Karla Barker 
AI-focused hedge fund Situational Awareness lost roughly 67% of its value in July. The fund, created by former OpenAI researcher Leopold Aschenbrenner, was forced to sell its public portfolio after margin calls it could not meet. Three prime brokers, including Goldman Sachs and JPMorgan, called in loans as positions fell sharply. Citadel bought the public equities book.
The fund had made 439% in the first half of 2026. It used around 4x leverage on concentrated bets in AI infrastructure and chipmakers. That leverage became the problem. Positions dropped between 35% and 47% in July. The margin calls came quickly. The fund had no liquidity buffer strong enough to wait out the downturn. Aschenbrenner told investors the fund was positioned “to fight another day” and remains up 80% year to date.
The right bets, the wrong structure
The strange part is the underlying bets were not obviously wrong. By the end of June, close to half of the fund’s US equity portfolio sat in companies like SanDisk and Micron. Memory chip demand was rising. Counterpoint Research found enterprise solid-state drives made up 48% of global NAND sales in Q2 2026. TrendForce expected NAND prices to rise 10-15% quarter over quarter, and DRAM prices 13-18%. Micron posted record quarterly revenue of $41.46 billion, up from $9.30 billion the year before.
None of that saved the fund. The Philadelphia Semiconductor Index fell almost 30% from its June peak. A leveraged position does not allow much patience, even when demand data points to recovery.
Regulators saw this coming
Central banks had warned about leverage in equity markets before this collapse. The Bank of England’s July 2026 Financial Stability Report flagged a “substantial increase in the use of leverage in equity markets,” especially around a small group of AI-related companies. The Bank for International Settlements pointed to the Archegos collapse in 2021 and the UK liability-driven investment stress in 2022. The lesson is familiar: non-bank funds with concentrated leverage can break quickly.
The pattern has not changed much. Good ideas, yes. But fragile execution, no liquidity cushion, and too much reliance on borrowed money.
What comes next
The open question for the second half of 2026 is how the market clears this risk. JPMorgan thinks the tech trade may still be working through excess leverage. Options, margin accounts, and leveraged ETFs could remain under pressure. That could shift the balance between leveraged and unleveraged demand for tech stocks.
Direct AI investment is not slowing. Goldman Sachs estimates $7.6 trillion in spending on computing, data centers, and energy from 2026 to 2031. Real assets fundraising tied to infrastructure, energy, and data centers reached a record $206.6 billion in 2025, according to PitchBook. Investors prefer longer-duration structures with contracts and yield.
July changed the way the market thinks about bearing risk. The next downturn will test whether patient capital replaces leveraged holders who have to sell. This story is about more than one hedge fund manager. It is about the structure of the AI trade.