Here's a story from the last couple of weeks that a lot of people missed, mostly because the stock market looked totally fine while it was happening.
A hedge fund called Situational Awareness went from managing about $45 billion to around $10 billion. Not over a year. Over roughly three weeks in July.
The guy running it is Leopold Aschenbrenner. He's 24. He used to work at OpenAI, and in 2024 he wrote a 165-page essay about where AI was headed that got passed around Silicon Valley like required reading. The essay made him kind of famous. He named his fund after it and started taking money in July 2024. He had never managed money professionally before that. Earlier in his career he'd worked at FTX, which is a detail people are bringing up a lot now.
So what actually went wrong?
He borrowed a lot of money
This is the whole story, basically. Everything else is decoration.
When a fund buys stock, it doesn't have to use only its own cash. It can borrow from big banks, put up its existing holdings as collateral, and buy more. That's leverage. If you have $10 and borrow $30, you're controlling $40 worth of stuff with $10 of your own.
Reports say Situational Awareness was running leverage as high as 400%. That means for every dollar of actual money, there was something like four dollars of exposure in the market.
When it works, it's great. Your gains get multiplied. When it doesn't, the same math runs in reverse and it runs fast.
The fund was heavily concentrated in AI hardware companies. Memory chip maker SK Hynix, cloud company CoreWeave, that kind of thing. Not a wide spread of bets. A narrow set of bets on one idea, made much bigger with borrowed money.
Then the calls started coming
In July, tech stocks pulled back. Nothing catastrophic by normal standards. But when you're leveraged that heavily, "nothing catastrophic" is enough.
The banks that lent him the money (Bank of America was one of them) started asking for more collateral. That's a margin call. Your positions dropped in value, so the loan is now riskier than agreed, so you need to put up more money or sell something.
He had to sell. Selling pushed the prices of those stocks down. Lower prices meant more margin calls. More margin calls meant more selling. Round and round.
By the time it was over, the fund had lost about 67% and had to hand off basically all of its public stock positions to Ken Griffin's Citadel, at a discount. Citadel bought the problem. And once the market knew the forced selling was done, a lot of those AI stocks bounced right back, which meant Citadel made money almost immediately on what it had just picked up.
That part stings.
The weird part: the market was fine
While all this was happening, the S&P 500 was sitting near record highs. If you only checked the index, you'd have no idea anything unusual was going on.
What was going on underneath was that the stocks everyone had been piling into suddenly reversed hard, and the stocks nobody wanted went up. One analyst at BTIG called it the fastest momentum crash he'd seen. For a fund betting on the popular stuff and betting against the unpopular stuff, both sides lost money at the same time. There was nowhere to hide.
Why anyone outside of finance should care
Bank of America's CEO, Brian Moynihan, called this a warning shot. That's a fairly loaded phrase coming from a guy whose bank was one of the lenders.
His point is that borrowed money is all over the place right now, not just at one fund. Margin debt at retail brokerages is up 49% according to FINRA. South Korea had 1.2 million margin calls from people using leveraged ETFs. Plenty of professional investors are carrying near-record borrowing too.
None of that is a prediction. Leverage being high doesn't mean something bad happens tomorrow. It means that if something bad does happen, it gets worse than it otherwise would, and it spreads to people who weren't involved in the original bet. Aschenbrenner's forced selling dragged down chip stocks for everyone holding them, including people who'd never heard of him.
The lesson everybody's repeating
You'll see a version of this everywhere: he might have been right and still lost.
The AI buildout thesis in his essay isn't obviously wrong. Chip demand is real, data centers are real. If he'd held those positions with his own cash and waited, he might have been fine. But he couldn't wait, because the loans didn't let him. The banks decide when you're done, not you.
There's an old line about markets staying irrational longer than you can stay solvent. That's basically what happened here, just compressed into a few weeks and with a very young guy at the center of it.
Worth noting that not everyone thinks this is over. Michael Burry, the guy from The Big Short, has been arguing for a while that a lot of AI demand is being propped up by financing arrangements rather than actual customers paying for things. He's been early before. Sometimes very early.
Either way, one fund nearly disappearing because of borrowed money is a fairly old story wearing new clothes. It happened with Archegos in 2021. It happened with Long-Term Capital Management in 1998. Now this article.