By Rajat Baijal, Part-Time Lecturer, M.S. in Enterprise Risk Management Program, Columbia University School of Professional Studies
In early July 2026, Leopold Aschenbrenner’s AI-focused hedge fund Situational Awareness stood at roughly $45 billion in assets, up 439% net return through June 2026 and drawing comparisons to some of the great trades in modern markets. By the end of the month, the fund had been forced to sell its entire public equities book to Citadel at a discount, with assets cut to a fraction of their peak. The unraveling took less than three weeks.
The fund’s strategy was straightforward in concept: go long on the physical infrastructure of the AI buildout — chips, power, data centers — while shorting software incumbents seen as vulnerable to AI disruption. What made the strategy fragile was not the thesis but the way it was expressed: reported leverage as high as 400%, concentrated positions in names like CoreWeave, Nebius, and SK Hynix, and a short book that was supposed to hedge the longs but instead moved against the fund at the same time the longs collapsed. When both legs of a “hedged” trade lose money simultaneously, the hedge has failed, and the portfolio behaves like an unhedged, maximally leveraged bet.
The broader equity market barely flinched. The S&P 500 stayed near record highs throughout the entire episode. That divergence is perhaps the most important lesson from this case: a fund can suffer a catastrophic, career-defining loss while the index that most risk dashboards watch shows nothing unusual at all. Several critical factors emerge from a closer look at the Situational Awareness collapse:
1. The Illusion of Diversification: A long/short structure is only a hedge if the two legs are genuinely uncorrelated in stress. Here, AI infrastructure longs and software shorts were both, in effect, bets on the same underlying factor — AI sentiment — expressed from opposite directions. In a sentiment reversal, “hedged” became “doubly exposed.”
2. Leverage as an Amplifier, Not a Strategy: Leverage did not cause the initial drawdown; it transformed a manageable price correction into a forced liquidation. As collateral values fell, prime brokers made margin calls, and meeting them required selling into an already falling market — the classic deleveraging spiral, and a mechanism nearly identical to LTCM, Amaranth, and Archegos before it.
3. The Fallacy of Being Right Eventually: The fund’s investment thesis may ultimately prove correct. AI infrastructure demand may continue to grow. Computing requirements may continue to increase. Power availability may remain a critical constraint. But being right about the long-term direction of a market does not guarantee survival as an investor.
4. Risk Models Can Fail Precisely When They Are Needed Most: Traditional risk measurements often rely heavily on historical volatility, correlations and observed liquidity. But the most dangerous events occur when those relationships change. During a market regime shift, correlations can converge toward one, liquidity can disappear and volatility can increase precisely when leverage must decrease.
The question for risk managers is therefore not simply: What is our expected loss?
It is: What happens to the portfolio when our assumptions about volatility, correlation and liquidity are simultaneously wrong? That is a scenario-analysis question rather than a conventional VaR question.
5. The Speed of the Unwind Matters: Perhaps the most striking feature of the episode was the speed at which the situation changed. The fund reported a 439% net gain through June. On July 24, an investor letter characterized the market dislocation as creating attractive investment opportunities and sought additional capital. Within days, the firm was dealing with margin pressure and ultimately sold most of its public equity portfolio. This illustrates the importance of risk velocity.
Traditional risk assessments often focus on probability and impact. But for highly leveraged institutions, a third dimension is critical: velocity. A risk that can move from manageable to existential within days requires fundamentally different controls from a risk that develops over months or years.
Looking Ahead
The Situational Awareness collapse will likely join LTCM, Amaranth, and Archegos in the standard canon of leverage-and-concentration case studies — not because the underlying thesis was necessarily wrong, but because the position sizing and hedge construction left no room for a normal drawdown to remain survivable. For risk professionals, the lesson is not about AI specifically; it is about the perennial gap between a fund’s stated risk framework and its actual factor exposure under stress, and about how quickly market-implied trust can evaporate once a leveraged position becomes visible to counterparties with an incentive to test it.
Ultimately, Situational Awareness is a case study in the difference between being right and being resilient.
In enterprise risk management terms, this episode is a case study in the difference between measured risk and realized risk — and a reminder that a fund’s best-ever quarter often deserves the most scrutiny of its leverage and hedges — not the least, as is usually the case.
Views and opinions expressed here are those of the author and do not necessarily reflect the official position of Columbia University School of Professional Studies or Columbia University.
About the Program
The M.S. in Enterprise Risk Management (ERM) program at Columbia University prepares graduates to inform better risk-reward decisions by providing a complete, robust, and integrated picture of both upside and downside volatility across an entire enterprise. For both the full-time and part-time options, students may take all their courses on Columbia’s New York City campus or choose the synchronous online class experience. Learn more about the program here.
