A 439% return, then a fire sale: How a 20-something's $45B AI fund cratered in days

It took less than a week. Leopold Aschenbrenner’s AI-focused hedge fund, Situational Awareness, went from reporting a gaudy 439 percent return for the first half of the year to selling nearly its entire public stock portfolio to Ken Griffin’s Citadel. The trigger wasn’t a flawed thesis—it was a margin call.

Aschenbrenner, the mid-20s former OpenAI researcher who gained a cult following for his essay predicting explosive AI growth, had loaded up on borrowed money to bet big on AI infrastructure plays. When stocks like memory maker SK Hynix dipped and his short against Adobe went sideways, the banks came knocking. The fund, which had swelled to $45 billion in assets with just eight employees, couldn’t meet the calls. The forced sell-off happened right before some of those same positions bounced back.

The irony is hard to miss. Aschenbrenner’s core argument—that powerful AI will demand a staggering buildout of chips, data centers, and energy—isn’t wrong. That buildout is underway. But conviction without cash management is just gambling. His private holdings, including a stake in Anthropic, remain untouched. The public market lesson, however, was brutal: you can be right about the destination and still get wiped out on leverage before the train leaves the station.

This isn’t just a story about one fund imploding. It’s a flashing warning sign for the wave of AI investment vehicles that have launched with more enthusiasm than trading experience. When the cost of financing rises and investors start asking when the billions in infrastructure spending will actually pay off, heavily levered bets become kindling. Aschenbrenner’s saga suggests the AI trade is entering a phase where timing and risk management matter as much as having the right big-picture call.