# Leopold Aschenbrenner's AI Hedge Fund Collapsed From $45 Billion to $10 Billion > Source: > Published: 2026-08-03 03:33:30+00:00 *Leopold Aschenbrenner wrote the essay that convinced Silicon Valley to bet everything on AI infrastructure. This week his own fund got margin-called out of that exact trade, shrinking from $45 billion to roughly $10 billion in six trading days.* Situational Awareness LP, the hedge fund run by the 25-year-old former OpenAI researcher, was forced to sell almost its entire public stock portfolio to Ken Griffin's Citadel this week after margin calls from prime brokers including Goldman Sachs, JPMorgan and Bank of America, according to CNBC. The fund's assets fell from a July peak of $45 billion to about $10 billion by Thursday, July 30. Six trading days undid a run that had made Aschenbrenner one of the most talked-about investors in the world. The irony writes itself. In 2024, Aschenbrenner published "Situational Awareness: The Decade Ahead," a widely circulated essay arguing that AI progress would demand an enormous buildout of chips and power, with the data centers to house it all bolted on. That thesis became something close to scripture in parts of Silicon Valley. He turned it into a fund, and the fund turned $225 million into as much as $45 billion in under two years, according to CNBC, printing a 439% net return through June 2026. Then the same trade that made him famous is the one that broke him. Situational Awareness ran leverage of roughly four times its capital, per CNBC and Yahoo Finance reporting. It held concentrated long positions in AI infrastructure names, SK Hynix, Micron, Sandisk, CoreWeave and Nebius among them, while shorting software stocks like Adobe as a hedge. In July, that hedge failed in both directions at once. The AI infrastructure longs fell more than 35% as semiconductor and data-center stocks reversed sharply. The software shorts rallied instead of falling. At 4x leverage, that combination doesn't just cost you money. It triggers margin calls. Prime brokers came calling. Rather than raise fresh capital, the fund chose to sell. Citadel picked up the bulk of the public equity book, reportedly worth around $16 billion, at a distressed price, according to Yahoo Finance and Seeking Alpha. The trade cleared a major forced seller out of the market, and the affected stocks bounced sharply on July 30, the same day the sale closed. Aschenbrenner had no professional investing background before this. He was an OpenAI researcher who left the company and turned a manifesto into capital. That's the part of this story that should unsettle anyone who treated conviction as a substitute for risk management. A 439% return by June looked like proof the thesis was right. It wasn't proof of anything except that leverage amplifies whatever happens next, good or bad. ## What's left of the fund Not everything is gone. Situational Awareness still holds a private stake in Anthropic worth roughly $5 billion, acquired when the AI lab was valued at around $60 billion in February 2025, according to TechCrunch. That position wasn't marked to market the way the public portfolio was, and it wasn't subject to the same margin calls. The firm will keep operating, just smaller, as a private investment vehicle built around that Anthropic holding rather than a leveraged public-markets fund. So the man who wrote the manifesto still owns a piece of one of the labs building the future he predicted. He just no longer owns the leveraged bet on the infrastructure underneath it. This is the AI trade's first real blowup, not a rumor of one. Semiconductor and data-center stocks have carried the market's AI narrative for two years without a genuine stress test. Aschenbrenner's fund just supplied one, in public, with real numbers attached. Whether SK Hynix and CoreWeave, along with the rest of that basket, recover from here says more about the AI infrastructure bull case than any essay could. The market already answered the question about leverage. What it hasn't answered yet is whether the underlying thesis, the one Aschenbrenner wrote down before he ever traded on it, still holds. **Also read:** [Alibaba Says Its New Qwen3.8-Max Model Trails Only Anthropic's Claude](https://startupfortune.com/alibaba-says-its-new-qwen38-max-model-trails-only-anthropics-claude/) • [ASML Is Pushing Price Hikes on TSMC the Same Week Chip Stocks Lost $1 Trillion](https://startupfortune.com/asml-is-pushing-price-hikes-on-tsmc-the-same-week-chip-stocks-lost-1-trillion/) • [A Deadly Chile Storm Just Shut Down a Fifth of the World's Copper Supply](https://startupfortune.com/a-deadly-chile-storm-just-shut-down-a-fifth-of-the-worlds-copper-supply/)