Situational Awareness was built to profit from an AI buildout that its founder believed was inevitable. Instead, a brutal turn in the infrastructure trade forced the young hedge fund into a rapid public-market retreat.

Leopold Aschenbrenner, a former OpenAI researcher, launched the fund in 2024 around a thesis that surging demand for chips, computing power, memory and energy would underpin the race toward more capable AI. The wager initially looked spectacular: the fund returned 439% through June and, at its peak, was reported to oversee as much as $45 billion. ¹

Then AI-linked equities slid. Holdings including Nebius, Sandisk, Micron and CoreWeave were hit hard, while the fund’s use of borrowed money magnified the downturn. A sharply skeptical account cast the episode as a warning about turning sweeping AI predictions into concentrated, leveraged trades, noting Aschenbrenner had said that “not blowing up is task number one and two.” ²

By late July, Aschenbrenner was still urging investors to see the selloff as a buying opportunity and seeking fresh commitments. But the capital did not arrive at the scale needed. The Financial Times reported fundraising efforts amid heavy losses; within a day, Axios reported that Citadel had bought the fund’s entire public-equities portfolio. ³

Reports differ on whether Citadel acquired all or merely most of the listed holdings, but they agree on the core reversal: Ken Griffin’s firm stepped in as Situational Awareness unwound. The fund’s assets were subsequently reported at roughly $10 billion, far below recent estimates.

The sale did not erase the fund’s private-market exposure. Most importantly, Situational Awareness retains an Anthropic stake Bloomberg valued at about $5 billion, leaving Aschenbrenner with a potentially significant claim on the AI boom even after his public-stock bet buckled. ¹