A correct bet on AI infrastructure was not enough to survive a margin call — leverage, not conviction, decided the fund's fate.
A correct bet on AI infrastructure was not enough to survive a margin call — leverage, not conviction, decided the fund's fate.

A correct bet on AI infrastructure was not enough to survive a margin call — leverage, not conviction, decided the fund's fate.
Situational Awareness, the hedge fund run by former OpenAI researcher Leopold Aschenbrenner, lost 67% in July after nearly 4x leverage collided with an AI stock rout, forcing a $16 billion sale to Citadel.
"We let you down this month," Aschenbrenner wrote in a letter to investors reviewed by The Wall Street Journal, taking full responsibility for the drawdown and vowing to "learn the necessary lessons" from the experience.
The fund, launched in late 2024 with less than $1 billion, swelled to more than $20 billion in under two years after returning roughly 439% in the first half of 2026. Its top first-quarter holdings — Nebius Group, Sandisk, Micron and CoreWeave — each fell more than 35% in July, while shorted software names like Adobe rebounded, squeezing both sides of the book. The fund used money borrowed from prime brokers including Goldman Sachs, JPMorgan, Bank of America and Citigroup to enlarge its positions.
The episode shows how a correct long-term thesis can collapse under short-term liquidity pressure. The fund retained roughly $10 billion in assets, including about $5 billion in Anthropic equity, and Aschenbrenner has pledged to remove all leverage. Whether the AI infrastructure trade resumes depends on whether other levered funds face similar margin pressure.
The collapse began with a market-wide reassessment of AI capital expenditure payback periods. As investors questioned the returns on debt-funded data-center construction, AI hardware and infrastructure stocks saw concentrated selling in July. Asian names such as SK Hynix weakened sharply, while the software stocks the fund shorted rebounded, creating a pincer move against both long and short positions.
More critical was the leverage. The fund borrowed from Wall Street banks to enlarge positions, amplifying gains while AI stocks rose but reducing the collateral supporting those loans when they fell. As losses mounted, margin calls came in rapid succession, forcing the fund to either raise capital quickly or sell assets under duress.
Around July 30, the fund agreed within 24 hours to sell the bulk of its roughly $16 billion public stock portfolio to Citadel at a discount of more than 10 percent. Prime brokers including Goldman Sachs, JPMorgan, Bank of America and Citigroup helped arrange the transaction, which the Financial Times described as one of the largest rushed stock transactions in Wall Street history. The fund also scrapped a planned $3.5 billion sale of Anthropic equity to a consortium led by Greenoaks and Sequoia Capital, according to people familiar with the matter.
The core lesson is not that Aschenbrenner's long-term thesis is wrong. Microsoft Azure's continued growth, persistent supply-demand tightness for Nvidia's high-end chips and AMD's repeated upward revisions of AI server market forecasts all point to a macro demand curve for AI infrastructure that is still sloping upward.
The real risk was never about direction but about the alignment of leverage, position concentration and entry timing. History offers precedents: Long-Term Capital Management, staffed by Nobel laureates, imploded in 1998 during the Russian debt crisis under extreme leverage, while Peter Thiel's Clarium Capital made a correct but premature bet on oil prices that proved unsustainable.
Business Insider summarized the failure precisely: "Even if you can accurately identify the biggest fundamental shift of a generation... once you add sufficient leverage and conviction, the market no longer judges whether your ideas are smart — it judges whether you can survive when liquidity dries up."
The fund's rapid expansion from a small operation to tens of billions of dollars in assets massively amplified operational difficulty. Highly concentrated public positions stacked on high leverage left almost no room to maneuver when faced with the dual shock of sudden Asian market volatility and a global selloff in AI stocks. Large-scale redemption pressure combined with instantaneous liquidity evaporation forced the fund to liquidate passively at the worst possible moment.
For the broader market, the AI narrative remains compelling, but leverage is a double-edged sword. Global hedge funds suffered their largest monthly drawdown on record in July, while Asia-focused fundamental long-short funds lost an average of 18.6 percent through July 28, according to Reuters. If other levered funds face similar margin pressure, the unwinding could inject further volatility into the Nasdaq and S&P 500 in coming weeks.
This article is for informational purposes only and does not constitute investment advice.