Last week the Federal Reserve looked at a hot economy and decided to sit still. Rates stayed at 3.5% to 3.75%, and the vote was nine to three, with all three dissenters wanting a hike rather than a cut. That is the first time since September 2016 that three committee members have pushed against the chair in the same direction. Kevin Warsh told reporters afterward that he had asked for a good family fight and got one.

It surprised me a little. I had expected the conversation to be about when rates come down, and instead a third of the voters wanted them to go up. The bond market seemed to have a similar reaction. The 30-year Treasury yield jumped after the decision and closed the month at 5.27%, its highest level since 2007.

The stock market has been rough on top of that. Global semiconductor stocks have shed around $3.3 trillion since June, a slide that picked up speed when TSMC raised its full-year capital spending guidance to $60 to $64 billion and investors started asking who actually pays for all of that equipment. And in San Francisco, Situational Awareness, the $45 billion AI hedge fund run by former OpenAI researcher Leopold Aschenbrenner, was margin-called down to around $10 billion and forced to sell its leveraged positions to Citadel at a discount.

Put those together and you can build a reasonable case that a crash is coming. A friend texted me the other night and asked me whether I'm worried about a recession. The short answer is no.

Money is expensive right now

A 30-year Treasury pays 5.27% today, backed by the full faith and credit of the US government. The interest is exempt from state and local taxes, although the federal government still takes its cut, so it is not quite tax-free. Even so, an investor can lock in better than 5% for thirty years while taking close to zero credit risk. Stocks have to promise meaningfully more than that to be worth owning, which is a big part of why they have wobbled. And because nearly every other borrowing cost in the economy gets priced off the Treasury curve, mortgages, corporate bonds, and small business loans are getting more expensive too (but nowhere near the highs we've seen).

Investors have not faced a risk-free hurdle rate this high since 2007. Expensive money is how economies cool, whether or not the Fed ever adds the hike those three dissenters wanted.

Why I think inflation is living on borrowed time

We do have inflation today, and those three dissenting votes tell you how seriously the Fed takes it. I think that inflation is going to fade, and I think AI is the reason.

I've written before about the long pattern of manufactured goods getting cheaper. Televisions, computers, clothing, cutlery, anything that can be made at scale and shipped in a container has fallen in real price for decades. The US Bureau of Labor Statistics reckons television prices, adjusted for quality, have dropped about 6.5% a year since 1950. In AI Won't Make Us Rich. It Will Make Everything Cheap. I worked out that a house cost about 75 dinner sets when my parents married in the 1980s and costs about 1,400 dinner sets now. The plates got cheap. The things that need humans did not.

That second group is what has driven the inflation of the past few years. Doctors, lawyers, accountants, consultants, software developers. Their prices have risen far faster than the price of goods, and services have stayed expensive for a simple reason. Services are made of skilled human time.

AI is the first technology that puts direct price pressure on skilled human time, because what AI sells is intelligence, and intelligence is the costly ingredient in most of those services, from legal advice to medical triage to writing code. If the cost of knowledge work falls the way the cost of televisions fell, the cost of living starts to drift down for a lot of people. I'd put the odds of that comfortably above even. I'm not predicting prices across the whole economy fall next year, and I'm not predicting a deflationary bust. I expect the pressure behind these high inflation prints to fade, with a new structural force underneath pushing the price of knowledge down for the first time.

The margin call was a leverage story

Situational Awareness was reportedly running leverage of as much as 400% into a sector that had climbed about 130% in a year. At that gearing an ordinary pullback becomes a forced sale, the forced sale pushes prices lower, and lower prices trigger the next margin call. Goldman Sachs and JPMorgan have spent the past month making similar calls to other funds.

A blowup like that says a great deal about how much borrowed money was riding on AI stocks and rather less about what AI is worth. An unleveraged investor holding the same names is sitting through an uncomfortable correction. Nobody is forcing them to sell anything.

The 87%

I started my first business in 2009, in the middle of the last serious recession, so I have some feel for building when the headlines are ugly. When I look at a downturn I start with one number. The worst single year for US output in the past hundred years was 1932, when GDP fell about 13%. The full slide from 1929 to 1933 took roughly a quarter of the economy, and that remains the worst case on record. The 2008 to 2009 recession, the one I started into, knocked off about 4%.

Turn the worst of those numbers around. Through the hardest year of the Great Depression, 87% of the economy kept operating. People still ate, shipped, built, hired, and bought. It was harder. Customers wanted more for less, and more competitors were chasing less work. That pressure hones a business. In any recession there is a group that keeps trading and a group that disappears, and you have to decide which one you are going to fight to be in. I will fight my tail off to be in the 87%, the same way I did in 2009.

So no, I'm not worried. The economy is running hot, money is expensive, and some heavily leveraged bets are being unwound in public. Parts of the ride will be bumpy. But in interesting times there are chances to make opportunities, and this is an interesting time.