Leopold Aschenbrenner's Situational Awareness hedge fund nearly collapsed in July, lost $35 billion in assets, and was forced to sell all of its public equity positions to Ken Griffin's Citadel. By any reasonable measure, that should have been a wake-up call. Instead, the fund manager used its return to financial markets in September 2024 to do essentially what got him into trouble in the first place: bet heavily on AI stocks. The 13F filing released Monday reveals a portfolio that looks less like a chastened investor learning from catastrophic losses and more like someone doubling down on the exact same thesis that nearly destroyed his fund.
The holdings tell a stark story. Aschenbrenner took a $440 million equity position in Australian neocloud company SharonAI, along with significant stakes in Cleanspark, IREN, Bitdeer Technologies, and Fermi. He also maintained call options in companies like Nebius and Sandisk—the very neocloud and memory chip plays that nearly took his fund down over the summer. These weren't small positions meant to test the waters; they were substantial bets that the AI trade would continue its torrid pace. The problem is that every single one of these holdings has been crushed since September 30.
SharonAI has lost 58% in October alone and fallen another 11% this month. IREN dropped 41% in October. Nebius, one of the original problem children that nearly sank the fund, declined 35% and then plummeted more than 16% in a single day last week after issuing a weak full-year forecast. The timing could hardly be worse. Industry leaders like Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman spooked investors in September with comments about AI development slowdowns. Meanwhile, Democratic election wins have raised concerns about data center development, questions persist about whether technology companies are using excessive debt to finance AI buildout, and Jensen Huang's departure from his CEO role at Nvidia has sparked additional uncertainty.
What's remarkable here is the apparent lack of learning. Aschenbrenner's fund peaked at $45 billion in assets before evaporating down to $10 billion. That humbling experience should have prompted serious reflection about concentration risk, sector rotation, and the dangers of being overly tethered to a single narrative—no matter how compelling. Instead, he rebuilt by piling back into the same AI-centric positions at precisely the moment when sentiment was shifting and rational investors were beginning to question whether the AI hype had run too far ahead of reality.
The 13F filing only shows positions as of September 30, so it's unknown whether Aschenbrenner exited these positions during October's selloff. But even if he has since cut losses, the decision to re-enter so aggressively into positions that had already nearly destroyed him raises serious questions about judgment and risk management. For investors considering whether to trust this manager with their capital again, the recent moves offer little reassurance that anything has fundamentally changed.