Happy Midterms Monday! It’s a big week in America as we decide our political, economic, and moral future. It reminds me of what my Dad always asked me, “John, what do you want to be when you grow up?”

America will tell us on Tuesday.

Thanks to everyone who voted in last week’s poll. This newsletter will stay on its law, finance, and geopolitics trajectory, with a personal essay or two sprinkled in from time-to-time. Because occasionally I need to wax poetic on the state of masculinity, my wilderness years, or the man who raised me.

Now onto my U.S. Midterm Elections preview, which blends law, finance, and geopolitics in the only way I know how.

Subscribe now

Everyone will be watching the electoral map on Tuesday night in America. But more of us should be focused on which levers change hands and when. How does power transfer if Dems win the House, or if they win both the House and Senate? What could that lead Trump to do from the Executive Branch?

As I said in my last midterms piece:

There’s a consistent trend with Trump and his administration — the more they disregard or defy the rules of the game, the more threatened they feel. The defiance of institutions and rules that otherwise ordinary people must follow is not a signal of strength but fear; fear they will lose power.

I called these actions “costly moves.” The price of these moves is only worth paying if the incumbent (Trump) expects the vote to be close or lost. A great example of this is Trump instructing DOJ to preemptively challenge the Alaska Senate election, but only if the GOP loses.

You don’t make those moves — challenging election results before you have them — if you think you’re going to win.

And he’s made many of these moves, as I explained in my last piece. Democrats are also looking stronger in the polls, with the House well in reach and the Senate a real possibility.

The size of the wave, however, matters far less than which chamber flips.

Tier 1: when one chamber is enough

If Democrats flip the House on Tuesday, one of their biggest power grabs will be appropriations. This is Congress’s power of the purse. When most people hear this, they think about new spending, but few realize the Congressional purse has a leak.

If a president asks Congress to cancel funds less than 45 days before the date it’s set to expire (typically September 30th), the money is frozen until it dies, with no vote required (similar to a pocket veto — it’s called a “pocket rescission”). The Trump administration has already done this twice, and the Supreme Court let it stand without ruling whether it is legal.

An incoming Congress can close this loophole, but only by writing it into a spending bill that the president has to sign (or otherwise risk a shut down). A standalone bill would likely get vetoed by the current president. Although Republican appropriators are split on whether pocket rescissions should be legal (it’s likely one of those things where some are opposed if their party is not in power, but in favor if their party has power).

In addition to the purse, if Democrats win the House they get subpoena power. They can compel testimony and documents, which they might have a field day doing across Trump’s crypto adventures, suspicious trading activity, and his family’s activities, from his sons to Jared Kushner. The big questions will be how hard should Democrats push and will it even be effective?

Remember, criminal contempt against anyone who defies a House subpoena requires a full House vote and then a referral to DOJ (controlled by Trump for the next two years). So it’s unlikely that any Trump ally or family member would ever get prosecuted. They would likely just ignore the subpoenas while MAGA screams “lawfare.”

Democrats might be wiser to try the court of public opinion than anything formal like subpoenas that go through the Executive Branch. It’s a big flaw of our current system as no Executive Branch — regardless of party — is eager to enforce laws against itself.

That brings me to impeachment. Yes, this takes a simple majority. And you probably already knew that given the current president was impeached twice.

The challenge? Conviction and removal requires 67 senators. So it’s not a feasible tool for checking executive power or punishing abuses because the likelihood of getting some 17 Republicans to agree is almost zero.

Over at the Senate, their primary one-chamber power is confirmations. This applies to everyone from judges and agency heads to Fed governors. I bring up the latter because Trump is still trying to fire Fed governor Lisa Cook and his team has been hunting people he doesn’t like for evidence of “mortgage fraud.”

Future Fed board seats run through the Senate, so if the Democrats gain control, this will be a huge barrier for Trump to hurdle (he would need their votes to fill new seats). It would not only protect Cook, but also the institution’s independence from the Executive Branch.

Tier 2: both chambers required, plus the president

If the Democrats achieve the once unthinkable and win both chambers — the House and Senate — they will have additional levers of power available to them, but many also require the president’s signature. The reconciliation process, for example, is a filibuster-proof budget process that led to the One Big Beautiful Bill. It required both chambers to adopt the same budget resolution.

If you lose either chamber, that budget process is gone. You also need the president’s signature.

By far the biggest Tier 2 lever though is one almost nobody discusses — the debt limit. Projections estimate that the U.S. will hit the ceiling around May 2027, which last year’s law raised to $41.1 trillion. Whoever controls Congress in January 2027 will inherit this fight. And if either chamber flips, raising the debt limit will require votes from the other party, which is the most leverage any midterm result can buy.

Tier 3: what no midterm election reaches

This is also under-discussed — what the midterm elections cannot touch.

Remember how the Supreme Court struck down emergency tariffs in February? The administration used serious legal creativity to rebuild those tariffs on trade statutes that delegate authority to the president. No vote from Congress required.

Today’s tariffs of 10 to 12.5 percent on some 60 economies all rest on one of those statutes. To undo them, there would either need to be a legal challenge (there’s one pending) or a new law, which means a presidential veto, which means two-thirds of both chambers needed to override it. Which, of course, is not going to happen.

War powers have the same math. I’ve previously explained why Congress cannot actually check the President (under our current laws) when he engages in hostilities. So it didn’t matter that both chambers voted over the summer to end hostilities in Iran — the White House called the vote meaningless. And it was technically right, legally speaking.

The type of resolution Congress passed never actually goes to the President’s desk and legislative vetoes haven’t been a thing since the Supreme Court banned them in 1983. This left the President largely unchecked on engaging in foreign wars.

So no matter who controls which chamber, nobody is likely forcing this president to end the war in Iran (unless maybe the bond market finally wins).

Presidents follow a pattern when facing a final midterm

Every recent two-term president that faces a final midterm has responded in a similar way. After Democrats won both chambers in 2006, George W. Bush announced the Iraq surge anyway. Congress passed war funding with a withdrawal timeline, but Bush vetoed it and the surge proceeded.

In 2014, after Republicans took the Senate, Barack Obama announced deportation protections to shield millions of undocumented parents from deportation. Congress tried to defund the program through a must-pass spending bill and failed. A federal judge stopped it instead and the Supreme Court deadlocked.

Even if you go back to 1999, Bill Clinton pulled similar moves in Kosovo without Congressional authorization, launching a NATO air campaign.

The point is that a term-limited president facing a Congress he cannot control almost always shifts into Tier 3. He goes to where Article II constitutional powers are the strongest and the check is the courts, not some Congressional vote. So I expect a lot more foreign policy escapades, tariff threats, and executive orders as Trump moves into his final two years alongside a Congress he likely won’t have complete (or maybe any?) control over.

But I wouldn’t count out the Republican majority yet. The new Congress won’t be seated until early January 2027. Until then, the current majority holds every lever of power, including a December 11 funding deadline. And if a lame duck House could impeach Bill Clinton in December 1998, it could do bold things before December 2026 ends.

Tuesday will set America’s future course, but regardless of what happens, the president is still very much steering the ship.

Subscribe now

Share

Happy Friday. Today I’d love to hear from you. Short poll below on the intellectual and artistic direction of this newsletter moving forward.

But before we get there I have a lot of thoughts about AI and writing.

My brain currently lives at the nexus of these topics for two reasons — the angst I feel in the world and a recent Google event I attended this past week in NYC where publishers were very down on the health of their businesses. Organic search just ain’t what it used to be.

When I think about AI and what I’m doing right now (writing) though, I’m actually not thinking about macroeconomics. I’m not thinking about how AI could take our jobs and displace millions of workers.

It could easily do all of that, although I imagine I’m more optimistic than some. Just look back at every agricultural, industrial, or technological revolution and you’ll find a net positive creation of jobs following every wave of innovation.

Did some stagecoach drivers lose their jobs? Undoubtedly. But now we have industrialized farming and a billionaire who wants to merge his electric car and rocket companies. Lucky us.

What I’m most interested and concerned about at the nexus of writing and AI is the reality that with AI now it’s easy not to write. For those who never liked writing in the first place that is.

For the more insane among us who live to write - who do not feel fulfilled until they’ve opened a vein and bled on the page - this AI era, I think, creates opportunity.

LLMs revert to the mean. The average output of human expression. They’re not novel, never weird, and rarely profound. Solid B+ writing.

But that’s not what captures the mind. What fills the soul. What makes you remember something for days or weeks or months later.

It’s like most short form video. Do you ever remember what you watched? Unless it’s substantive stuff I put out (of course!).

Do you ever remember what your favorite LLM told you?

My reason for writing this note though is not actually to lambast LLMs. They’re fine. Good editors and researchers for us writers.

My purpose here is to warn those who think in this new AI world that they don’t need writing skills any longer. This will be a growing perspective if it’s not one already. Tools that make anything easier tend to morph into crutches.

People with this perspective miss a fundamental truth about writing - it’s arguably more about thinking than the craft of writing itself. There’s simply no better way to organize your thoughts as I’m trying to do now. So it logically follows that those who delegate writing to LLMs will also be delegating much of their thinking.

And the syllogism is complete.

Subscribe now

Now for the poll on the future intellectual and art direction of this newsletter:

Thank you for being here. Have a good weekend.

Imagine that while you’re sleeping, your trusted AI agent is working for you. But there’s a problem. It fails.

Instead of following the script you specified, it takes irreversible actions like deleting a client database. Then it executes unauthorized financial transactions. Maybe that was other people’s money you were holding.

This is what I mean in the title when I say too agentic. An AI agent — a program that doesn’t just answer questions but actually acts, supposedly on your behalf — can cause serious trouble for yourself and others.

Don’t take my word for it. Read Anthropic’s S-1. It spends almost 100 pages going through numerous risk factors from its products (or services) causing the end of humanity to running wild with broad systems access that can potentially go unsupervised for days.

Yet through all of these risks Anthropic is reportedly targeting a $1.5 - $2 trillion valuation. It’s expected to go public in late November, just in time for Thanksgiving in America.

You can probably imagine why these risk factors are not making many feel thankful.

So much of the coverage I’ve read about Anthropic’s S-1 was fixated on its first half of 2026 loss of $80.7 billion. Most of that, however, was tied to a $72.4 billion paper charge from revaluing convertible notes and similar instruments. I think the more honest framing is roughly breakeven once you also account for stock compensation to employees.

But for me, the number that matters isn’t in the financials. It’s in the 94 pages of risk factors, where Anthropic admits that efforts to limit its liability may not be enforceable or sufficient, especially when it relates to its agents’ actions. How the law applies to AI and agent behavior specifically is unsettled to say the least. And as I keep arguing into the void, our current laws will not protect us sufficiently from AI.

Given the importance of Anthropic’s legal liability — not only for the firm’s future, but humanity — let’s dig into it. There are four candidates who could pay the price here if something goes wrong. Here’s how each one can get out of paying and who could be left holding the bag.

Read more

One of my main takeaways from studying political science as an undergrad came from an entry-level course taught by a grad student. This guy pounded the following phrase into our brain. Repeatedly.

“The rules of the game are important.”

“Why are they important?”, you might ask.

Because they’re bigger than any one man or woman. They check the powerful as much as the powerless. They check the haves as much as the have-nots. They set expectations, create order, and when executed well, lead to a functioning society.

At least that’s how the rule of law should work in theory.

This grad student warned us (back in 2007) that the second the rules start to become meaningless — or applied differently to different people — that’s the second we start to lose democracy. It fractures. It frays at the edges. And in a similar way to how Hemingway described going bankrupt, democratic decay happens gradually, then suddenly.

We’ve witnessed a lot of this in America over the past decade. Now ahead of the U.S. midterm elections in a few days, we’re seeing an acceleration of an administration bending or outright breaking the rules so they can win the game at almost all costs.

The Trump administration is deeply unpopular:

While many point to Trump’s unpopularity as a strong indicator that Democrats will ride a blue wave over the midterm elections, I’m analyzing what his administration’s actions signal. What his continued disdain for the rules of the game indicate. What his expectations are for the midterm elections.

Subscribe now

The signal test I use

All politicians pull moves ahead of major elections. They’ll kick the can of difficult decisions down the road beyond election day. They might take cheap shots at their opponent or use unflattering imagery like the infamous Dukakis in a tank photo.

We expect this. It’s the dirty side of politicking in America, all powered by the firehose of unlimited money thanks to Citizens United.

But this category of what I’ll call “routine incumbent moves” should be distinguished from the “costly moves.” The latter includes moves that carry a real price, whether to the rule of law, a likely court loss, bipartisan backlash, reneging on a campaign promise, policy reversal, damage with allies, etc. When I say “costly” in this context I mean: the price is only worth paying if the incumbent expects the vote to be close or lost.

Basically, costly moves are those a sitting president wouldn’t pay if he was confident about the midterm election results.

So let’s analyze some of Trump’s recent moves leading up to the midterms to discern whether they’re costly moves or simply routine incumbent moves. The more costly, the more likely Trump thinks his party is at risk in the midterms and vice-versa.

The costly moves

Suing Alaska in advance to halt certification. There are two people named Sullivan running for Senate in Alaska and only one is the Republican Trump wants to win. That Sullivan is neck-and-neck with the leading Democrat, Peltola (who has a narrow lead in recent polls). The DOJ, at Trump’s request, filed a lawsuit on October 22 arguing that a second candidate with the same name (Sullivan) could siphon votes from the Republican incumbent. The lawsuit requested that the court block certification until an investigation into voter confusion is done. But they only want this done if their Republican Sullivan loses. If the Trump administration were confident in its Alaska chances, it wouldn’t be buying insurance by using the DOJ as its personal law firm to halt election certification that only applies if the Republican loses.

Reversing Russia policy for diesel prices. I’ve covered this one extensively, but it cannot be emphasized enough that Trump invoked a waiver for a law that he had just signed weeks earlier. It caused friction with Ukraine and Europe and contradicted tariff threats he’s made against India. This is at best a short-term fix with barely any structural impact on diesel supply, all for a fairly high foreign policy price by waiving sanctions on Russia when Putin didn’t concede or do anything for it.

Making $5,000 payments conditional on the election result. Trump recently pledged checks to every American if Republicans keep control of both the House and Senate. This would cost taxpayers some $1.2 trillion if Congress enacted it. In the process, Trump has seemed to ignore federal vote-buying laws, and many vulnerable Republicans like New York’s Mike Lawler have refused to commit to it. Offering what appears as an outright transaction to get votes is not what a confident sitting president does going into midterm elections.

Taxpayer-funded promotional ads followed by a retreat. Trump dipped into federal taxpayer coffers to fund promotional ads that cost at least $1.5 million. A federal judge temporarily barred them before Trump shifted them to MAGA Inc., which has led to separate complaints. Dipping into federal taxpayer funds for advertising is not the move of a secure incumbent.

Federal intervention in ballot counting while voting already started. On October 15, the DOJ sued California to block ballots from more than 20,000 suspected noncitizens. This comes months after the Trump administration’s mail-voting executive order failed (note: Trump has voted himself by mail). The mail-voting order lost in court, and appellate judges sounded skeptical of the California lawsuit. While this may be consistent with Trump’s long-held beliefs on fraud and elections, the timing is the signal: the DOJ suing mid-vote, after early voting already started.

Mid-decade redistricting. Although not a recent move from 2026, last year Trump called on Republican-controlled states across the country to draw new maps to protect the GOP’s narrow House majority. Prior to 2025, only two states had conducted voluntary mid-decade redistricting since 1970. It triggered reciprocal redistricting from Democratic-led states like California and was Trump’s original costly move signal. Who redraws election maps for a majority they think is safe?

Routine incumbent moves

“Pretend I’m on the ballot.” Trump is obviously not on the ballot this cycle, but he’s urged voters to pretend like he is. Almost every sitting president nationalizes midterm elections. Obama, for example, did it in 2014.

Releasing oil from the Strategic Petroleum Reserve. This is similar to the move Biden made in 2022 to rein in fuel prices ahead of the midterms that year. Every president fights gas prices before election season.

Extending the China trade truce beyond Election Day. Kicking the can of difficult decisions down the road is a time-honored tradition. No president wants a market shock before a major vote.

Promising voters money. I want to distinguish something from the $5,000 costly move above. There’s a general practice of promising tax cuts or other promises of relief. What’s not common and arguably illegal is paying Americans cash on the condition that the president’s party wins control of both the House and Senate. That’s not a policy promise, but a transaction.

What the signals tell us about Trump and the midterms

I’m not one to gaze into the crystal ball and proclaim I know what will happen. If I’ve learned anything over the past decade it’s to take polling data with a giant grain of salt. Question everyone, even the most respected experts.

But whenever I sit down to analyze Trump as objectively as one can, I keep coming back to that grad student whose mantra I never forgot.

“The rules of the game are important.”

There’s a consistent trend with Trump and his administration — the more they disregard or defy the rules of the game, the more threatened they feel. The defiance of institutions and rules that otherwise ordinary people must follow is not a signal of strength but fear; fear they will lose power.

So when I see Trump ordering the DOJ to sue Alaska and halt their certification, but only if the Republican loses, when I see him undermining the Lindsey Graham Act he had just signed to side with Putin (who has no respect for the international rules of the game), when I see him offering to pay voters on the condition of a Republican win, when I see taxpayer-funded promotional ads and federal intervention in state ballot counting and mid-decade redistricting efforts, I see someone who is not secure in his power.

I see an administration making costly moves, with payoffs that only exist if the race is close. Multiple of these examples are conditional on the election results. You only make those conditions if you think the outcome is in serious doubt.

The unfortunate reality is that it’s not just the Trump administration that will pay the price for its costly moves. It’s all of us. It’s the entire democratic world. It’s anyone who wants to live in a rules-based world order.

When the rules of the game lose their importance, the game stops. Only for a new game to start, this time with more Hobbesian rules. Where might makes right. Where the law of the jungle prevails over the rule of law.

That’s why the rules of the game are so important. It’s why we must never stop working to protect them and make them better.

Subscribe now

Share

It’s time for an odd kabuki dance.

Those aren’t my words, by the way. They’re from the man himself, Elon Musk, CEO of both Tesla and SpaceX. He was describing how he can’t be involved when the recently formed special committees explore a merger between his two publicly-traded tech giants.

I had to look up what a kabuki dance was.

It’s apparently a classical Japanese dance-drama, but let’s be real here. In America, kabuki really means a show that’s put on when everyone already knows the ending.

Can you imagine the Tesla and SpaceX special committees doing this dance together?

Well, you’re in luck. Because I did. Each company name refers to their respective special committee.

Subscribe now

Tesla: “Thanks for meeting with us guys. So you’ve had a taste of public markets for not even 6 months and you want to merge with us?”

SpaceX: “We were actually thinking of buying you. I mean, we’re really the future of SI. You know, Super Intelligence. We’re also bigger than you. Rockets, data centers in space, satellites. We almost ended the war in Ukraine, what have you done lately?”

Tesla: “We were the original SpaceX. We’ve promised things for decades that still haven’t come true.”

SpaceX: “Ha, very funny. Can’t wait to see your face when we colonize Mars in 2126.”

Tesla: “You’ve clearly never seen or read The Martian. But in all seriousness, we have so much data, we’re way less volatile than you, and we have better consumer products. We’re equals at worst.”

SpaceX: “I don’t think so. Just this past year we’ve had our market cap hit $2.6 trillion. You’ve only reached $1.6 trillion.”

Tesla: “You forgot to mention your market cap also dropped to $1.4 trillion during that period. Are you even a real company?”

SpaceX: “What do you mean?”

Tesla: “Like, could you even exist if it weren’t for government contracts and bond offerings? Do you have any free cash flow?”

SpaceX: “We’re colonizing Mars and our TAM is infinite!”

Tesla: “Not if Democrats win the House in a few days. Then the only thing infinite will be your investigations, subpoenas, and regulatory inquiries.”

SpaceX: “Look, we need to cut the BS. Elon told us that he wants full control of any future entity. I’m sure he’s told you the same.”

Tesla: “Yes. I mean, we do pray at the same master’s altar.”

SpaceX: “OK then it makes the most sense for us to acquire you. Elon has 84% of the voting power in our dual class stock. He also has extra rights to appoint board members and our governance basically allows him to do whatever he wants.”

Tesla: “We love letting Elon do whatever he wants. But you’re going to have to pay our shareholders a premium so Elon can have even more control over them. They won’t approve otherwise.”

SpaceX: “Why do you think that? They approved that insane pay package for Elon after all. And can you imagine what Elon could do with this merged entity? Electric, self-driving cars. Robotaxis. Cars with rockets on them!?”

Tesla: “OK, you got us with the flying cars. But we’re still demanding a 20% premium on our current stock price so everyone’s compensated for Elon’s greater control.”

SpaceX: “You act like Elon having more voting power is a bad thing. Since when? He’s amassed more individual wealth than practically anyone in history. He’s made all of us rich over the long run. Why not give him complete control? Why does that cost a premium? If anything, you should be thanking us!”

Tesla: “You know what? You’re right. Let’s just give him 100% voting power over the new entity. Carte blanche governance-wise. Let’s build some craziness and promise even more of it.”

SpaceX: “Thank God we have that settled. Now we just need to decide on his pay package.”

Tesla: “$2 trillion.”

SpaceX: “Huh?”

Tesla: “Yeah, the combined contingent pay packages at both Tesla and SpaceX. Why not just add them together. We’re combined now aren’t we?”

SpaceX: “You’re right. And I mean, one of our contingencies is colonizing Mars with 1 million people. So yeah, that’s worth $2 trillion at least, right?”

Tesla: “The promise itself is worth $2 trillion.”

SpaceX: “We’re going to be great together.”

Note: Even if the Tesla committee wins a 20% premium on behalf of shareholders, it would likely be cheap for Elon. Tesla shareholders would receive SpaceX’s low-vote shares, while Elon keeps his super voting shares — so every dollar of premium is paid to Tesla shareholders in stock that carries economics but far less voting power.

The committees will negotiate hard over this premium, but the man who isn’t in the room will be the biggest beneficiary of their kabuki dance.

Subscribe now

Share

Credit markets get jumpy when traders smell blood in the water. That’s exactly what’s been happening the past week with CoreWeave, a company that’s known as a “neocloud”, which provides GPU-accelerated infrastructure for AI training and workloads.

All it took was a chart from Chamath Palihapitiya on October 13th and a response from Elon Musk — “yikes” — to send CoreWeave’s credit into a tailspin. Its 2032 bonds touched a 15% yield that day.

A week later, its 2031 bonds now trade near 80 cents on the dollar with a 15.58% yield, which is up from 11% in August. CoreWeave’s stock is also at its lowest level since May 2025.

Why does any of this matter to the broader AI trade, which has sustained much of the world’s growth and economic activity in 2026?

Because what traders are pricing here is who gets paid last in AI. CoreWeave is the glaring example.

A landlord that rents out AI chips

Earlier this week I wrote about Starlink and the risks of critical infrastructure running on a private company’s terms. With CoreWeave, the story is somewhat similar. Except it’s more a story of critical infrastructure running on a private company’s debt. It tells us what happens when the people who lent CoreWeave money start to lose faith.

The day after Chamath posted CoreWeave’s rising yields, its stock fell about 15% on October 14th. Its credit default swaps started to rise, basically traders betting that the risk of a CoreWeave default was getting higher.

But then this week we appeared to be in the clear. Goldman upgraded the stock to buy, and the rest of the AI trade rallied. The only problem was that CoreWeave closed slightly lower, with its bond yields still elevated near 15%. And now its stock is at its lowest levels since May of last year, begging the question why it’s performing so differently from the rest of the AI trade?

Many people describe CoreWeave as the landlord of artificial intelligence. The company rents out chips to Microsoft, Meta, OpenAI, and the like. So its customer base is heavily concentrated.

It also bought most of those chips that it rents out with borrowed money.

Business has been booming though. Its second quarter revenues from this year doubled to $2.6 billion, with $104 billion in contracted future revenue. Indicators that this stock should be rocketing alongside other AI darlings.

The problem is the price of CoreWeave’s future dollars. CoreWeave is the type of credit stress I’ve been talking about in a world of rising yields and interest rates. They’ve been borrowing at 9-10% in a hot market. Now the next dollar costs CoreWeave ~15%, with lenders demanding more from a company they’re less sure can pay them back.

If CoreWeave wants to keep growing, they’ll need to continue borrowing.

Amidst this world of rising yields and rates, maybe we shouldn’t be asking whether there’s an AI bubble, but where does a lender stand in the AI line?

Read more

Look, I’m a Starlink customer. I love it, especially in rural areas without fiber networks. For many folks in those communities, Starlink may be their best internet option.

You know who also relies on Starlink? Ukraine.

Last Tuesday, October 13, Ukrainian officials said drone missions failed mid-flight. By the end of the week, two different people had two different answers about who was responsible.

Nobody disputes that Ukraine’s Starlink connection was throttled. Fingers are just getting pointed around the table, from Elon Musk to President Donald Trump.

Shouldn’t we be concerned when a weapon of war can be switched off or slowed down (throttled) and we cannot identify who made the call? Especially if that someone was unelected like Elon Musk or is someone nobody can name.

At a minimum, it’s a major national security vulnerability.

President Trump denied throttling Ukraine’s access; the same throttling that opened a window where Russia unleashed some of its most aggressive attacks on Ukraine to date. And this all came just days after President Trump announced a diesel deal with Russian President Vladimir Putin, weeks ahead of the U.S. Midterm elections with U.S. diesel prices near record highs.

Subscribe now

The Starlink throttling, in four acts

• Act I: Ukraine was hammering Russian refineries and the Trump administration demanded it stop. A few days later, the throttle on Starlink hit, basically a speed limit for their drones. Inside Ukraine’s borders, drones previously flying at 400 km/h lost connection at 161 km/h. Ukrainian officials literally drove cars across the border to test the speed changes.

• Act II: The finger-pointing. Elon Musk argued that Ukraine cannot use Starlink for “offensive” purposes, but pointed the finger at Trump and the government, who claimed they didn’t do it. A senior Ukrainian official said the White House communicated it was their decision.

• Act III: The Trump administration waived sanctions on Russian diesel. And now there are even reports that Putin wants a summit with Trump and China’s Xi Jinping to decide Ukraine’s fate (without Ukraine at the table, which is eerily similar to the 1938 Munich Agreement that decided Czechoslovakia’s fate without them at the table either — we all know what happened to Europe after).

• Act IV: Ukraine is still being throttled as of Saturday, October 17. Zelensky even clarified their use of Starlink and how they use it to defend Ukraine, not for action inside Russia.

Remember Elon Musk from 2022?

The minute I heard this Starlink throttling news I thought back to 2022. It was the early days of the war and Ukraine was trying to bring the fight to Russia’s Black Sea fleet (which was badly damaged after this episode).

When Elon Musk caught wind of what the Ukrainians were up to, he refused to switch on Starlink for them. Elon said he would have switched on Ukraine’s Starlink access if the Biden Administration had asked.

Apparently this week, the Trump administration asked.

Just instead of switching on Ukraine’s access, they reportedly asked to switch it off. Although they, of course, deny it.

We need more transparency on critical military technology

There is no public process for how someone decides to use critical military technology like Starlink. There’s no notice requirement, nobody is required to own the decision publicly, and Congress isn’t required to be involved.

Nobody has even publicly documented Starshield’s contract terms. That’s the military version of Starlink.

So the American people signed themselves and their allies up to a billionaire’s internet service that’s designed to connect them and keep them safe, but it’s a complete mystery on who has power over the throttle that determines the speed of that connection.

How much does that billionaire, Elon Musk, have a say in it?

None of this is fully known and that’s precisely the problem.

Yet we don’t have this problem in other areas of military contracting, specifically airplanes. Airlines contractually pledge aircraft can be activated when needed, with the commander of U.S. Transportation Command having activation authority in case they need civilian air support in emergencies. It was last used in the botched evacuation of Afghanistan in 2021.

There’s a big difference though between the terms of use and who has activation authority with civil aircraft and how this works for Starlink and Starshield. First of all, they’re written down and specific people are accountable.

When it comes to satellite systems though, the Space Force started writing rules, but never finalized the crucial “denial of service” clause. This leaves the door open for billionaires, presidents, and other government officials to exercise personal grievances or preferences in unaccountable ways.

Hence all the finger pointing about Starlink and Starshield in Ukraine.

We need to write the rules down

Satellite internet service is the future. Data centers in space are probably the future too. Both have immense national security implications. And both are currently being pushed and/or operated by SpaceX, a publicly traded company where Elon Musk holds 82-85% voting power.

The European Union is already seeking to rid itself of dependence on SpaceX for critical infrastructure like satellites and internet. The Trump administration is threatening them with “devastating” retaliation if they restrict SpaceX in any way.

But the Europeans aren’t wrong here. The whims of Elon Musk and President Trump are simply too unpredictable. And when there’s little to no governance in place when determining whether to throttle Starlink access, there’s huge risk for anyone relying on it.

Including the U.S. military.

What if Elon and Trump find themselves in a spat again, only for Elon to throttle Starlink or Starshield to make Trump look bad, putting U.S. servicemen in danger? I’m not saying this would happen, but it’s an unacceptable possibility of putting so much power over U.S. national security in the hands of basically one private businessman.

If there’s any sort of throttling or off switch on critical military technology, there also must be a process for who, when, why, and how it should work. That process must be transparent to the public.

Otherwise, we’re at the whims of a billionaire and a president. And it doesn’t matter what their names are.

Subscribe now

Share

If you’re interested in the AI boom’s off switch, stay tuned for Wednesday’s planned piece which will show how it may be buried in otherwise boring loan documents.

Thanks for being here. If you want to support this publication, consider becoming a paid subscriber if you’re not already.

Good afternoon. Yesterday the National Assembly in France started debating a budget most people will consider a snoozefest.

Bond traders won’t.

We previously discussed who will save France, including the European Central Bank (ECB) backstop and how the market is underpricing the risk of no stable French government and no ECB safety net. Simultaneously.

How this budget debate plays out in France matters more to bond traders in London, New York, and Tokyo than anything going on at the ECB.

France is about to show the world what happens when a central bank’s rescue comes with legal conditions, and why America’s version of this problem looks completely different.

Subscribe now

Why France is in trouble

It’s important to understand why France is in trouble in the first place. We discussed soaring French debt last week, which has put pressure on its bond market.

France borrows by selling government bonds, which traders call OATs. Germany’s bonds — called Bunds — are the benchmark “safe” asset in Europe given the fact markets view Germany as the continent’s strongest economy.

The spread is the extra interest France pays over Germany, which is measured in basis points (100 basis points = 1 percentage point). You should think of this as the market’s price for the risk of lending to France instead of Germany.

Now look at how that spread has jumped recently:

Back in February of this year, the spread was about 55 basis points, but now it’s closing in around 140. This is the highest it’s been since the 2011-12 eurozone crisis, and this sentence is crazy to say — it’s a wider spread than Italy’s.

A G7 country now pays more to borrow than the old problem children of Europe, Italy and Greece.

Why the world should care about France

As I said last week, France is not Greece 2.0. It’s the euro area’s second-largest economy, and it’s deep and diversified. The challenge is that when a market of France’s size endures a repricing, global markets feel it. The world feels it.

This tension has started spilling into other high-debt eurozone countries and is putting pressure on the euro more broadly, in addition to equity markets. This pressure weighs on the euro while U.S. yields support the dollar, making this French budget fight an American exchange rate story too.

The budget fight faces a divided parliament ahead of next year’s election to replace Macron (who has reached his term limit). The far right National Rally is looking at its best chance yet.

So what happens in this French budget fight will have major financial and political implications for Europe and the world.

What’s interesting is that short-term French debt has remained relatively stable, with spreads around 65 basis points. But when you look at the 10-year yield, that spread widens to about 140 basis points. The market is charging for the years after this next French election, which is a bet on uncertain politics, not fear of French insolvency.

Everyone assumes the ECB will step in and save France from its mess. This is the central bank that sets one interest rate for every country using the euro. And while it has tools to counter “unwarranted” and “disorderly” market dynamics, its rescue tools have two big pieces of fine print — bailout conditions — that almost nobody covering this has read.

I have.

Read more

Good morning. Deadly strikes over the weekend by the Houthis targeted King Khalid International Airport in Riyadh, killing at least 12 people and injuring hundreds. It was the deadliest attack in a Gulf Arab state since the U.S.-Israeli war with Iran began in February of this year.

The strikes coincided with the arrival of energy executives and government officials in Riyadh for the World Petroleum Council Energy Congress, with BP among those that pulled out.

The thing is that oil prices barely moved. They’re slightly up as I write this, with Brent crude trading around $106/bbl, but they haven’t moved much since Friday despite the attacks on the Saudi airport.

Markets only care about barrels though, not airports. And if you want to watch barrels, you have to keep an eye on Yanbu, the end of the East-West pipeline that’s supposed to allow Saudi Arabia to bypass the Strait of Hormuz via the Red Sea.

That’s how it’s supposed to work. On paper and in theory. But as we’ve learned in Ukraine and throughout this war against Iran, theory does not always work or accurately account for asymmetrical power imbalances.

This Saudi pipeline bypass of Hormuz has two main constraints. The first is at the port in Yanbu. The pipeline that feeds the port is rated at 7 million barrels per day, but about 2 million of those barrels per day go to west coast refineries. The port at Yanbu can load about 4 to 4.5 million barrels per day onto tankers, so it’s not like the entire pipeline capacity can be exported daily.

The second constraint is that the pipeline route is terribly exposed and has been hit repeatedly. The Houthis announced their blockade of Saudi Red Sea oil in July 2026. Then they struck the refineries at Yanbu and Jizan, their first direct hits of Saudi oil infrastructure since 2022.

August pipeline flows fell to 2 million barrels per day, their lowest monthly level since January (before the war). The Houthis didn’t need to match the Saudis strength for strength — just like the Ukrainians against Russia and the Iranians against the U.S. military in Hormuz — they just needed to make the Red Sea difficult and dangerous for the Saudis to use.

Brent crude has steadily traded higher since August as pressure built on both chokepoints, moving from ~$80/bbl to ~$106/bbl today. The Saudis have tried to find a workaround for this disruption — using eastern Gulf terminals and ship-to-ship transfers near Oman for export — but this only partially offsets the disruption to their Red Sea bypass of the original disruption (the Strait of Hormuz).

The Saudi energy minister says that pipeline flows have returned to some normalcy, pumping ~5.8 million barrels per day. But just as flows recovered, fighting on the Yemeni coast escalated and the Houthis turned to deep strikes on the Riyadh airport.

Most Saudi crude goes to Asia, and a tanker leaving Yanbu for Asia has to exit the Red Sea through Bab el-Mandeb. If the Saudis can help the Yemeni government fully retake the coast around it from the Houthis they can protect a crucial export route.

Note the timing here. Deep strikes by the Houthis intensified against some of Saudi Arabia’s most precious assets — oil facilities and the Riyadh airport — just as the Houthis lost coastal ground near Bab el-Mandeb.

Saudi Arabia’s Hormuz bypass insurance only pays out if this second strait stays open via the Red Sea. This is why the fight for the Yemeni coastline matters more to oil markets than anything happening at the airport in Riyadh. And like I’ve argued about actuaries in the Strait of Hormuz before, commercial and legal factors can have bigger impacts than the war itself on whether a strait stays truly open.

Share

Subscribe now

Good afternoon from New York City, where I’m looking towards Paris. And while I normally dream of bread when I do that, today I’m thinking about bonds and debt.

I’m wondering who will save France from a crisis, from within or without. There’s a hierarchy to this great repricing that’s afflicting much of the developed world, and France occupies the messy middle that doesn’t own its currency and shares its central bank.

So who will save her?

Subscribe now

France borrows at a higher rate than Italy. Think about that. If anyone told you this would happen over a decade after Italy nearly broke the euro, you probably would have laughed in their face.

If I had also told you that French bonds have a spread over Germany that’s at its widest since that eurozone crisis, you likely would have asked what alternative reality I’m living in.

But this is very much France’s current reality.

France is under intense financial and political pressure because it hasn’t balanced a budget since 1973 (5 decades!). And as you’ll see, this is made worse by where France sits in the sovereign bond hierarchy.

At the top of this hierarchy are countries that borrow in a currency their own central bank controls (e.g., U.S., Japan, Britain). At the bottom are countries that borrow in someone else’s currency, like emerging markets with dollar debt. The eurozone, including France, sits in the messy middle, sharing a central bank with other countries while borrowing in a currency they don’t fully own or control.

Just look at the numbers.

France has missed its last two budget deadlines and carries a deficit around 5% of GDP. Overall, its public debt is near 120% relative to GDP, with interest costs climbing from roughly €65b to €74b in 2027 (which is the government’s own projection).

On top of this financial pressure, France has a presidential election in April 2027 with no incumbent (Macron has reached term limits).

But for all of those dire numbers, we shouldn’t jump to the conclusion that France is Greece 2.0. For one, it has legal fallbacks that can keep the state funded, unlike us Americans who just shut down the government when budget bills fail.

Its sovereign bond auctions are still getting well-covered. Investors are still buying French debt. And its average debt maturity is long, so higher rates now will take time to feed into the budget.

For all the comparisons with Germany, many neglect to mention that German Bunds have outperformed. This isn’t just a story of French bonds selling off.

France has a deep, diversified, and large economy, especially when compared with the eurozone’s Greek and Italian problem children from the 2010s.

While I doubt France will become the next Greece, there are important complicating factors for France — namely, that it borrows in a currency (the euro) that it cannot unilaterally print. So the question we should be asking is not whether France passes a budget, but who backstops a country that gave up sovereignty over its currency?

The answer comes with a price.

Read more

Good afternoon from New York City where fall is in full swing. These are some of the busiest months of the year until we hit Thanksgiving and the holiday season.

Today I’m covering last week’s AI Summit that reminded me of Bretton Woods, minus the institution building and global collaboration of that event, which shaped the modern financial world. Later this week we’ll analyze the French bond market (for paid subscribers, so consider becoming one!), yet another example of the great repricing happening in developed markets globally.

But although the “Bretton Woods” AI event last week at the White House was smaller in scale than its financial predecessor, it could prove just as consequential. It charts the course for how the U.S. approaches the AI revolution.

The big problem, however, is how this summit differed from Bretton Woods. While Bretton Woods built financial institutions and systems with global participation, this AI summit built a pledge. That pledge chose not to regulate AI while deciding what shape AI should take — closed, with 6 big winners.

Subscribe now

“We cannot get in the way of innovation”, said every executive from every industry since the beginning of the American experiment. We heard this in agriculture, manufacturing, finance, and technology.

You know what happened each time we stayed completely out of innovation’s way?

Horrid conditions in fertilizer and meatpacking plants that workers had no right to organize against before the National Labor Relations Board. Adulterated food before the Food & Drug Administration. The Great Depression before federal securities laws. And we’re still trying to sort through the consequences of our social media and internet age. Meta just entered into an up to $18 billion settlement over Facebook and Instagram and the mental health problems suffered by teenagers.

During each era we were convinced by industry that they could self-regulate. Or that they needed to move fast and break things. That someone else was bound to win or that innovation would slow to a standstill if we even tried to regulate industry even a little.

Look, I’m a lover of capitalism. But with controls. With a conscience.

I’ve seen people blow themselves up - figuratively - on Wall Street trading floors when they thought they knew better. When they thought controls didn’t apply, that compliance was optional, or maybe my favorite — that it’s easier to ask for forgiveness instead of permission.

With a consequential technology like AI though, we may not have the chance to ask for forgiveness. While I don’t subscribe to much of the doomerism out there, history has shown us how foolish it is to approach any sort of Industrial Revolution without baseline minimum standards.

And this is particularly important with a technology like AI that permeates across everything, often with something resembling a mind of its own.

You would think an all-powerful technology like that would at the very least demand minimum standards not just created and self-enforced by industry, but by government of the people.

To be clear, I also don’t buy that the efforts of Dario Amodei (Anthropic’s CEO) and others to regulate AI are purely motivated by self-interest. That it’s all a conspiracy for him to achieve regulatory capture and build a competitive moat around his leading position.

That certainly could be a consequence, but if you have ever read any of Dario’s essays, you know he’s a pretty thoughtful guy who seems genuine about his concerns for the technology he has a big hand in shaping.

Perhaps we should be listening closer to his concerns coming from inside the AI house, while thinking critically about the best solutions.

But instead of setting minimum standards or solutions applicable to the AI industry as a whole, President Trump recently hosted some of the biggest AI players at the White House and did the exact opposite — he told each of the six companies that signed the accord that they’re responsible for their own technology.

This AI accord makes regulatory capture more likely

The accord’s four layers are built for companies that control their AI models, like the closed models of the 6 signatories.

There’s also been a rebrand by the way. It’s no longer called AI, but “super intelligence.” And the one page accord with its name is “morally binding”, according to President Trump.

So let’s get this straight — we’ve got a morally binding and voluntary super intelligence accord that’s really only built for companies that can control their models. It’s not designed for the Kimi K3s of the world that rely on open weights, basically the equivalent of scattering seeds — they can grow wherever someone takes them.

The big question is what happens to Kimi K3 and open weight models? Will they just be banned at some point? Is this accord simply the precursor to a law that requires full control of AI models in a way that only closed source can achieve?

Not only are we not regulating AI formally, we’re now demanding control in a way that’s only realistic for closed source systems.

The Trump administration doesn’t see it that way though so let’s take their view seriously. Over the weekend they appointed Jay Clayton, former SEC chair and the current director of national intelligence to lead the government’s AI initiative. Which means President Trump views AI through the lens of national security.

The biggest AI risk is China. They could beat us Americans in the technology race and/or use our AI technology to develop weapon systems or cybersecurity capabilities that threaten our national security.

The Trump administration prefers this self-policing accord approach because there are many existing laws that already apply to AI, as demonstrated by the current FTC probe into many of the top companies.

Of the problems that have recently occurred, like the Hugging Face incident or the recent alarm bells rung by former Anthropic and OpenAI employees (see David Robinson’s Atlantic essay), culture may be more of an issue than missing laws or rules.

The Hugging Face incident is the most serious AI incident to date. Guess where it happened? Inside a closed lab at OpenAI; a lab with exactly the kind of internal controls this accord leans on. And OpenAI didn’t detect it for about a week!

I know the Trump administration would probably argue I’m “over-lawyering” AI and that I’m too much of a former compliance officer who wants to push minimum standards where none are needed.

But in addition to that closed Hugging Face incident, I have history on my side. We’ve almost always waited to regulate new innovations properly. And too often we’ve had to endure pain, tragedy, and controversy before finally agreeing to put appropriate controls in place. From the Industrial Revolution to the internet.

Now with this AI (or super intelligence) accord, we’re choosing not to regulate the most powerful new industry of all time. And worse, we’re telling the industry what shape AI is allowed to take, which appears to be leading to more closed systems that can be controlled (and captured). As opposed to the open systems that the modern internet was largely built on.

Subscribe now

Why this AI accord deters open source

Take a look at the four layers of checks each of the six AI companies agreed to implement:

1. Robust internal controls

2. Internal oversight team

3. Independent external auditor

4. A committee within each company’s board of directors to evaluate reports from internal and external auditors

You cannot have robust internal controls or an internal oversight team over AI that’s fully open and released into the wild. Any programmer or maybe even vibe coder can take that program and use or modify it in unintended or unforeseeable ways.

What this accord really targets are the closed systems offered by OpenAI (ChatGPT), Anthropic (Claude), and Google (Gemini). The type of systems that operate as walled gardens that companies have complete autonomy and control over what goes on inside. These are far easier for auditors and boards to examine and evaluate, unlike an open model released outside the walls that companies lose control over once it escapes.

Notably, Meta and Nvidia signed the accord too, both of whom have released and championed open models. But will there be a separate accord for open source? Or are we setting up a future regulatory regime that basically bans or outlaws that technology?

“Frontier model” is also not defined under the accord. So do Meta and Nvidia get to keep their open models outside the accord’s scope? Does everyone else with a closed model they can control and oversee get slapped with the “frontier” label?

Perhaps we need a separate framework for open weights. One with rules on deployers and uses rather than on control of the model itself (which really only works for closed systems).

If we want smaller entrants and new startups to the AI race, just as Anthropic was a few years ago, we need to promote and appropriately govern open source technology. We need to leverage our collective knowledge to push and advance super intelligence.

It’s also important for free speech and freedom of expression. If there are only a handful of “regulated” models in the world, it’s easier for companies or governments to censor or regulate what they dislike. Whereas the beauty of the internet on the whole is that it’s mostly open and free. It was built on a general philosophy of open source that has been a boon for speech and innovation.

I fear we’re putting all of that at risk with an accord like this one — concentration of AI power is a precondition for this government model, not a mere side effect. It presumes established companies with strong infrastructure and sophisticated auditors and boards while ignoring the startups that largely built everything in the previous internet era (remember Google’s early days?).

The no regulation era for super intelligence is here

Did you notice who was not at the table for this Bretton Woods style AI summit?

The little guy.

The teacher concerned about students using AI. The average white collar worker worried about their job security. The junior software developer wondering if they have a career path any longer.

Almost every signatory at this table was a billionaire. And surprise, they all said they could police themselves.

Some fourteen months ago the Trump administration created an AI Action Plan that called open-weight models strategically important. But now this signature document of the no regulation era for “super intelligence” points in the opposite direction.

This is after months of folks from all corners of the AI universe accusing the likes of Sam Altman and Dario Amodei of trying to create an AI duopoly. Or trying to achieve regulatory capture.

This summit wasn’t an order of AI for two, but of super intelligence for 6. Please.

It focused on companies that can completely control their models. It never asked whether or how to best regulate AI to keep people safe. It just asked whether its governance runs through a special set of 6 companies or through everyone.

Surprise, they only ordered super intelligence for the six of them.

But we will all have to pay for their meal when something inevitably goes wrong.

Subscribe now

Share

Good morning and TGIF.

If you’ve followed me long enough, you’ll know I love bread. Baking it, talking about it, researching it, exploring its history, you name it.

Because just like law, finance, and geopolitics, when I’m interested in something, I go deep.

My home here in New York City may not have the same sourdough scene as San Francisco or, of course, Paris, but it’s competitive. So much so that it attracted a revered Austrian family bakery, RYE by Martin Auer. Their first U.S. outpost is just downtown here on Lafayette Street.

But I need to make one thing clear — I refuse to buy Martin’s bread.

For one, we need to make waiting in lines taboo. Because social media has ruined us, especially in New York City. I once passed the 57th street Chanel store on a random Tuesday and when I saw a line around the block I thought: (i) does nobody work? and (ii) we’re finished as a culture.

People have lined up for many other places in New York City, including Cafe Panna in nearby Gramercy. But at least there’s novelty in their ice cream as the flavors change. It also carries the Danny Meyer name, which is revered in this town.

Like RYE, Cafe Panna has received a lot of attention on TikTok. It started in 2019 and boomed serving ice cream during the pandemic.

But do people really go for the quality of the ice cream or to take part in the experience of the line and street cred of, “yeah, I tried it”? It’s not like they’re using magical or unique ingredients. Although the rotating flavors help keep the novelty alive.

RYE, meanwhile, sells one type of loaf. There are variations of its sliced bread with different toppings, so I suppose that could retain novelty, but the prized luxury item — the $60 loaf — is only one type of rye sourdough loaf for now.

That same loaf sells for $16 back home in Austria.

If you’re trying to lure the New York City luxury bread crowd, you’ll need to offer more than a fairly standard rye sourdough loaf for $60. Maybe a different flour. Maybe unexpected techniques or inclusions like chocolate chips or rosemary, cheese, jalapeños, chives, something! Sourdough customization is part of what makes bread so special and interesting.

But do the folks in line really care about that? I think not.

Because there are many great bread bakeries across the city’s five boroughs that don’t have lines around the block. They’re only lining up at RYE because it’s new and viral and slightly controversial with that $60 price tag.

So it’s not even really about the bread. It never was.

If you don’t opt for the entire $60 loaf, here are some of your other options:

• $6 for a slice with butter and chives

• $14 for a slice with roe on rye

• $15 for a quarter loaf

• $30 for a half loaf

$30 for a half loaf!? Inflation is still running hotter than the Fed’s 2% target, but that’s an insane price even for an oversized half loaf.

Thank you to Lidl for trolling them. They promoted their $3.99 loaf instead.

Why do I even care about any of this?

Because as I’ve written before, we have a bread problem in America. Here’s what I said over a year ago:

Scanning the average American grocery store bread aisle is a journey through preservatives and artificial flavorings. Calcium propionate. Potassium sorbate. Emulsifiers. Vegetable gums. Even ingredients that are banned across much of the world like potassium [bromate] are used in many commercial sourdough breads in the U.S.

RYE’s $60 loaf doesn’t fix any of this. In fact, it might even deter people from wanting to buy, eat, or dare I say bake, better bread. It forces bread into the wrong conversations about luxury signaling and coastal elites lining up for the next fad that illustrates how out-of-touch they are with normal humans.

Remember the Meadow Lane craze? Yes, the luxury grocery store selling $15 chicken nuggets. I refused to shop there too because these places care more about marketing than food quality.

You can find evidence of this at RYE from its ingredients alone. None of which are bad, but none of which justify the exorbitant price tag either. It’s made from 100% rye sourdough, which includes a family starter that’s apparently 50+ years old. They import the organic Austrian rye flour, but I assume they use New York City water (which is amazing) and salt (of unknown origin).

You know what all of those ingredients create? A good, standard loaf of sourdough bread.

So you’re setting yourself up for disappointment if you expect anything other than typical bread when visiting RYE. And yes, I can say that without having tasted the bread just from the ingredient list alone.

It will be hard for RYE to sustain the $60 price point. Unless, as The Washington Post review described it, the shop continues to operate more as a luxury boutique than a bakery.

Look, I know Manhattan rent is expensive. Labor is expensive. It’s not cheap to import all that organic Austrian flour. But no amount of stretching and folding and artisanal shaping can convince me that a standard loaf of sourdough is worth $60. Even if it’s a double loaf made from a 50+ year old family starter.

Instead of debating $60 loaves, we need to be discussing how to bring better quality bread to America. How to raise the minimum standard to bread that doesn’t include additives and substances, some of which are banned around the rest of the world.

We need to show people how good sourdough bread can be for a 1/6 of the price at RYE. How high quality flour is available and possible to consume if only we demanded it more instead of all the shelf stable alternatives.

I won’t be trying RYE’s $60 loaf, but I will keep baking. And I will keep hunting for the next great sourdough loaf. I’m not even opposed to paying $60 if the bread tells me a convincing story why — unique starter (that’s not just old), special inclusions, novel shaping, unusual grains, advanced milling techniques, etc.

If it has none of those things, it’s stunt food. Marketing. Status. Social media signaling.

Not bread.

Which is what we should really be talking about in America.

Subscribe now

Share

Good morning. I’m sitting here in New York State petrified by the heating oil prices that I and many other consumers in the Northeast are facing this winter. Millions of us in this part of the country rely on heating oil to heat our homes.

Now you might think, “What does this have to do with an essay on diesel?” Heating oil is almost chemically the same product. So whatever diesel prices do this fall, heating oil will likely do the same this winter.

And we’re at record diesel prices in America and around the world. Last week we hit an all time record of $6.52 per gallon. With few signs of relief not just because of geopolitical conflict and uncertainty — in Iran, Ukraine, etc. — but because of refining capacity.

I made a video about a week ago about this diesel crisis, wondering why the White House hadn’t started a debate about export controls (because we refine and export a lot of diesel in America). Well, that debate is reaching its climax now and I think pulling the export control lever hard is a terrible move.

Before explaining why, we have to understand why diesel of all products is the issue here. Most analysts and commentators focus on the price of crude, right? So why is diesel the problem?

Subscribe now

We’ve lost significant diesel exports from the Middle East and Russia. Refineries in these countries produce middle distillates (of which diesel is a type) for export and for various geopolitical reasons, these countries have stopped refining at their usual levels.

This differs from crude oil, which mostly travels on tankers and can be rerouted amidst geopolitical conflict. Diesel needs complex refining units like hydrocrackers and other special equipment so refiners can create it. You can’t just conjure up a hydrocracker overnight and put it online when there are supply disruptions.

We’re already near maximum output of diesel here in America. Diesel itself trades on a global market, so a refinery outage in the Gulf shows up in my local heating oil price.

These global dynamics incentivize U.S. diesel refiners to export to the highest bidder, which right now is often in Europe. The U.S. price therefore converges on the global diesel price, which is the economic reality at the core of the U.S. diesel export ban debate.

Now consider what happens if the U.S. restricts its diesel supply, keeping what it refines for itself. That move would probably work in favor of U.S. consumers. Temporarily. Oxford Economics, for example, projected that an export ban could cut prices roughly 30% within weeks (in at least some U.S. regions).

But you know who it screws over? Europe.

Many of the same folks who are incensed at the U.S. for attacking Iran, threats to take over Greenland and make Canada the 51st state, and overall undermining the rules-based world order that has kept the world largely safe since 1945.

According to Oxford Economics, wholesale European diesel prices could spike 40-50% if America stopped exporting. This would lead to increased geopolitical tension and less support for what once was America’s most important international allies.

Although Republicans might be happy with temporary reductions to U.S. diesel prices just ahead of the midterm elections.

Notice I keep saying temporary. Because refiners can’t sell or even store all of the trapped diesel that an export ban would cause. This would incentivize refiners to cut runs, shrinking supply for jet fuel and gasoline as well. Therefore, a diesel export ban could backfire incredibly on U.S. consumers with potentially higher gas prices.

The problem here is also uneven because in the U.S. our diesel surplus is in the wrong place. The East Coast actually relies on imports for about 10% of its diesel because we don’t have pipeline capacity that allows it to flow from Gulf refineries near Houston.

The New England region has no operating refineries and leans heavily on imports from — guess who! — Canada.

A diesel export ban may be great for Houston, but it won’t be great for heating oil customers like me in the Northeast.

What America should really do instead of creating more geopolitical conflict and economic disruption with a diesel export ban is extend the waiver of the Jones Act. This is an old maritime law that restricts foreign flagged vessels from shipping products between U.S. ports. There’s currently a waiver in place of this law that has helped push, for example, some 200,000 barrels to ship per day in April alone from the Gulf to the East Coast.

Instead of imposing an export ban, we need to extend the Jones Act waiver that expires on November 15th. This is the cheapest marginal solution, although it’s not a silver bullet fix.

But I wouldn’t bet on the U.S. opting for a simple legal waiver extension when the blunt force option of an export ban is on the table just ahead of midterm elections. Especially when it involves a president who has declared many national emergencies, and he probably has better legal support for this one than almost any of the others (such as tariffs). A national emergency on oil and diesel shortages globally is arguably more squarely within executive emergency powers.

Even if it’s a problem largely of our own creation in America.

The law may let the president implement a diesel export ban, but the long term economics and geopolitical consequences might actually deter him this time.

At least they should.

Subscribe now

Share

Good evening from a UN-gridlocked Manhattan.

Diplomats, dignitaries, and just about every politician has been in town this week. But while everyone focuses on an unaccountable UN Security Council and the soaring price of diesel fuel, we’re completely ignoring something important closer to home - the U.S. deficit.

Nobody likes to talk about the deficit because nobody knows what the deficit is. When I say that I don’t mean the actual definition — spending minus revenue for a fiscal year — I literally mean nobody knows what the deficit is.

Just look at the featured image for this piece. Three different primary sources. Three different estimates for how deep in the hole America is on her deficit this year.

Two of the estimates (the Congressional Budget Office vs the Office of Management and Budget) are some $212 billion apart!

Some of this variance comes from refunds the administration has been forced to issue for tariffs imposed under its now unconstitutional use of IEEPA (an emergency powers law the Supreme Court basically said cannot be used to tariff the world).

How can we not know how much more we spent than what we earned for a fiscal year? And why does nobody seem to talk about it, especially as it compounds our gross national debt, which recently passed $40 trillion with a “t.”

Net interest on that debt was $1.05 trillion over the past 11 months, up $111 billion year-over-year (or ~12%). While the U.S. Department of Defense outlay during the same period was $833 billion. So the total cost to service our debt (just the interest payments!) for the past 11 months exceeds U.S. military outlays by $219 billion.

This is not sustainable. We’re entering the height of U.S. midterm election season, yet you’ll hear little to nothing of our deficit because nobody can decide on the actual number. Treasury’s final FY2026 statement arrives in mid-October in a routine monthly statistical release, roughly three weeks prior to the midterm elections. But these statements regularly get buried and almost no campaign mentions them. And meanwhile, Treasury’s next quarterly borrowing estimates arrive on November 2nd, the afternoon before Election Day.

So read on if you want to understand why unattributable means unownable for U.S politicians, how this compares and contrasts to emerging market countries, and where our lack of financial discipline in America will end up instead (hint: it’s somewhere in the bond market).

Read more

Programming note: I’m speaking about bonds on Medium Day, this Friday, September 18th. You can register here and plan to virtually attend my session at 10:30 AM EST (it’s free!).

Good evening from Manhattan. I’ve been painting and refurbishing a pre-war apartment, so haven’t been as dialed in as usual. But I can’t stay on the sidelines here.

First, I want to commend Fed Chair Kevin Warsh. I honestly didn’t think he had it in him. I know the market priced in about 90% odds of a rate hike leading up to the decision announcement, but I was convinced that political pressure would be too much. In the weeks prior to the rate decision, the President, Vice President, and Treasury Secretary all dismissed inflation concerns and said the Fed should hold (if not lower) rates.

That’s a lot of pressure to defy. Especially in the wake of Trump taking shots at previous chair Jerome Powell for resisting his calls to lower rates. And don’t forget Fed Governor Lisa Cook, whom the Trump administration is still trying to fire.

Maybe Warsh doesn’t want to become Arthur Burns 2.0. The discredited Fed Chair from the Nixon era who kept rates artificially low to juice the economy in the short term, which we all of course paid for dearly later. Paul Volcker had to tow the American car out of the stagflation ditch with double digit interest rates.

Few people want to risk that historical record.

Despite Warsh’s defiance of any political pressure though, I don’t think the rate hike decision matters much against the macroeconomic and geopolitical backdrop we now face. But before we get there, it’s important to remember that Warsh is not the only Trump appointee on the Fed Board of Governors. Michelle Bowman and Christopher Waller both voted with the unanimous majority to raise rates.

Yet Trump still thinks the Fed’s Board is “hostile.” He also had this to say following the rate hike announcement:

This is why financial literacy is so important in our current moment. It’s a big reason why I write about this stuff because the premises here are just plain wrong.

Trump wants to treat the U.S. like a corporation with a great credit rating. Who therefore should get the best interest rates.

But that has nothing to do with where the Fed sets interest rates. A corporation isn’t tasked with keeping inflation at 2% or maintaining employment for the U.S. economy.

And even if you did argue it’s comparable, Treasury yields are the risk-free benchmark that practically everything else is priced off, and the policy rate is set for inflation and employment, not as a reward for creditworthiness. Even if we went the creditworthiness route, Moody’s cut the U.S. from AAA in May 2025, so none of the big three rating agencies say that America is “the best credit.”

He’s also confusing a trade surplus and deficit with income and loss. A deficit is not a fancy word for loss. For example, I have a deficit with Costco where I love to shop. Does that mean I have a loss because of Costco? No, it just means I give Costco money in exchange for a membership and whatever I discover on my weekly treasure hunt.

The same is true for America. We buy and import things from other countries largely because we need them. And then you know the best part of that trade? Foreigners purchased Treasuries, bringing dollars back into the U.S. (which hardly amounts to carrying any other country).

But all of that is starting to change, which should concern Trump and everyone else far more than a single Fed rate decision. It’s also precisely why today’s Fed rate hike can’t fix rising yields.

For one, the Iran war shows no signs of slowing down. The Strait of Hormuz situation remains uncertain at best. And as I’ve argued in the past, it’s not like Trump can wave a magic wand to open it — insurance actuaries and commercial shipping companies have to be comfortable to do business in those waters. It’s also relatively cheap for Iran to continue to disrupt the strait with missiles especially as it knows the U.S. interceptor stockpiles are dangerously low.

The energy inflation caused by this war is not going away anytime soon regardless of any Fed rate hike. Crude sits around $100/barrel with diesel trading as if it’s closer to $200/barrel.

The other main reason why I’m convinced that the Fed’s latest rate hike can’t fix rising yields is because it only sets the policy rate, as I explained here:

The Fed can't control the long end of the curve. Which like many other developed economies is suffering under the weight of higher global inflation (largely thanks to the Iran war), rising deficits, and outsized debt-to-gdp ratios.

What it’s really lost control over, however, is something even more fundamental that it cannot entirely influence — confidence in U.S. financial system. The Fed cannot unilaterally assure the market that U.S. debt is a good bet 30 years from now. Investors are increasingly demanding more term premium (higher yields) for purchasing that debt.

Term premium, which is the extra yield investors demand to lock up money for a decade, has turned positive after years below zero, but at ~0.76%, it’s about half its 60 year average.

Most of this year’s move in the 10-year Treasury has been from the market pricing a more hawkish Fed path. But that also means term premium has room to run.

It easily could keep running because there’s more supply now in this market coming from the hyperscaler big tech firms. They’re issuing so much debt at longer duration that it’s almost competing with Treasuries. Google, Meta, Amazon, and others are all issuing long-dated bonds that are competing for the same pool of long-duration capital.

And so long as the AI trade keeps humming, expect more inflation across the economy, with more demand for power, construction, chips, and data centers.

So rate hikes? Sure, they affect short-term debt like your credit card rate or what you can earn from a money market fund, but otherwise they don’t really impact the long-end of the curve that’s seeing the most action right now.

The 10-year Treasury breached 5.041%, its highest since July 2007.

The 30-year Treasury peaked at 5.401%, the highest it’s been since June 2007.

Now, I still commend Kevin Warsh here for doing the right thing and letting the objective metrics be his guide — as he said he would at Jackson Hole this summer — because this situation could be much worse if he didn’t. Had the Fed even held steady, the number on that 30-year yield would likely be much higher.

Bonds would have sold off.

Warsh has also added some certainty and confidence back to the Fed given his consistency since that Jackson Hole speech. He had already warned about persistent inflation there.

I don’t envy his position. Especially as his counterpart in the Treasury Department is intent on intervening in markets, disrupting the unfiltered data feed the Fed relies on to make decisions.

And especially as he’s likely to hear more noise coming from the White House.

But he must have known that when he accepted the job.

Subscribe now

Share

Good morning to what promises to be another wild week. Oil over $100/barrel. AI firms admitting they’re losing control of their powerful bots. And the Fed meets to debate raising rates.

So how can we make sense of this chaos?

There’s no stock price for America. Or any other country in the world for that matter. But we have bond markets, or what I like to call “the original prediction markets.”

Although James Carville may have my favorite line about the bond market:

“I used to think if there was reincarnation, I wanted to come back as the president or the pope or a .400 baseball hitter, but now I want to come back as the bond market. You can intimidate everybody.”

This former political advisor to U.S. President Bill Clinton said this back in 1993, and it’s very relevant today. Not just in America, but globally.

Just look at what’s been happening in the past few weeks:

• The UK just sold a 30-year bond at 5.8168%, the highest at any UK Debt Management Office auction or syndication since the office was founded in 1998

• Japan’s 10-year hit 3% for the first time since 1996, the exact long term rate Japan’s budget assumed for debt servicing costs (on a huge debt load that’s ~230% of GDP)

• The US 30-year is at post-2007 levels, with the Fed meeting Tuesday and Wednesday this week to discuss rate hikes

I’ve written multiple pieces on what I call the great repricing: the phenomenon of rising yields globally, how they check governments, and how we can understand the world through the bond markets.

You don’t need a PhD in economics or experience working on trading floors like me to understand the basics. And while most people are obsessed with tracking the stock market, there are no stock prices for America, the Eurozone, Japan, India, or any number of developed and emerging market countries.

But they all have bond prices on their sovereign debt. How a country’s long-dated debt trades is the closest thing we have to a verdict about how the world views that country. It’s something no president, prime minister, central bank chair, or any governing body can set — long-dated debt is set by the market.

Which is why more people should and can understand the basics.

So that’s why I’m talking about bonds on Medium Day on September 18th.

It’s free to watch and listen.

We’ll first start with a primer on bonds so anyone without any background can understand how it all works. How bonds get priced. What the yield curve means. Why a central bank only really controls the “short end.”

Then we’ll move into the “great repricing” and my view that the era of cheap money that has persisted since the financial crisis of 2007/2008 is over. You’ll see why I have this thesis as we circle the globe from the U.K. and Germany to the U.S. and Japan.

We’ll look at how the bond market has checked some of the most powerful people in the world — recently! From former UK Prime Minister Liz Truss (who had to resign within a month of announcing her mini-budget) to the case I’ll make Friday about how it’s shaped U.S. President Donald Trump’s decisionmaking, from tariffs and Greenland to the war in Iran.

Sometimes the bond market isn’t explicitly referenced amidst these world events. One of its most powerful influences is invisible. Like the invisibility cloak from Harry Potter, you may not know it’s in the room but a finance minister feels its presence during a budget debate and a president or prime minister feels its heat at the podium as they deliver shrinking promises to the public.

If there’s one thing I hope you’ll take away from the talk it’s that the bond market is omnipresent. Even when it’s not mentioned. Or cited.

After the talk you may start to look at yield curves more. You may consider what politicians say juxtaposed to where their country’s 30-year yield sits. You may think about their performance in office in more empirical terms — did yields rise or fall?

The bond market doesn’t care about political ideology. It has taken down politicians on the right, left, populist, conservative, expansionist, you name it.

What it cares about most is credibility and financing needs.

It tells the truth about constraints even though it doesn’t care about your values. And it doesn’t have to — it’s an unelected market that’s prone to overshooting or overreacting.

Which may be why governments try to fight the bond market all the time. They try buybacks, central bank interventions, yield curve control, or other creative maneuvers.

The bond market won’t tell you that your financial games or your spending is wrong. It will just communicate a price. And the market that sets that price can’t be voted out.

So is that a healthy check on a country or an undemocratic one?

I guess you’ll have to join me on Friday to discuss.

Subscribe now

Share

Good morning and welcome to the liveliest months of the year. We have major central bank decisions, elections, and market stories on deck.

I discussed some of these events, in addition to today’s topic — what’s driving the great repricing of global bond markets — on a recent podcast. Give it a listen for some good markets and political economy banter between me, a banker, and a political consultant.

I’m also speaking about bonds at Medium Day on September 18th. I’ve been writing on Medium for many years, and they’ve been doing this virtual event for the past 4 years or so, bringing together top experts on the platform. Registration is free and you can sign up here. My talk is at 10:30 AM EST: “How I understand the world through the bond market.”

Let’s now discuss what’s driving the great repricing of global bond markets and why the upcoming Fed meeting on Sept. 15-16th is unlikely to resolve it.

Subscribe now

You might think we’re in an oil shock. That three western central banks are simply being hawkish, likely to raise rates. Fed Chair Kevin Warsh indicated as much at Jackson Hole recently, saying this summer’s inflation readings didn’t show that underlying trends have meaningfully improved. The European Central Bank is also expected to hike rates (the market is pricing ~90% odds by year-end). And most expect not one but two Bank of England hikes by year-end too, although its governor, Andrew Bailey, pushed back on this recently.

The Eurozone’s inflation, like America’s, is persistently hovering over 3%, above the 2% target. Add in the Strait of Hormuz uncertainty and you exacerbate an already precarious situation.

There’s also less demand for long-dated government debt. As I have argued in previous pieces, AI is a new long-bond player competing for financing, and auctions have been weaker relative to pre-2022 levels. They haven’t broken though and foreign holdings of Treasuries, for example, are still around $8.5 trillion.

This leads some to believe that America “is a much stronger credit than it was a short time ago.”

Is he really threatening the Fed with more tariffs and trade wars?

What this ignores is the great repricing synchronization across the Western world, including those like Japan. Higher bond yields and elevated inflation are not just a single country’s story.

• Japan’s 10-year hit 3% on September 1 (the first time since 1996)

• UK’s 10 year gilt hit ~5.25% (the highest since 2008)

• Germany’s 10-year bund is around 3.35% (highest since 2011)

• France’s 10-year OAT is around 4.21 (highest since 2008)

• America’s 10-year Treasury is marching toward 5%, currently around 4.8%, with the 30-year around 5.25% (highest since 2007)

This is not an isolated story. We have five different sovereigns with five unrelated domestic political situations, all moving in unison on rising yields.

It’s the pattern I named the great repricing back in May and it’s only intensified. I expect that trend to continue for one simple reason — we are entering a new era where investors are no longer willing to ignore fiscal problems, soaring national debts, and no plans to control either.

This is a problem for us Americans in particular because we’re not being honest with ourselves about what’s happening. We have the President screaming from his perch on Truth Social that the opposite is true, while others announce doomsday prophecies that catastrophe will strike once yields breach 5% on the 10-year Treasury.

I think neither extreme is accurate. The 10-year yield still sits at or below the economy’s nominal growth rate, which means the federal government with its 6% deficits, above-target inflation, and full employment, is still funding itself at about the rate the economy grows.

That does not scream crisis.

But we have to be honest about what’s happening. Markets are repricing what they’re willing to pay for America’s — and many other’s — longer-dated debt. The era of cheap money is over.

And while we’re all busy arguing over what’s driving the move — term premium, inflation expectations, oil, the Strait of Hormuz — we’re missing a key point. Higher yields hurts America today because levels are now above what America’s fiscal machinery assumed. There’s little evidence this will change, even if the Strait of Hormuz miraculously opens (which I’ve argued for months that it won’t).

The Fed meets on September 15-16th. Whatever they decide on rates, they’re setting the front end of the curve, the policy rate. But the number that governs most of America is the 10-year.

It’s the number we have not been honest with ourselves about. It’s the number where every spending bill in Congress gets scored against at an assumed rate of 4.1%. All despite the 10-year trading near 4.8% since August.

If there’s any compounding line item in the fiscal story of America, it’s this one.

And it’s appearing in three places most people are ignoring.

Read more

Programming note: I will be back writing on September 9th, but you can find me on Instagram, TikTok, and YouTube in the meantime.

Good morning from an America battling over economic signals (among other things).

I still can’t get over how Kevin Warsh didn’t mention the U.S. Treasury even once in his Jackson Hole speech this past Friday. I’m not too surprised, I guess, but it’s odd that Treasury Secretary Scott Bessent seems to be diametrically opposed to Warsh when it comes to market signaling.

When it comes to long-dated U.S. Treasuries and certain FX markets like yen, market fundamentals are wrong and in need of “liquidity support” in Bessent’s view. Whereas Warsh seems to trust whatever markets tell him.

Warsh set up the case for higher rates in his speech on Friday. He didn’t give forward guidance but he did identify the asset classes and the signals he’ll be watching:

The Fed needs clear market signals, as unfiltered as possible . . . from market internals . . . the level and change in asset prices across sectors . . . the prices and trading volumes of Treasury securities. . . the foreign exchange value of the dollar . . . the cost and availability of credit . . . and the price of a broad set of commodities.

Bessent has intervened in two of these markets. So the U.S. Treasury is actively editing the data feed the Fed is relying on to make important interest rate decisions at its next FOMC meeting (September 15-16). And we’re in U.S. midterm election season.

Maybe Warsh doesn’t want to draw attention to the diametrical differences he and Bessent have when it comes to markets right now. Maybe he hopes it will blow over. Perhaps Treasury will surprise us by doing nothing in terms of buybacks starting on September 9th (they’re not required to buy, but if they do it must be at the new floor of $4 billion).

I was so fascinated by the Warsh and Bessent differences that I made a longer video discussing the implications of Warsh’s refusal to guide markets amidst Bessent’s efforts to edit their signals.

Warsh talked tough about inflation during his Jackson Hole speech, but part of me thinks that’s all it was ever meant to be — tough talk.

What happens to Warsh and his credibility if he holds rates after the market priced a 57% chance of a hike off his words?

Do we honestly expect him to vote to raise rates in three weeks with the backdrop of the U.S. midterm elections hanging over him? With the ire of the president hanging over his every vote, especially now with the electoral stakes so high.

The problem Warsh has, however, is that he will face the ire of the markets if the inflation data doesn’t improve and he refuses to raise rates in mid-September. Many Fed watchers and traders are currently working with the presumption that the Fed will raise rates unless the data indicates it’s not necessary. The same data Bessent is desperately trying to massage, with two interventions in August alone (yen and Treasuries).

Watch that August consumer-price index on September 11th. If that number comes in hot or hotter than expected or doesn’t seem to cool at all, Warsh will lose credibility if he doesn’t hike.

Markets took Warsh at his word though — for now. After his speech, the 2-year yield jumped ~12 basis points to 4.356%, the 10-year rose ~5 basis points to 4.726%, and the 30-year moved just 2 basis points to 5.211%.

So traders took him at his word on the policy rate (the short end of the curve that the Fed can directly influence), but ignored him on the long end (the exact part of the curve Bessent is intervening in).

As I listened to Warsh’s speech on Friday, I was hopeful that we have ourselves a Fed Chair who acts as an objective referee. Even if one of the biggest players (Treasury) is playing games with the numbers.

But I can’t help suspecting that the weight of political pressure will be too much to remain purely objective. The Trump administration has done everything possible to claim that inflation has been brought under control, despite the Strait of Hormuz still not being completely open.

Warsh hiking rates less than two months before crucial midterm elections would directly undermine this inflation messaging.

The official narrative is that the Fed is independent and nobody on the Board of Governors, including Warsh, take politics into account. But Trump is still trying to fire Lisa Cook, a Fed governor, and he threatened to fire former Fed chair Jerome Powell multiple times.

We’ve moved beyond the politicization of the Fed to the more pointed question of — how much political influence and pressure is on Warsh?

So much that I would be surprised if he votes to raise rates unless the data is so hot that he’s forced to.

Even with the Treasury editing the feed.

Share

Subscribe now

Good morning and here’s what I’m thinking about.

What’s one thing many Republicans and Democrats increasingly have in common?

They’re against data centers.

In today’s piece I want to explore how the data center has become symbolic of the “AI Boom Haves” vs the “Have Nots.” How it’s become symbolic of billionaires like Sam Altman, who have far worse favorability ratings than founders of the past like Steve Jobs.

Why?

The data center anger is justified for many reasons. But I also think much of it is overblown, and that moratoriums or pauses (like in my home state of New York) are a fool’s errand.

When what we should be arguing about is who pays for grid buildout. Because the AI revolution is happening whether we like it or not. And we need to make sure that the public isn’t subsidizing purely private gains with minimal social benefits.

The free part of the following essay explores data center economics and why who pays for grid buildout is the key question. The paywalled section is how state and federal regulators can solve that grid buildout problem.

So let’s cut through the partisan web surrounding data centers and address their pros, cons, and the solutions to their biggest problem — who pays to power them?

Subscribe now

When most people see data centers today they immediately think: tech wealth. These buzzing behemoths have become symbols of what so many despise about the AI revolution.

Billionaires and now even trillionaires behaving like the rules don’t apply to them. How they construct these megaliths of concrete, wires, and chips in our neighborhoods and communities with little concern for the consequences.

Perhaps this all started from the way AI was first sold to us. I wrote one of my first essays about AI a few years back when Sam Altman first testified to Congress after launching ChatGPT. There was a lot of doomsday prophecy and warnings of calamity, like AI accessing nuclear weapons or creating novel biological agents.

Then there were the warnings of mass job loss. How AI was like “social media on steroids.” How it could unduly influence our elections and disseminate mountains of misinformation on everything from medicine to politics.

We were programmed to hate data centers before they started popping up everywhere.

Now they’re easy targets because we can see and hear them in the physical world. Just head down to the data center alley in Loudoun County, Virginia to experience their size and scale first hand.

So it’s no wonder why so many people are protesting the presence of data centers in their communities. Or why this family of farmers in Kentucky, for example, refused to sell their land once they learned the buyer was Meta (who had tried to shield its identity through LLCs).

I understand this aversion and anger. Maybe the private equity guys who chop up companies and sell them for parts have finally met their match — data centers.

But I have to wonder whether the anger towards AI and data centers arrived before most people could identify a specific harm stemming from them. The data center symbol seems to have overshadowed reality. And if we cannot build data centers, say goodbye to the U.S. lead in AI.

Data center problems are real but the blame isn’t

First, the jobs. The tech industry oversells them, as if they’re building railroads or cars of the past that employed millions. When in reality subsidies have run around $2 million per permanent job. That’s a huge sum to support a single job that only pays a small fraction of that in salary.

And that general estimate is on the low end. States like Virginia, home of data center alley, report that roughly one permanent job is created for every $13 million invested. Think tanks like Brookings have found that county-level net employment effects are near zero.

So we’re not seeing a job boom from data centers. The data just doesn’t support it.

That’s before we even get to the water, noise, and electricity consequences.

Data centers drink a lot of water to cool their hot machines

People love to cite the stat that data centers consume less water nationally on average than golf courses. What they neglect is the fact that few of us experience anything nationally. Water usage concentrates regionally, so local impacts can be substantial even when the aggregate national share is trivial.

Roughly 57% of data centers draw from potable water supplies, with an estimated 80-90% coming from watersheds (lakes, rivers, aquifers) that are the same sources feeding public drinking water. Large data centers consume up to 5 million gallons of water per day, which is equivalent to a town of 10,000 — 50,000 people.

Most of this water goes through the municipal main at municipal rates. Reclaimed water infrastructure typically doesn’t exist or it was never required. Although a few big tech firms like Google use reclaimed or non-potable water at over 25% of its facilities, with Amazon using purified wastewater at about 20 of its data centers.

Using reclaimed or recycled water is doable for the data center industry, but it’s just not required and it’s certainly not the default approach. This adds stress to local water supplies because data centers tend to cluster together whereas golf is widely distributed across the country, which is why the local water impacts drastically differ.

Data centers aren’t too loud, but their consistent hum can be a nuisance

Imagine living next to a quiet dishwasher that never stops. It may not be the block party that your local zoning ordinance was designed to stop, but the consistent hum could definitely get annoying.

Nearly a third of Virginia’s data centers sit within 200 feet of residentially zoned property. Like many other locales, Virginia zoning ordinances put data centers in the same non-industrial category as office space.

Have you ever encountered a dentist office that sounds like a dishwasher around the clock?

As I’ve said before, many of our current laws fail to address the risks and challenges we face from an increasingly AI world.

Local zoning laws are just one example. Most of them in America measure a peak in decibel level. When they were written nobody considered the consistent, steady hum from a cluster of data centers.

The good news is companies like Amazon are already trying to adapt. They replaced 424 rooftop exhaust fans at a Manassas-area complex after neighbors measured up to 65 decibels at night. Their retrofitted fans reduced the sound by 10 decibels.

But there was no rule or law requiring Amazon to take this step. They likely just saw the writing on the public opinion wall — more people everyday hate data centers.

Class actions have formed in New Jersey, Wisconsin, and Mississippi to bring data center noise nuisance lawsuits. I’m not aware of any of them reaching summary judgment or trial — for now they are just allegations — but it shows you how angry people are who have to live within hearing distance of these complexes.

I can’t say I blame them. I would feel the same way if one of these humming megaliths were in my backyard, but I’m hopeful noise reduction strategies like the one used by Amazon can work more broadly. Like water and electricity, solutions require more than courtrooms and social media comment sections — they demand updates to zoning codes and local and state laws. This is especially true for electricity.

Data center power costs hit everyone’s bill

This is the big one. The real fight. Unlike the job numbers, nobody disputes the bills went up. Instead the argument is over how much of the demand behind those bills is real and who should pay for it.

PJM runs the wholesale electricity market for 67 million people across 13 U.S. states and Washington, D.C. Its independent market monitor — Monitoring Analytics — estimated that data centers accounted for 63% of price increases in the 2025/2026 auction. That amounted to roughly $9.3 billion that was recovered from customers, as capacity costs are paid in advance and spread across customers.

In a normal market, a spike in demand would be met with an increase in supply, but in power markets this isn’t easy to do. Even a developer with their docs and permits and proposal put together wouldn’t be able to bid into the power capacity auction for 3 years.

So it takes approximately 3 years to go from price signal to supply response (and going live with a new plant takes even longer).

In the meantime, customers have to absorb the price spikes and electricity supply shocks. Which is why people are pissed about data centers.

The electricity load shouldn’t be unmanageable though. Neither should that queue. But these are issues I rarely hear being raised from data center protests or other op-ed pieces.

Across the board we’re dealing with outdated laws that don’t address our data center challenges. Water runs through the potable main because reclaimed water was never required. Data center hums reach bedrooms because zoning called it an office. And the electricity costs land on your bill because no rule addresses what happens when data centers come to town and drain power capacity.

None of this is a billionaire problem despite the data center symbolism. That’s coming from someone who has been critical of the billionaires running the AI show.

The data center problem is really a local and state government problem. A ratemaking problem. A permitting problem. And they all have solutions that don’t involve tarring and feathering billionaires even if that’s what you want to do.

Read more

Good morning from yet another eventful weekend. This time, Canada.

Before we dig into the trade war that was sparked, yet again, between Canada and the U.S., recall this line from Mark Carney’s Davos speech earlier this year:

"[T]here is a strong tendency for countries to go along to get along. To accommodate. To avoid trouble. To hope that compliance will buy safety . . . Well, it won't."

I commended Carney at the time for invoking Vaclav Havel in his speech and calling for the middle powers like Canada to band together. How those powers should stop performing and pretending that the rules-based world order exists for them.

America and other great powers have exempted themselves from this rules-based system whenever convenient, and that’s especially true today. Just this year — Venezuela, Iran, and the continuation of tariffs all illustrate how America will exercise power however it chooses.

On that tariffs point, Canada had tried to comply. They tried to “go along and get along”, as Carney had said in his speech. They accommodated Trump himself in the deal he personally negotiated, certifying certain Canadian goods under the rules of his trade agreement.

But just a few days ago, Canada got hit with 50% tariffs on about $20 billion of goods anyway. On products that were supposed to qualify under the deal.

Over a series of proclamations, the Trump administration erased months of negotiating with Canada. It set a deadline, paused the tariffs at the last minute to gain leverage, then executed when the leverage didn’t work.

Canadian companies had gone to great lengths to prove their products were made in North America, documenting their origin via the required paperwork. Only for the Trump administration to reach for a 1930s law that never considered trade agreements. They’ve now applied it to tariffs regardless of whether the required paperwork existed or not.

The administration had to reach for this arcane law because the Supreme Court ruled the emergency powers statute Trump had been using didn’t authorize tariffs at all. So they searched for legal authority that might still work, landing on one that predates modern trade agreements altogether.

The stakes here are about more than the $20 billion in Canadian goods that just got hit with 50% tariffs. They’re about what every other government watching this ordeal just learned — even a signed agreement with the United States is no guarantee.

Subscribe now

Compliance does not buy safety

If Canada can’t secure safety and certainty from America right now, how can anyone expect to?

Canada is probably the closest U.S. ally. It has the longest border with America. Its trade deal was negotiated by the U.S. President himself. And Trump even called it the best deal in history at the time.

So even if some form of agreement is eventually reached in the coming days, the damage here is done. The toothpaste has erupted from the tube. And nobody else in the world will want to use it to brush their teeth.

This is the channel through which a country’s risk premium widens. It won’t happen in a single trading session, but it will eventually appear in the spread between what a borrower pays and what the math says they should be paying.

Firms and governments now have to price political risk into every North American supply chain, just as they would for supply chains in say Latin American countries. Maybe not to the same degree, but it’s definitely a consideration, regardless of whether the U.S. Supreme Court struck down the emergency powers tariffs.

Trump has demonstrated a willingness to find new legal mechanisms to attempt to justify new tariffs. And the courts are always playing catch up. Even if they do check his executive actions, the damage is usually done.

This is exactly the type of uncertainty and volatility I regularly heard emerging market traders and salespeople explain when I covered an EM rates trading desk. Once risk premium increased, spreads widened and many bond buyers balked, sending yields higher.

I just always heard this about countries like Argentina. I never expected it to apply to America.

“Agreements bind you, not us”

If you really want to understand the crux of Canada’s frustration, look no further than the U.S. asking for terms limiting Canada’s ability to make trade deals with other countries. So the trade agreements bind Canada, but not America.

There may be an asymmetrical power imbalance, but the ask only makes sense if you assume the U.S. has permanent leverage over how Canada negotiates bilateral trade terms with a country like China.

Canada is betting that America won’t have this leverage forever, if they even do today.

Notice too how the Trump administration isn’t saying Canada technically violated any trade agreement with the U.S. They’ve instead attempted to route around the question of whether the agreement applies and use a law from the 1930s in a way that’s never been done before.

Remember, the Supreme Court held this past February that the emergency powers statute didn’t authorize tariffs at all. This took away the administration’s primary tool, and what’s followed is a winding path through older and narrower statutes until they found this one from the 1930s. It had never been used for tariffs like this, and it only works because it predates the modern trade-agreement system (by not recognizing it).

Of course this will be challenged in the courts, but that, of course, takes time.

In the meantime, we have a trade war.

Some American grievances are real

Now here’s the part that Trump is somewhat right about. Over the past 18 months while this U.S.-Canadian trade deal was being negotiated, Canada has been boycotting American goods. I attribute this mostly to Trump’s manner of dealing with Canada (like threatening to make them the 51st state).

Regardless, the economic effects on U.S. businesses have been real. Canadian provincial liquor monopolies, for example, stopped buying U.S. spirits, with U.S. alcoholic beverages down about 81% ($582 million). Canadian imports of U.S. motor vehicles are also down roughly 22% ($5.6 billion).

But even if you argue that Canada could have done more to smooth over U.S. business relations while the agreements were negotiated, this still doesn’t excuse the U.S. making unreasonable ultimatums, including restrictions on making trade deals with other countries. Or trying to make other 11th hour material changes.

The fact sheet itself from the administration states that the U.S. did not agree to renew the agreement because it wasn’t sufficiently beneficial. Months of negotiations be damned.

Alternatives to America are being built

This pains me. Yet we need to be real about something — 73% of Canadian exports go to the U.S. Only 4.4% of Canadian exports go to China.

So on the numbers alone, it’s not like China is going to swoop in overnight to replace America. But there is a second alternative.

Europe. The EU has been positioning itself as the go-to alternative for those middle power countries that Carney was talking about in his Davos speech. It has tried to market itself as a welcoming place for those who want to band together and avoid the binary choice between the U.S. and China.

The EU has even signed trade agreements with Australia, India, and Indonesia. And Canada is the only non-European member of SAFE, which is the EU’s defense procurement program.

Some may argue that these aggressive and unreasonable moves against Canada will push them into the arms of China, but that’s unsupported by the trade data. The more likely reality is the risk premium added by every country that does business with America following this Canada debacle. A risk premium that doesn’t go away even if a deal is resolved this week.

Uncertainty will continue to linger, just as it does after the many failed deals to reopen the Strait of Hormuz. Because just as Trump can’t force insurance actuaries to change the math on the page, he can’t force trading partners to amend what their new risk premium for America is telling them.

This is the greatest consequence of the trade war with Canada. Not the loss of business or tit-for-tat tariffs with America’s peaceful northern neighbor, but the loss of confidence everyone else now has for the U.S.

The world has seen now what compliance buys from America.

It’s not safety.

Subscribe now

Share