The Cautionary Startup Tale of Leopold Aschenbrenner: Inexperience Means Inexperience

Leopold Aschenbrenner's Situational Awareness Fund lost $35 billion of its $45 billion assets under management in about a month after running 4x leveraged bets on AI infrastructure stocks including SK Hynix and CoreWeave, triggered by a July 2026 selloff that saw 1.2 million margin calls in Korea alone. Despite having no prior trading or risk-management experience, Aschenbrenner had attracted billions from investors based on his AI essay and AGI-timeline credibility, and Silicon Valley is reportedly still willing to back him.

TL;DR: Leopold Aschenbrenner turned a famous AI essay into a $45 billion hedge fund, then lost $35 billion of it in about a month by running 4x borrowed money on a handful of AI infrastructure bets. Silicon Valley is now lining up to give him more money anyway. Fourteen months ago, Leopold Aschenbrenner was the closest thing Silicon Valley had to a prophet. Twenty-five months later, he was on the phone with prime brokers explaining why they needed to seize his portfolio. Both of those things are true, and neither one cancels the other out. That's the part people keep skipping past. Aschenbrenner graduated valedictorian from Columbia at 19, joined OpenAI's Superalignment team, and was fired in April 2024 over what he says was a mishandled security memo and OpenAI says was a leaked internal document. Two months later he published "Situational Awareness: The Decade Ahead," a 165-page essay arguing AI systems would be doing their own AI research by 2027 and that superintelligence was closer than almost anyone in power wanted to admit. It became required reading from Sand Hill Road to the Pentagon. He also spent part of 2022 at the FTX Future Fund, handing out grant money, and resigned on November 11, 2022, the exact day FTX filed for bankruptcy. He has never run a trading desk, never managed outside capital through a down cycle, and had no professional risk-management background of any kind before he started doing it with other people's billions. None of that stopped investors from handing him a hedge fund. Situational Awareness Fund launched off the strength of the essay and Aschenbrenner's AGI-timeline credibility, not a track record, because he didn't have one to speak of. It didn't matter. By the first half of 2026 the fund was up 439%. Since inception, gains ran as high as 1,551%. Assets under management hit $45 billion. He was, by the only metric Silicon Valley actually tracks in real time, winning. The borrowed money was the whole strategy Here's what a lot of the coverage buries: the returns weren't from being right about AI. They were from being right about AI while borrowed four times over. Situational Awareness ran concentrated positions in AI infrastructure names, including SK Hynix and CoreWeave, using leverage reported as high as 4x. That's not a hedge fund technique. It's a bet dressed up as one. Borrowed money doesn't change whether your thesis is correct. It changes how violently you get punished if the timing is off by even a few weeks. Dario Amodei wants a harder China chip ban that would only speed up the loss he is trying to prevent https://startupfortune.com/dario-amodei-harder-china-chip-ban-would-speed-up-the-loss-anthropic-fears/ Anthropic's Dario Amodei is urging Washington to tighten chip export controls, crack down on distillation, and mandate safety tests. It reads less like confidence than like a company looking over its shoulder, and the ban he wants would only push China to build its own chips faster. The distillation complaint also lands awkwardly a week after... - dario amodei china chip ban https://startupfortune.com/dario-amodei-harder-china-chip-ban-would-speed-up-the-loss-anthropic-fears/ - anthropic export controls safety policy https://startupfortune.com/dario-amodei-harder-china-chip-ban-would-speed-up-the-loss-anthropic-fears/ The timing was off. In July 2026, SK Hynix's U.S. IPO coincided with a broader AI infrastructure selloff, and semiconductor stocks came apart at the same moment a wave of retail margin trading was already unwinding across markets. In Korea alone, one day in July saw 1.2 million margin calls and roughly 320,000 to 360,000 forced liquidations. U.S. margin debt had grown 54% year over year going into it. Situational Awareness Fund was sitting in the middle of exactly the kind of crowded, borrowed-money trade that turns an orderly pullback into a spiral. Falling prices shrink your equity cushion. Brokers demand more collateral. You sell more to raise it, the selling pushes prices down further, and the cushion shrinks again. The fund lost 67% of its value in a single month. $35 billion gone. Assets fell from $45 billion to roughly $10 billion. On July 30, prime brokers forced a full liquidation of the fund's public stock positions, sold at a discount to Ken Griffin's Citadel, just to stop the fund from being wiped out completely. Aschenbrenner's own summary of it: "These were very expensive scars, but I am dedicated to ensuring they will be invaluable lessons for our organization and for myself as we move forward." That's a fine sentence for a Tuesday. It is also $35 billion of other people's money, gone in the time it takes most funds to schedule a quarterly investor call. What actually survived To be precise about what happened, because precision is the thing this story keeps losing: the fund's private stakes weren't part of the forced sale. That includes positions in Anthropic, Fluidstack, and MatX. They're illiquid, marked differently, and not subject to the same margin mechanics as the public book. So the $35 billion figure describes the public portfolio's collapse. It is not a claim that every dollar Aschenbrenner ever raised evaporated. Not even close. It's still one of the fastest multibillion-dollar unwinds in the AI trade's short history. And it happened to a fund run by someone whose actual qualification for the job was writing a very good essay. The part that should bother you more than the loss Two weeks after the fire sale, Redpoint Ventures' Logan Bartlett was on record explaining why Silicon Valley wants to give Aschenbrenner more capital, not less: "There is such a thing as the archetype of a hero," and "there is a movement rallying around Leopold Aschenbrenner." Aschenbrenner, for his part, wasn't sitting still licking the wound. He deployed $500 million into private companies after the collapse. $400 million of it was committed on August 4, days after Citadel finished picking through his public book. Wall Street's entire culture is built around protecting principal, because principal belongs to someone else. Silicon Valley's culture treats a correct-but-catastrophically-timed thesis as a résumé line. Those are incompatible ways of thinking about other people's money. Aschenbrenner's fund is what happens when the two mix. It's what happens when the second one gets access to borrowed money the first one would never have signed off on. Being right about where AI is headed and being competent to run billions of dollars against that view are two different skills. One of them Aschenbrenner has clearly demonstrated. The events of July 2026 are the clearest evidence yet that he hadn't demonstrated the other one, and that the people funding him didn't especially care. That's the actual lesson here, and it isn't really about Aschenbrenner. It's about an investing culture that keeps mistaking conviction for competence, and keeps getting surprised when the difference costs $35 billion.