AI Hedge Fund Situational Awareness Dumps Stocks to Citadel
02 Aug 2026
A meteoric rise, then a steep fall
Situational Awareness, the AI-focused hedge fund founded by former OpenAI researcher Leopold Aschenbrenner, has sold most of its public stock portfolio to Ken Griffin's Citadel following a sharp downturn. The fund's assets under management fell to roughly $10 billion, down from around $20 billion in recent months and from a peak of $45 billion at the start of July.
The drop caps a dramatic run. Situational Awareness launched in 2024 with several hundred million dollars from backers including Jane Street, Stripe co-founders Patrick and John Collison, and Meta AI executives Daniel Gross and Nat Friedman. Through June, the fund returned 439% for the year — a performance that pushed its AUM to as much as $45 billion despite a lean team of just eight staff, four of whom were investment professionals.
From superalignment to a $45 billion bet
Aschenbrenner's path to running one of the highest-profile AI-themed funds is unconventional. He enrolled at Columbia University at 15 and graduated as valedictorian at 19 in 2021, then worked at the FTX Future Fund before joining OpenAI's superalignment team in 2023. About a year later, he was dismissed from OpenAI for leaking internal information. Two months after his departure, his "Situational Awareness" essays were published, and the hedge fund bearing that name launched soon after. Carl Shulman, previously of Peter Thiel's Clarium Capital, served as the fund's director of research.
What triggered the losses
The report ties the fund's decline to a broader slide in AI-adjacent public equities over the past month, with declines exceeding 30% in some names. Sources differ on the specifics: one account lists Nebius, Sandisk, Micron and CoreWeave down more than 35%, while another cites SK Hynix, Sandisk, Bloom Energy and Nebius down more than 30%. The fund's losses were reportedly amplified by leverage — the common hedge fund practice of trading with borrowed money — which can magnify both gains and losses.
The situation has drawn comparisons to Archegos Capital Management, the family office that lost $8 billion in ten days in 2021 due to concentrated, leveraged bets. That comparison underscores the risk profile associated with concentrated, leveraged exposure to a single theme — in this case, AI.
On July 24, before the worst of the decline, Aschenbrenner sent investors a letter describing the selloff as one of the best buying opportunities since early last year. Aschenbrenner has also said, "Obviously, not blowing up is task number one and two," describing the firm as "kind of like a brain trust on AI."
What's left standing
Despite offloading most public holdings, Situational Awareness retained its private company investments. That includes a $5 billion stake in Anthropic, last valued at $965 billion in a May Series H round and expected to go public as soon as October. The fund also holds stakes in chipmaker MatX and AI data center startup Fluidstack, which was reportedly in talks in April to raise a new round at an $18 billion valuation.
Sources differ
The report notes a couple of discrepancies worth flagging: one source states Aschenbrenner is 24 years old, while another puts him at 25 and identifies him as German-born. Sources also differ on exactly which stocks drove the losses and by how much, as detailed above.
What's not yet known
Several details remain unclear from available reporting: what specific information led to Aschenbrenner's OpenAI dismissal, the precise trigger for the recent losses, whether investors are pulling money from the fund or how they responded to the July 24 letter, the fund's strategy going forward, and how much leverage contributed to the losses.
Why founders should care
For early-stage founders, this episode is a useful data point rather than a direct warning. It's likely that AI-themed hedge funds using leverage will continue to see amplified volatility in both directions as public AI-adjacent stocks swing — a dynamic founders raising from or partnering with such funds should weigh when assessing investor stability. At the same time, the fund's decision to retain private stakes in Anthropic, MatX, and Fluidstack — rather than liquidate them — suggests institutional conviction in private AI infrastructure and model companies may remain intact even amid public market turbulence. Founders building in AI infrastructure or model layers may find this reassuring, though the Archegos comparison is a reminder that concentrated, leveraged bets on any single theme carry real tail risk that could eventually affect capital availability across the sector.