$20B AI Bet Collapses: 13F Reveals Shift from Hedging to Leverage

Woofun AI reports that Situational Awareness LP, the investment vehicle managed by Leopold Aschenbrenner, disclosed its 13F quarterly holdings report to the SEC on August 15th, revealing the structural vulnerabilities that precipitated the fund's recent collapse. The filing provides a critical snapshot of the portfolio as of June 30th, just before the entity faced severe market headwinds that forced a distressed liquidation of assets to Citadel, marking the end of an era for the manager once hailed as the 'AI stock god.'

The trajectory of Situational Awareness LP was defined by explosive growth in asset accumulation, a pattern clearly documented in sequential regulatory filings. The fund began with public holdings valued at merely $225 million in Q4 2024, a modest starting point for an entity that would soon attract global attention. By the time the Q4 2025 holdings report was released in February of this year, the portfolio had expanded to $5.5 billion, signaling early momentum in the AI investment thesis. This growth accelerated dramatically in the following quarter, with the Q1 holdings report released in May showing a surge to $5.

5 billion. The momentum continued unchecked, reaching $13.7 billion in subsequent filings, before the latest data indicated a nominal holding value of $20.2 billion. It is crucial to note that under SEC regulations for 13F reporting, funds must submit these documents within 45 days after the end of each calendar quarter, and the valuation of option assets is based on the notional value of the underlying stocks rather than the actual premium paid, which inflates the apparent size of the portfolio.

The pre-collapse portfolio snapshot as of June 30th reveals a strategy characterized by extreme concentration in specific sectors of the AI supply chain. Situational Awareness LP disclosed 26 holdings in total, with a combined nominal value of approximately $20.2 billion. The most dominant feature of this balance sheet was the heavy allocation to storage chip manufacturers, specifically SanDisk (SNDK) and Micron (MU). SanDisk represented the largest single position, accounting for $5.674 billion, or 28.

0% of the total portfolio. Micron followed closely as the second-largest holding, valued at $5.574 billion, representing 27.5% of the fund's assets. Together, these two entities constituted $11.2 billion of the portfolio, equating to 55.5% of the total nominal value. This data indicates that more than half of the fund's public market exposure was tied to the performance of just two companies in the storage sector, creating a fragile dependency on the continued appreciation of these specific equities.

Beyond the primary storage holdings, the portfolio included significant positions in other segments of the AI infrastructure ecosystem, reflecting a broad but correlated bet on industry growth. Bloom Energy held a position valued at $1.899 billion, targeting the power demand associated with AI data centers. TSMC, a key player in advanced semiconductor manufacturing, accounted for $1.265 billion. Nebius, a provider of AI cloud computing services, held $1.233 billion.

Additionally, the fund maintained exposures to CoreWeave, Core Scientific, Applied Digital, and IREN, entities involved in AI cloud computing, data centers, and energy infrastructure. These holdings, while diverse in terms of company names, were unified by a single thesis: that the demand for AI computing power would drive sustained growth across the entire value chain, from chips and storage to electricity and data center operations.

The strategic risk inherent in this portfolio structure lay in the combination of high concentration and high leverage, a pairing that proved fatal when market conditions shifted. The fund's thesis relied on the continuous expansion of AI infrastructure construction, a narrative that supported elevated valuations during periods of market optimism.

However, this approach lacked diversification across uncorrelated asset classes, meaning that a downturn in the AI sector would impact the majority of the portfolio simultaneously. The use of leverage amplified both gains and losses, reducing the fund's resilience to market volatility. When the AI infrastructure narrative faced skepticism, the lack of hedging mechanisms left the portfolio exposed to rapid devaluation, leading to a liquidity crisis that forced the manager to seek emergency capital solutions.

Woofun AI data shows that a critical question surrounding the fund's collapse was the disappearance of protective hedges that had been present in earlier quarters. In the Q1 holdings report released in May, Situational Awareness LP demonstrated significant caution regarding the overheating of chip and storage stocks. At that time, the fund held put options on leading companies in these sectors with a notional value exceeding $8 billion. These positions covered major entities including SMH, NVDA, ORCL, AVGO, AMD, and ASML, accounting for more than 60% of the fund's total nominal holdings. This hedging strategy was designed to offset potential losses in the long positions if the market corrected. The presence of these puts suggested that the manager anticipated a possible downturn and had taken steps to mitigate downside risk, a prudent move given the high valuation multiples prevalent in the AI sector at the time.

However, the Q2 13F report revealed a drastic reversal in this risk management approach, as the fund liquidated nearly all of its protective put options. The sell list showed a complete abandonment of the hedges that had previously shielded the portfolio. Specifically, the fund reduced its SMH Put position by 14.94%, its NVDA Bear ETF Put by 11.47%, and its ORCL Put by 7.84%. Further reductions included AVGO Put (-7.36%), AMD Put (-7.09%), TSM Put (-3.91%), and ASML Put (-3.61%).

Additionally, the put options on MU and SNDK, which had been structured as call options in Q1, were closed out with reductions of 3.09% and 2.84%, respectively. These positions were replaced by direct stock holdings, effectively removing the downside protection and exposing the fund to unlimited loss potential in the event of a market decline.

This shift from a hedged strategy to an unprotected long-only approach significantly increased the portfolio's vulnerability to market shocks.

The buy list in the Q2 report further illustrated the fund's commitment to an aggressive long strategy, with new additions reinforcing the exposure to AI infrastructure. TSMC ADR saw an increase of 6.19%, while Nebius, a newly added position, accounted for 6.09% of the portfolio. STMicroelectronics STM was added with a 2.89% allocation, and SharonAI entered the portfolio with a 2.12% stake. Keel Infrastructure was also added, representing 0.75% of the holdings. These additions spanned various segments of the AI value chain, from European wafer fabrication to Russian-backed cloud services and power infrastructure. The consistent theme across these new positions was a belief in the long-term growth of AI demand, but the lack of corresponding hedges meant that any short-term correction would directly impact the fund's net asset value without any offsetting gains from short positions.

The consequences of this strategic shift became apparent in July, when the AI sector experienced a systemic correction. Without the protective put options that had been liquidated in Q2, the fund was fully exposed to the downward pressure on stock prices. Micron (MU), one of the largest holdings, saw its price drop from $1,154 to $739 by July 29th, a decline of 35.9%. SanDisk (SNDK) suffered an even steeper fall, dropping from $2,274 to $1,016, a decrease of 55.3%. These losses, amplified by the fund's leverage, rapidly eroded the capital buffer, triggering margin calls and forcing a distressed sale of assets. Ken Griffin, the owner of Citadel, stepped in to purchase the majority of the public market holdings at a discount, effectively rescuing the fund from total collapse but at the cost of significant value destruction for the original investors.

The aftermath of the collapse leaves Leopold Aschenbrenner, now 25 years old, with a diminished public portfolio but still holding significant private assets. The fund's public market holdings have been largely liquidated, but the manager retains investments in private companies, including shares in Anthropic, a leading AI research organization. These private holdings represent a crucial remaining asset for the young investor, offering potential for future recovery despite the recent setback. The incident serves as a cautionary tale about the risks of high leverage and concentrated positions in volatile sectors, highlighting the importance of robust risk management strategies in the face of market uncertainty. While the 'AI stock god' narrative has been disrupted, the underlying thesis on AI growth remains intact, suggesting that the manager may yet rebuild his portfolio in the coming years.