MacroVoices #556: He Was Laughed At for Six Months, Then Started Doing a Different Sum

Contents

1. The six months he spent being laughed at, and the disaster that didn’t arrive
2. Thursday morning, running down his usual row of prices
3. $19 trillion, $1 trillion, $2.6 trillion
4. There’s only one buyer
5. Being right and making nothing: the problem is tempo
6. Worth a look
7. The one thing to take with you

In the old days coin was struck and counterfeiters hunted down; now lead and tin are allowed in with the bronze. Cast in a clay mould it comes easy, and good from bad no longer tells apart, so the deceit runs long.

—— Du Fu, “Song of the Year’s End” (Tang, c. 768; translation mine)

In episode 556 of MacroVoices, released 29 October 2026 and recorded on the 27th, Erik Townsend brings back Luke Gromen, founder of Forest for the Trees. When Luke was on in April he said the closure of the Strait of Hormuz could run as long as 4 July, and listeners mocked him for it; the strait is still closed, now twice as long as that call. What he spends most of this hour on is a different sum: $19 trillion of US consumer debt, where on his arithmetic a 5% default rate is about $1 trillion, against a total equity base of $2.6 trillion for the entire US banking system — the first number is 40% of the second. It holds only if AI removes white-collar jobs at the pace he expects, which he calls an inference, and he repeatedly calls himself a tourist on the military and oil-market questions.

The six months he spent being laughed at, and the disaster that didn’t arrive

Erik opens with the hard question: you were right about the duration, and none of the consequences you and I expected have shown up.

Luke gives the situation a name first. Schrödinger’s closure. The Iranians say it’s closed, the Americans say it’s closed, and according to the president it is up and running — depending on who you ask, it is closed and open at once.

Then he spends ten minutes on where he was wrong, which is my favourite stretch of the episode. The direction was right and the speed was wrong. He underestimated how far China would constrain its own demand and run down its reserves, and he underestimated the coordinated releases from Western strategic reserves. That extra supply absorbed the shock.

The food section is the counterintuitive one. What he hears is that farmers in lower-income places have done the arithmetic and simply aren’t ordering fertiliser — it costs too much, the maths doesn’t work. So the first stage of global food inflation doesn’t show up this year; it shows up as lower yields next year. Bearish in the near term, deeply bullish for crop prices later. He flags his own first-world bias here: he assumed someone who knows trouble is coming would stockpile, and the real-world response is that if you can’t afford it, you don’t buy it.

Erik pushes on refineries: Russian, Saudi and Iranian refineries keep being struck, a refinery takes ten years to build, so why isn’t the market frightened? Luke says the ledger doesn’t balance for him either. The president says the strait is nearly open, reserves are being drained, and refineries taken offline means less demand for crude from those refineries — add it up and there is a lot of incremental crude versus six months ago, while the price of crude is much higher. The only explanation he can make work is that China has been buying it all off the lows, and he says he doesn’t know.

Thursday morning, running down his usual row of prices

Then comes a scene I kept thinking about afterwards.

Thursday 22 October, he sits down at his screens and clicks through his usual list of prices. Gold, down 50 dollars. His reasoning: if gold is down like that, oil must be up big. Pull up oil, up 3%, fine. Gold down big and oil up big means bitcoin — which he treats as a rough read on liquidity — should be hit. Bitcoin down 3% or 4%, fine. The dollar index should be up; it’s up, not by much, but up.

Follow that chain to the last box and the 10-year Treasury yield should be up big. He pulls it up, and the yield is down big.

Down as much as 20 basis points intraday. He has been in investment research for 31 years and says he can count on one hand the days the 10-year fell 20 basis points: 9/11, Lehman, COVID, catastrophes of that order. And this was a Thursday with nothing on it.

He then offers a string of guesses, with the hedge stated up front: he has no facts here, only strong suspicions. It could be France, it could be the Middle East, it could be someone getting advance word on the midterms. He says Scott Bessent has warned — his word is “warned slash threatened” — that if the Democrats win he will be forced to compress his entire fiscal consolidation into the two months from 4 November to 3 January 2027. Luke’s description: it fits both men’s personalities, if I can’t win at Monopoly I tip the board over. In that world stocks fall, commodities fall, and you want to own bonds.

His own verdict on the passage is not to play five-minute macro — three trading days in total, and the yield has been pinned at 4.99 since. But the break in pattern set off his radar.

$19 trillion, $1 trillion, $2.6 trillion

Then the main course, the part he says nobody is discussing publicly while he has been writing it up for clients for months.

The setup is layoffs. His list: Kroger cutting 37% of its corporate workforce (he flags it as the number he remembers), a company that sells food and whose business model hasn’t changed in a long time; American Express saying last week it is cutting 5% because of AI; Barclays saying last week it is cutting 5% because of AI; Microsoft the previous day announcing buyouts for 27% of its workforce, a figure he states two different ways on air, so I’m relaying it as he said it. Microsoft, he says, is the “we’re going to need a bigger boat” moment for the AI employment argument.

Then the arithmetic. $19 trillion of consumer debt, mostly mortgages, credit cards and auto loans. Assume AI-driven job losses and wage deflation push 5% of it into default over the next two or three years: roughly $1 trillion. The comparison: per the Fed, the entire equity capital base of the US banking system is $2.6 trillion. So every 5 percentage points of consumer default eats 40% of the equity base of the US banking system.

He walks the banks’ four options into walls. Stop lending. Sell the Treasuries they hold as high-quality liquid assets, which puts another seller alongside the Treasury department. Issue equity, which puts another claimant on capital alongside the Treasury department and AI. Or slow the economy down a lot — and the US debt position depends on growing out of it.

He says the sum has already lodged in his own head: managing his own money, he is sitting with 30% of his liquid net worth in cash and T-bills. He doesn’t want to short into a rigged market, because under fiscal dominance liquidity gets supplied the moment the Treasury market acts up. He notes that banks and consumer lenders have badly underperformed the S&P over the last three to six months, then asks the question — if AI works, why would I want to own any of this?

There’s only one buyer

He pushes the same logic out to the fiscal side. The US is running $2 trillion deficits. A bursting AI mania brings a recession, and he says the last three recessions widened the deficit by 6, 8 and 10 percentage points of GDP; using 6 to 8 gives $1.8 to $2.4 trillion, which turns $2 trillion into $4 trillion.

From $4 trillion he keeps adding. Banks selling Treasuries to offset credit losses, $200 to $300 billion a year. Foreign central banks on pace to sell nearly $600 billion since February. Hedge funds own 8% of the Treasury market and will de-gross into a recession, call it another $200 billion. That stacks to $5.1, $5.2 trillion of supply.

So he asks who the buyer is. Retail, as they lose their jobs? Pensions, as they pay out early withdrawals? His answer is the line the episode opens with: there’s only one buyer, the Fed. They will print money and buy Treasuries, and they will say “we’re not monetising the debt, we’re not monetising the debt,” and it will all be bullcrap.

The sentence I most wanted to write down is next: that policy response can’t happen before a crisis, because otherwise it’s just handouts for the banks. You have to scare people first.

Erik then reads from a filing. Anthropic’s newly filed S-1 says concerns have been raised inside the organisation and by researchers, regulators, policymakers and industry participants that sufficiently advanced AI systems, if not developed, deployed and operated within appropriate safeguards, oversight and security measures, could pose catastrophic or existential risks to humanity. Erik’s comment: he is fairly sure no S-1 ever filed with the SEC has said, by the way, our stuff could end humanity.

Luke answers with a cartoon: three people in rags around a campfire, saying we ended up destroying the Earth, but for a moment there we created some shareholder value.

Then the darkest stretch of the hour. Suppose Bessent and David Sacks hold AI up and won’t let it fail — he thinks they are saying exactly that and will probably succeed — what have they won? China’s one-child policy leaves it with collapsing demographics; the US has a growing and heterogeneous population instead. As AI takes jobs out in 5% and 10% and 27% chunks, the collapsing-demographic societies are homogeneous, while the US is a growing heterogeneous society with not enough jobs. Google France, he says. Google the Middle East. He reaches for Peter Turchin’s “End Times”: elite overproduction plus massive wealth inequality always ends the same way. His expectation is that from 2027 this gets litigated with Molotov cocktails in American city centres. He adds a line straight after: people will call me a doomer, and this is critical thinking.

Being right and making nothing: the problem is tempo

Here’s where this episode hit me. You get the judgement right, you make nothing from it, and you spend six months being laughed at on the way.

The strait is the specimen. Luke had the direction, the cause and the duration right — closed twice as long as his call — and the oil price didn’t travel the path he and Erik expected, with long stretches where the market looked untroubled. The difference is tempo: three buffers absorbed the shock in sequence — China constraining demand, reserve releases, and refineries taken offline which cut demand for crude — and until the buffers are used up, price doesn’t follow structure.

That changed how I write a judgement down. Two layers. One is the structure: where this thing goes and why. The other is a buffer list: what absorbs the shock before it reaches price, and how much of each is left. Fertiliser is the cleanest example — one structure, bearish in year one (nobody can afford it, demand falls) and bullish in year two (yields come down). Write only the structure and you will conclude in year one that you were wrong.

The hole I’ve fallen into runs the other way: a detailed structure with the buffer column blank, so every time price refuses to cooperate I re-litigate the structure, and after a few rounds I cut the position at the ugliest possible place. Now I ask one question first — is this a buffer working, or is my structure broken? The two call for opposite actions. One asks you to wait, the other asks you to leave.

Worth a look

• MacroVoices episode 556, released 29 October 2026 and recorded on the 27th, the source for this piece
• The same show’s 16 April 2026 episode, where Luke made his original call on how long the Strait of Hormuz could stay closed
• Forest for the Trees (fft-llc.com), which he describes on air as aggregating publicly available data to identify developing economic bottlenecks
• Anthropic’s S-1 filing, where the risk disclosure Erik reads out can be found via the SEC’s EDGAR system
• Peter Turchin, “End Times”, the source of the elite-overproduction and inequality argument he cites
• Fed statistics on the equity capital base of the US banking system, the origin of the $2.6 trillion in his sum

The one thing to take with you

One line: in a set of things that normally move together, the moment one of them breaks ranks is the moment that carries information.

That’s what Luke’s Thursday is worth. Gold, oil, bitcoin and the dollar all behaved exactly as his chain predicted; the fifth box didn’t, and he neither explained it away nor added to anything. He wrote down “this doesn’t make sense” and turned his radar on. I took none of the guesses that followed as answers, but I did take the move: the break in pattern was visible to him because he had a list.

Here’s something I’ve tried that has nothing to do with investing. Tonight, take a sheet of paper and write down the four or five things you check without thinking every morning — hours slept, your mood getting up, yesterday’s screen time, the number in your account, whether a particular person replied to you. After each line, write the link you believe in (“slept badly, so that’s a day I buy things I don’t need”). Stick the paper somewhere you’ll see it. On whatever day one of those boxes breaks ranks, write one line on the back — which one didn’t follow the rule today — before you decide whether to act on anything. Flip the paper over a week later; the lines you managed to write are the thing you couldn’t see before.

Raw Markdown

This article is an educational discussion of investment method. It is not advice to buy or sell any individual security, offers no target prices, and does not analyze any current holding. Investing carries risk; make your own decisions or consult a qualified professional.