Situational Awareness hedge fund loses $35B in months on AI bets

Leopold Aschenbrenner, a 24-year-old former OpenAI employee, founded Situational Awareness as a hedge fund focused on artificial intelligence bets. The fund's strategy rested on a series of essays Aschenbrenner published arguing that machines capable of true thinking and reasoning would emerge by 2025–26, outpacing college graduates, and that superintelligence in the genuine sense would arrive by decade's end. He predicted that national security forces unseen in half a century would be unleashed and that an all-out race (or war) with China might follow. The entire thesis was that artificial general intelligence was coming, and the path to wealth lay in dumping capital into AI stocks and riding the wave.

Aschenbrenner's essays impressed influential Silicon Valley figures. Tyler Cowen, a libertarian economist with reach in certain circles, had called him an economics prodigy at age 17 and his Emergent Ventures fund gave him a grant. Aschenbrenner had graduated from Columbia University in 2021 as valedictorian at age 19, worked briefly at the FTX Future Fund (before FTX collapsed), and then joined OpenAI's superalignment team. He was later fired from OpenAI for leaking internal information. Two months after his departure, he published the Situational Awareness essays. Despite having no prior money management experience, Aschenbrenner attracted prestigious backers: Patrick and John Collison (Stripe cofounders), Daniel Gross and Nat Freedman (Meta AI leaders), Carl Shulman (formerly of Peter Thiel's Clarium Capital) as director of research, and eventually Jane Street, a Wall Street firm that rarely allocates capital to outside money managers. Ivanka Trump called his essays "an excellent and important read."

The fund operated with extreme leverage. Hedge funds typically borrow money to maximize their bets, and Aschenbrenner followed this playbook aggressively. By early 2024, he told podcast host Dwarkesh Patel that "obviously, not blowing up is task number one and two," and that timing and sequence of bets mattered. At one point, the fund claimed to be up 439 percent. But leverage magnifies losses as well as gains. The fund was worth $45 billion at the start of July. By the time AI stocks began declining—driven by market nervousness about AI risk—margin calls from lenders demanded more collateral. Situational Awareness held calls with investors and lenders to raise capital. Its largest holdings at the end of the first quarter were Nebius Group, Sandisk, Micron, and CoreWeave; all fell more than 35 percent in the month of the crisis. Facing a margin squeeze, the fund first tried to sell parts of its portfolio to existing investors, then sold what the Financial Times described as a large portion of its $16 billion public equity portfolio to Ken Griffin's Citadel. After the sale, the fund's value had crashed to $10 billion—a loss of $35 billion. According to CNBC, the fund was also negotiating to sell its $5 billion stake in Anthropic, though it was unclear whether that deal had closed. If the reported figures hold, Situational Awareness's loss would exceed the previous record by Archegos Capital Management, which lost $8 billion in ten days in 2021. The irony embedded in the fund's name—chosen after Aschenbrenner's essays on gaining "situational awareness" about AI's imminent arrival—was not lost on observers.