Original | Odaily Planet Daily (@OdailyChina)
Author | Qin Xiaofeng (@QinXiaofeng 888)
Key Takeaway: Situational Awareness, the hedge fund run by former OpenAI researcher Leopold Aschenbrenner, suffered a roughly 67% single-month loss in July after high leverage collided with an AI stock pullback, forcing it to sell roughly $16 billion in public positions to Citadel — underscoring how even a correct macro narrative can prove fragile under leverage and liquidity shocks.
Key Facts:
a. The fund ballooned to roughly $20 billion in assets under management in under two years, posting around 439% returns in the first half of 2026, with heavy concentrations in AI infrastructure names like CoreWeave and Bloom Energy.
b. July's concentrated sell-off in AI hardware stocks — with top Q1 holdings like Nebius, Micron, and CoreWeave falling over 35% — combined with a rebound in shorted software names, delivering a double blow on both sides of the book.
c. The fund used nearly 4x leverage to amplify gains, which sharply magnified losses on the way down. Margin calls forced passive liquidations, ultimately leading to roughly $16 billion of the portfolio being sold to Citadel.
d. Remaining assets shrank to roughly $10 billion, including about $5 billion in private Anthropic equity, with plans to continue operating as a private investment firm.
e. Analysts note that the long-term demand thesis for AI infrastructure remains intact, but leverage, concentration, and poor timing alignment were the primary causes of the blowup. Historical cases like LTCM reinforce that "survivability, not being right, is the real moat."
In the final week of July, Wall Street witnessed the meteoric collapse of a hedge fund once hailed as "legendary."
Situational Awareness, helmed by Leopold Aschenbrenner — a German-born investor in his mid-20s and former OpenAI researcher — suffered catastrophic losses as AI-related stocks sharply corrected on top of heavy leverage. The fund was forced to massively unwind its public-market positions, selling the bulk of its stock portfolio to Ken Griffin's Citadel.
The episode quickly became the focus of global financial media, dragging the young manager — once dubbed the "AI stock whisperer" in Chinese-language circles — from myth down to a harsh reality check.
The Rise of a Prodigy
Leopold Aschenbrenner's background is itself the stuff of legend. Born in Germany in 2001, he entered Columbia University at age 15, majored in economics and math-statistics, and graduated at the top of his class.
After graduating, he briefly worked at FTX's Future Fund before joining OpenAI's Superalignment team in 2023, focusing on safety and alignment for superintelligence. In April 2024, he was terminated over alleged information leaks (which he denies), widely believed to be tied to his repeated security memos to the board and criticism of company practices.
Just two months after his dismissal, Leopold self-funded a roughly 165-page paper titled Situational Awareness: The Decade Ahead. In it, he extrapolated deep learning trends to predict that artificial general intelligence (AGI) could arrive around 2027, arguing that the true bottleneck wasn't the algorithms themselves but the explosive demand for compute clusters, electricity, data center space, and energy infrastructure.
The essay spread rapidly through Silicon Valley and investing circles, earning Leopold a reputation as an "AI prophet." In September 2024, he founded the hedge fund Situational Awareness LP — named directly after the paper — focusing on publicly listed companies tied to AI infrastructure: chips, storage, cloud computing, power equipment, and data center operators, while shorting legacy software and outsourced services that AI might disrupt. The fund attracted marquee backers including Stripe founders Patrick and John Collison, former GitHub CEO Nat Friedman, Daniel Gross, and Jane Street.
Starting with relatively limited capital, the fund scaled to roughly $20 billion in under two years. In the first half of 2026, returns reached approximately 439%, with heavy positions in CoreWeave, Bloom Energy, and other storage/compute-related names, cementing his status as the new "AI stock whisperer."
The July Storm: Leverage Meets the Correction
But the good times didn't last.
As markets began reassessing the payback cycle of AI capex, AI hardware and infrastructure stocks saw a concentrated sell-off in July. The fund's top holdings from Q1 — Nebius Group, Sandisk, Micron, CoreWeave — all fell over 35% during the month; Asian names like South Korea's SK Hynix also weakened sharply. Meanwhile, software stocks the fund had shorted (such as Adobe) bounced back, delivering a pincer move against both sides of the book.
More critical was the leverage. According to the Financial Times, Situational Awareness used borrowed capital to amplify returns — reportedly close to 4x leverage — a common hedge fund strategy that generates spectacular gains in uptrends but dramatically magnifies losses in drawdowns.
On July 24, Leopold acknowledged in a semi-annual letter to investors that the fund had "not been spared" from the recent market turbulence (particularly in Asia), while still framing it as the most attractive opportunity since early 2025, offering an August 1 window for additional capital.
But conditions only worsened. Positions kept sliding in the following days, with some names down 30% or more. Margin calls came in rapid succession, leaving the fund with a stark choice: raise capital quickly or sell assets at fire-sale prices.
In the end, Situational Awareness engaged with existing investors and lenders, then quickly turned to large buyers. Around July 30, the fund sold the bulk of its roughly $16 billion public portfolio (both long and short) to Citadel, with Goldman Sachs, JPMorgan, and Bank of America acting as prime brokers facilitating the deal. Fund assets shrank dramatically to around $10 billion, with July's monthly loss hitting roughly 67%. The fund retained some private investments (including roughly $5 billion in Anthropic equity) and plans to continue operating as a private investment vehicle.
Notably, several related AI stocks rebounded sharply immediately after the deal closed — widely interpreted as the forced selling overhang being flushed out, quickly lifting sentiment.
Deeper Analysis: Why the Right Narrative Can Still Get Blown Up
The core lesson here isn't that Aschenbrenner's long-term thesis is necessarily wrong. His repeated emphasis on compute, power, and infrastructure bottlenecks aligns directionally with current fundamentals: Microsoft Azure's continued strong growth, persistent tight supply-demand for Nvidia's high-end chips, and AMD repeatedly raising its AI server market forecasts. All of these point to the same fact: the macro demand curve for AI infrastructure is still sloping upward.
Many deeper analyses conclude that the long-term demand trend for AI infrastructure hasn't been invalidated — only that short-term valuation froth and the pace of capex have undergone a significant correction. The real risk was never about whether the direction was right; it was about whether leverage, position concentration, and timing could be precisely matched. When any one of these falls out of balance, even the sharpest macro narrative can collapse under liquidity pressure.
History is littered with cases of being right on the big picture but dying on timing. Long-Term Capital Management (LTCM) boasted Nobel laureates and precision mathematical models, yet blew up instantly in 1998 under extreme leverage during the Russian debt crisis. Peter Thiel's Clarium Capital made correct but premature bets on oil prices, suffering massive drawdowns before the market came around, ultimately forcing it to wind down. These precedents are a recurring reminder: being right is not a moat — survivability is.
Business Insider attributed Situational Awareness's failure to a brutally precise summary: "Even if you can correctly identify the biggest fundamental shift of a generation, build the right portfolio, and generate spectacular returns in the short term, once you add enough leverage and conviction, the market's final verdict isn't whether your ideas were smart — it's whether you can survive when liquidity dries up." That sentence could serve as the epitaph for leveraged macro strategies.
Moreover, the fund's rapid expansion from a small base to tens of billions of dollars in assets dramatically amplified operational difficulty and systemic fragility. High concentration in public positions, layered on top of leverage, left almost no room to maneuver when Asian markets suddenly swung and global AI stocks sold off in unison. Mass redemption pressure and instant liquidity evaporation compounded each other, forcing the fund to deleverage at the worst possible moment, feeding a loss spiral that turned what might have been a temporary correction into an irreversible exit.
Aftermath and Outlook
For now, Situational Awareness retains some private positions and remaining assets. Aschenbrenner's personal fortune remains substantial, and he's reportedly planning to get married this weekend (around August 1).
Whether this forced liquidation marks a definitive turning point for the fund, or merely a deleveraging pause before a reset, remains an open question.
For the broader market, the AI narrative can be deeply compelling — even exhilarating. But leverage is always a double-edged sword. Between short-term violent swings and long-term macro trends lies the cruelest gap of all: the margin call.
Leopold Aschenbrenner once rewrote the trajectory of his life with a single essay, then used high leverage to amplify that essay's conviction to the absolute extreme. July's liquidation may go down as the most profound — and most expensive — lesson of his career. It is also an emblematic warning sign for the entire AI investment mania.
The market doesn't care how smart you are. It only cares about one thing: whether you can still stand firm when the storm is at its worst.