Leopold Aschenbrenner was born in Germany to physician parents and attended the John F. Kennedy School in Berlin. He entered Columbia University at age fifteen and graduated in 2021 as class valedictorian with degrees in economics, mathematics and statistics. Early in his career he worked at the FTX Future Fund; the parent exchange’s fraud revelations and subsequent bankruptcy occurred at the end of 2022.
Fired from OpenAI and turning into a public critic
In 2023 Aschenbrenner joined the OpenAI Superalignment team, working under Ilya Sutskever and Jan Leike on aligning models with human goals. In April 2024 OpenAI dismissed him; the company said he improperly shared confidential internal information, while Aschenbrenner said he had sent an ideation document to three external researchers and had previously raised internal concerns about the company’s security practices. After leaving, he became an outspoken critic of safety and geopolitical risks tied to AI development.
The essay and its main claims
In June 2024 he published a 165‑page essay titled Situational Awareness: The Decade Ahead. The paper argued that general artificial intelligence (AGI) could become available around 2027 and that a rapid self‑improvement phase might lead to superintelligence by 2030. Aschenbrenner emphasized combined improvements in compute, algorithms and learning methods, advocated military‑style physical and cyber protections for the most advanced model weights, and proposed a tightly guarded, allied national program for top‑tier AI development. These claims were controversial, but the essay quickly became a widely discussed AI manifesto in Silicon Valley and on Wall Street.
From essay to hedge fund: strategy and early results
A few months after publishing the essay he launched the Situational Awareness hedge fund with about $225 million of seed capital from prominent Silicon Valley backers. He hired Carl Shulman as research director and assembled a team aimed at buying the ‘‘shovels and spades’’ of AI infrastructure: power generation, memory manufacturers, compute capacity, network optics, chip fabs and data‑center real estate. On the short side the fund bet against software companies it judged vulnerable if AI commoditized or supplanted traditional software offerings.
Early performance was notable: the fund reportedly built roughly $5.52 billion of public equity positions in one year and recorded an alleged 47 percent net return in the first half of 2025. By early 2026 the fund’s disclosed U.S. equity holdings reached about $13.7 billion. At its July 2026 peak, total assets — including public and private positions, with a large stake in Anthropic among the private holdings — were reported at around $45 billion.
Major disclosed long positions included SK Hynix, CoreWeave, Nebius, Sandisk and Micron; among the shorts was Adobe.
The collapse and forced liquidation
In July 2026 sentiment toward AI‑infrastructure names shifted sharply. Investors questioned the sustainability of extraordinary profitability at memory manufacturers and compute providers. Corporate developments also weighed on the sector: Meta signaled it might sell or lease excess compute capacity, and Apple sought U.S. approvals related to use of Chinese memory chips.
As a result many of the fund’s core long holdings — SK Hynix, Nebius, Sandisk, Micron and CoreWeave — fell 35–40 percent from their peaks in a matter of weeks. Simultaneously, the fund’s short positions moved against it because many market participants had paired semiconductor longs with software shorts, amplifying losses. The fund’s heavy leverage exacerbated the situation: margin calls arrived, brokers issued collateral demands, and temporary arrangements with brokers were insufficient to avoid escalation. Aschenbrenner asked existing investors for additional capital.
On July 30, 2026, the fund was forced to sell nearly its entire public equity portfolio in an emergency, 24‑hour transaction. Reports say Ken Griffin’s Citadel purchased much of the inventory at a substantial discount to market prices. The fund’s assets fell from roughly $45 billion at the start of the month to about $10 billion by month‑end. A spokesperson clarified that the Anthropic stake was not offered for sale during the liquidation.
Ironically, many of the holdings began to rally after the liquidation — gains that flowed to the buyers’ books rather than back to the fund. Nonetheless, reports indicate the fund’s investors still realized about an 80 percent year‑to‑date return as of that period.
Takeaways
Aschenbrenner’s story illustrates a familiar investment lesson in a new guise: conviction about a long‑term technological outcome does not substitute for prudent risk management. Position size, financing structure and timing matter as much as the correctness of the underlying thesis. Whether the world will indeed spend unprecedented sums on AI infrastructure, or whether AGI arrives on the timeline Aschenbrenner forecast, remains unresolved. July 2026 demonstrated, however, that short‑term market dynamics and leverage can overwhelm even the most persuasive long‑run bets.
Somlai‑Kiss Máté equity analyst, Concorde
This piece was published as a guest contribution; opinions are those of the author and do not necessarily reflect the editorial stance of Forbes.
