Leopold Aschenbrenner’s Situational Awareness AI hedge fund lost around 67% of its value in July. This was the result of margin calls leading the fund to sell its public portfolio. The collapse of such a fund poses a question for equity investors in regard to the second half of 2026. Despite accurately predicting the demand for AI, the fund almost experienced a catastrophe, making this event significant for reasons beyond a single hedge fund manager.
Whether or not the technology sector will remain as volatile going forward is one of the central questions now facing markets heading into the end of the year. The same names that have propelled stock indexes upwards can fall quickly as soon as leverage and liquidity meet a downturn, as was demonstrated in July.
Why a 439% winner lost two-thirds in a month
Situational Awareness was created by a former researcher from OpenAI and bagged returns of 439% in H1 2026 through about 4x leverage on concentrated bets on infrastructure and chips in the AI space, as per RCK Analytics’ paper on its collapse. However, that leverage had its downsides. As its positions fell between 35% and 47% in July, three prime brokers, including Goldman and JPMorgan, called margin, which the fund was unable to honor, leading it to sell its entire public equities book to Citadel.
Reports verified the 67% plunge in July and the sale of the fund’s majority public investments. According to Aschenbrenner, the firm explained that the fund is positioned “to fight another day” and presented the drop in terms of risk instead of thesis.
Breaking: Leopold’s full letter sent to his LPs last night
Leopold Aschenbrenner’s fund fell 67% in July but remains up 80% YTD and he announced he’ll keep investing in public equities https://t.co/n3Swjmp8Z6 pic.twitter.com/jswe9ROkTo
— Leopold Stock Tracker (@LeopoldTracker_) July 31, 2026
“We embrace volatility. But it should never jeopardize the fund.” And the company didn’t exit the industry, investing $400 million in Source Foundry, a chip-making start-up with a valuation of almost $5 billion that specializes in lithography, which is the bottleneck in chip manufacturing and the area that is under the monopoly of ASML.
The demand was real, but the financing was not
It was not the bet that was erroneous. Filings say that close to half of the fund’s US equity portfolio had actually come from investments in companies like SanDisk and Micron by the end of June, just when the demand for memory chips had been trending up. A study by Counterpoint Research indicated that enterprise solid-state disks accounted for 48 percent of the world’s total NAND sales in Q2 of 2026, during the shift of AI workloads from the training phase to the inference phase. Moreover, TrendForce predicted that NAND prices would go up by 10 to 15 percent compared to the previous quarter and that prices for DRAM chips would increase by 13 to 18 percent quarterly.
Micron supported its position in the market with statistics, reporting an unprecedented revenue of $41.46 billion for its fiscal third quarter, compared to $9.30 billion for the previous year. This, however, did not help the hedge fund. The Philadelphia Semiconductor Index suffered an almost 30% fall compared to the peak recorded in June, and an aggressive position did not allow to wait long for the recovery indicated in the demand figures.
Central banks flag the same leverage build-up
The weakness in question had already been identified by regulatory authorities before the case of Situational Awareness. The Financial Stability Report of the Bank of England, published in July 2026, identified the risk of a “substantial increase in the use of leverage in equity markets,” with a prominent emphasis on inflated estimations of a small number of companies related to AI. The Bank for International Settlements used the examples of the collapse of the Archegos investment fund in 2021 and the stress of liability-driven investments in the UK in 2022 to insist that non-bank funds should take measures to stay away from events such as the one in the case of Aschenbrenner. The greatest problems were identified by experts as being leverage and concentration risks.
Essentially, the pattern of behavior is unchanged: the right ideas, but the wrong form and way of realizing them without any liquidity buffer in place.
Whether the AI trade stays institutional
The unanswered question that remains to be explored in 2026 is how the gap is going to be filled. JPMorgan suggests that the technology trade may still be dealing with the leverage that has contributed to its rise in the past, and that leveraged ETFs, options and margin accounts might still be settling down. Such programming is expected to further modify how leveraged and unleveraged demand for technology stocks is balanced while the risk is being reduced by investors on their part. According to PitchBook, real assets leveraged fundraising related to infrastructure, energy and data centers reached the record level of $206.6 billion during 2025, with allocators choosing the structures with long duration and contracts over leveraged directional investments.
The investment funding in AI is not getting smaller. Goldman Sachs estimates spending of about $7.6 trillion on computing, data centers, and energy from 2026 to 2031. What has changed in July is how willing the market will be to bear that risk and whether the next downturn is greeted by patient capital or highly leveraged holders ready to sell off.
