically use repo markets, our exposure to interest rates is relatively neutral due to the way our positions are structured. For instance, suppose we are long $100 million in bonds and finance them through repo, paying the repo rate. At the same time, we might be short $100 million of different bonds and invest the proceeds in reverse repo, earning the repo rate.

If the repo rates on both sides are equal, the cash flows effectively offset each other, minimizing our sensitivity to fluctuations in the overall market-level repo rate. This balance allows us to operate efficiently in funding markets without significant exposure to interest rate movements.

ES: This is an interesting nuance. Most traditional models of market-making assume firms rely entirely on repo funding, but you’re saying  a firm’s balance sheet structure can significantly reduce that reliance—potentially making market makers’ funding costs almost independent of short-term interest rates?

MS: Exactly. If a firm runs a long-short book, its short positions naturally fund its longs, minimizing reliance on external financing.

However, it’s important to note that in this scenario, the firm must still pay the yield on its short positions. In this sense, compared to a long-only portfolio, a long-short portfolio represents a maturity transformation of the firm’s funding profile. In a long-only portfolio, long duration bonds are funded with short duration borrowing. In a long-short portfolio, the duration profile of the longs and shorts are more closely matched, which insulates the market-maker from risks associated with changes in the shape of the yield curve.

ES: You mentioned the concept of a blended cost of funding in large banks. Can you explain how that works? 

MS: Absolutely. Banks don’t operate with a single funding source; instead, their funding is blended from multiple channels—customer deposits, wholesale funding, repo markets, and internal treasury allocations. This diversification lowers and stabilizes funding costs compared to standalone trading firms, which are more sensitive to fluctuations in risk-free rates or repo costs.

That said, an important nuance is that while banks may have more stable and lower funding costs overall, this does not necessarily dictate how funding costs are allocated to trading desks internally. Each bank likely applies a different methodology, but in general, trading groups are charged a funding rate tied to the risk-free rate plus a spread that reflects the risk and liquidity profile of the assets they hold. This means that even within a bank, trading desks may face different funding costs depending on their specific positions and risk exposures.

ES: Beyond funding sources, how does the structure of a firm’s balance sheet—particularly the mix of long and short positions—affect overall funding needs?

MS: This is key. When market makers hold both long and short positions, they offset each other in financing needs. A short position generates cash, which can be used to fund a long position, creating an internal liquidity source that reduces reliance on external funding.

Combined with the bank’s blended funding structure, this further insulates market makers from dependence on any single funding rate. However, it’s important to distinguish between these two factors. The insulating properties of a balanced long/short portfolio are independent of the advantages of diverse bank funding. A standalone trading firm can achieve a similar balance through offsetting positions, but it won’t have access to the additional funding sources—such as customer deposits or internal treasury allocations—that banks rely on for greater funding stability.

ES: Given that market makers optimize funding through a mix of long and short positions, how do ETFs fit into this strategy? How do they enhance capital efficiency?

MS: ETFs play a crucial role in hedging and liquidity management. Market makers often use them to manage risk efficiently. For example, if we hold a long position in cash bonds, we can hedge by shorting an ETF that tracks the same sector or index. Conversely, if we’re short cash bonds, we can go long on an ETF to offset risk.

Because ETFs trade like stocks and closely reflect their underlying assets, they offer a flexible, cost-effective way to hedge in real time. This ties into the concept of synthetic liquidity—market makers are not just providing liquidity but also dynamically managing their own exposures. By integrating ETFs, options, and futures into market-making operations, we create a more resilient, adaptable liquidity structure while minimizing directional risk.

ES: Clearing fixed income instruments has been a growing focus for regulators. How does it factor into your business strategy? Do you see your firm expanding into clearing services?

MS: Not really. Clearing is mostly a solved problem for us. We settle trades through DTCC, the central clearinghouse for fixed-income markets, eliminating bilateral counterparty risk for bonds and ETFs. Some banks offer clearing services, but that’s generally independent of their trading operations and not our focus. Our competitive edge lies in liquidity provision and market-making technology, not clearing operations.

ES: One final question—since we discussed the hybrid traditional-prop market-making model your team is pioneering, could you clarify the key difference between proprietary trading firms and market-making banks?

MS: In the traditional view, these types of firms differ in a couple of key ways. However, the lines between these types of entities are starting to blur (and our business is one example of this).

First, non-bank trading firms traditionally have smaller capital bases and less-stable funding. As previously discussed, banks have large balance sheets and access to diversified funding sources - deposits, interbank lending, repo, and other secured and unsecured funding channels. By contrast, non-bank firms generally receive funding from banks, and lack the same diversity of liquidity sources, which can make them less stable during periods of volatility.

Second, proprietary trading firms have historically placed a heavier emphasis on technology and quantitative modeling, whereas banks traditionally focused on human-driven trading in less liquid markets, like fixed income.

But these distinctions are starting to become less clear. Some non-bank trading firms now have very large balance sheets, and perform core market-making functions with large portfolios in fixed income markets. At the same time, banks are investing heavily in technology and modeling in a bid to keep up. TD’s acquisition of our firm, Headlands Tech Global Markets, is one such example.

The traditional view is still directionally accurate today, but I suspect the lines will get even blurrier in the coming years.

ES:  This has been an incredibly insightful conversation, Matthew. Before we wrap up, do you have any final thoughts?

MS:  Thank you, Elham. I’d just emphasize that the fixed-income market structure is evolving rapidly. Electronification, alternative liquidity providers, and innovations in trading technology are reshaping how bonds are traded. While fixed income will never mirror equities exactly, we are seeing clear trends that point toward greater efficiency and accessibility. It’s an exciting time to be involved in this space, and I appreciate the opportunity to share these insights with you and your audience.

A (Long) Footnote From Elham Saeidinezhad

Because I’m focused on finishing my forthcoming book, Do Banks Matter?, by July 2026, I may write fewer analytical pieces to accompany the Market Microstructure Interview Series. That said, one of the early chapters will examine how financial engineering, historical developments, and the rise of private capital contributed to the transformation of banks into market makers. A key point from my interview with Matthew (@MatthewSchrager) speaks directly to this broader structural shift—and deserves special attention.

One of my central tasks in Chapter I of "Do Banks Matter?" book is to explain, as clearly as possible and in both financial engineering and simpler economic terms, what differentiates the new market-making model from proprietary trading. This distinction is not just technical—it’s foundational. The model pioneered by firms like TD Securities relies on technologies historically associated with prop trading (automation, alpha modeling, low-latency execution), but applies them in service of liquidity provision, not directional speculation.

Understanding this difference is essential because this is the model that will likely be adopted as we build out secondary markets in private capital, including private credit. Nonetheless, what I want to emphasize is that this new model of market making is neither proprietary trading nor traditional market making as we know it. It doesn’t resemble the prop desks that take directional risk. Still, it also departs significantly from the older Treynor-centric vision, where market makers continuously deploy their principal balance sheets to support prices. That’s why it’s essential to write extensively—and precisely—about what separates this new generation of market-making from both its predecessors and its close cousins.

The distinction is subtle but consequential, and making it legible will be a central goal of the book. This shift marks a structural break. The new model blends agency and principal roles into a single, seamless framework. And this is where terminology matters: I call these firms brokerdealers—deliberately dropping the dash from “broker-dealer”—to signal the collapse of that traditional divide. In this structure, every trade is simultaneously agency and principal. Balance sheet, execution, and client flow are no longer siloed—they’re fused.

TD’s model deserves close attention because it’s being implemented in the fixed-income markets. That’s critical. Fixed income remains the foundation of institutional capital markets and is already shaping how liquidity frameworks are being designed in adjacent spaces like private credit. This is not a side experiment—it’s the blueprint.

The brokerdealer model rests on three pillars. First, execution is driven by automation and predictive analytics. At TD, for instance, 75% of the team is composed of engineers and quants. Execution, pricing, and risk are all managed dynamically through code, enabling profitability even as spreads tighten. Second, ETFs are used not just for trading, but for hedging and arbitrage—allowing firms to manage risk without always transacting in underlying bonds. Third, funding is internally optimized through long-short portfolio symmetry and diversified sources like deposits, repos, and internal treasury. This minimizes reliance on short-term market-based financing.

To many observers, this setup looks like proprietary trading, but it isn’t. The use of sophisticated tools does not make it speculative. The brokerdealer is structured to manage risk continuously, not to take risk directionally. That’s what makes this model both so distinct—and so misunderstood. The brokerdealer model, pioneered by TD, reflects a deeper structural shift—one that integrates technology, balance sheet design, and agency execution into a unified market-making function. It is quietly becoming the blueprint for modern liquidity provision. But is it the future of market making?

I am excited to invite you to the Hedge Fund Association (HFA) - Market Microstructure Symposium 2.0: How Infrastructure Shapes Financial Stability, taking place on Friday, October 24, 2025, at NYU Stern.

Location: NYU Stern
Date: Friday, October 24, 2025
Organized by: Elham Saeidinezhad (NYU Stern/Columbia), Itai Dreifuss (Columbia), Nicholas Smith (HFA/Nasdaq)
Sponsored by: Nasdaq


Building on the success of last year’s symposium, this year’s discussions will focus on the evolving role of financial infrastructure in private markets, ETFs, and risk transfer mechanisms. With the rapid expansion of private credit ETFs, structured equity products, and listed derivatives, market participants face new challenges and opportunities in liquidity provision, capital efficiency, and risk management.

This symposium will bring together leading academics, policymakers, and industry practitioners to explore:

- The development of secondary markets for private credit and liquidity innovations
- The role of structured equity in bridging private credit and private equity
- The rise of ETFs with embedded optionality, including buffer and income strategies as alternative risk transfers
- The increasing use of Nasdaq-100 index options and listed derivatives in portable alpha strategies
- How proprietary trading firms are reshaping liquidity and market-making dynamics

Similar to Market Microstructure Symposium 1.0 in July 2024, this is going to be a small event and with invitation only. I will be sending the invitations soon. Looking forward to engaging in discussions and thought leadership on the future of market microstructure.

Stay tuned for the full agenda. In the meantime, attached is a copy of the program.

Had a fantastic time discussing my latest paper, Banks as Synthetic Hedge Funds, with two of the sharpest minds in financial commentary— Joe Weisenthal
@TheStalwart and Tracy Alloway @tracyalloway on Bloomberg’s Odd Lots.

Link on Apple:

Link on YouTube:

Link on Spotify:

Three key takeaways from the paper:

•Off-balance sheet, regional banks use swaps to replicate hedge fund strategies.

•On-balance sheet, they rely on alternative credit lines—like capital call facilities—to act as synthetic LPs in private equity.

•In a crisis, the Fed could find itself functioning as a synthetic alternative investment fund.

Link to the full paper:
Banks as Synthetic Hedge Funds