AI hedge fund Situational Awareness may have sold its public portfolio, but it still has its Anthropic shares

Forced Sale to Citadel Marks Reversal for High-Flying AI Fund

Situational Awareness, the hedge fund established by former OpenAI researcher Leopold Aschenbrenner, has sold the bulk of its public stock portfolio to Ken Griffin’s Citadel following a period of catastrophic losses. The sale marks a dramatic reversal for the fund, which had previously posted a staggering 439% return through June by making aggressive, leveraged bets on the AI infrastructure sector. This turn of events for the 25-year-old founder underscores the intense volatility in public markets for AI-related hardware and energy companies.

Key Factors in the Fund's Decline

• Leveraged Positions: The fund's strategy of using borrowed money to amplify its bets magnified steep losses when the market turned.
• Sector Downturn: Public AI infrastructure stocks, including key holdings like SK Hynix and Bloom Energy, plummeted by more than 30% amid investor concern over high capital expenditures and slow revenue growth.
• Failed Capital Call: A July 24 letter inviting clients to commit fresh capital failed to attract the necessary funds to stabilize the portfolio.
• Asset Collapse: Assets under management crashed from a peak of $45 billion to an estimated $10 billion after the sale to Citadel.

Despite liquidating its public holdings, Situational Awareness has retained its portfolio of private investments. The most significant of these is a stake in AI developer Anthropic, currently valued at $5 billion. With an Anthropic IPO expected as soon as October, a successful public offering could generate a windfall substantial enough to offset a significant portion of the fund's public-market losses. This situation creates a stark contrast between the perceived stability of premier private AI labs and the sentiment-driven turbulence of the public stocks meant to support them. Citadel’s purchase also signals that while Aschenbrenner's timing may have been off, his underlying thesis about the AI infrastructure build-out still has believers with deep pockets and longer time horizons.

The forced deleveraging of Situational Awareness illustrates the acute risks of betting on the picks-and-shovels of the AI boom in public markets, while its retained Anthropic stake shows where durable, long-term value is currently perceived to be: in the foundational model providers themselves.