Leopold Aschenbrenner, a former OpenAI researcher who warned publicly that China could steal advanced AI models, dramatically expanded his disclosed portfolio this spring — increasing reported equity exposure from $5.5 billion at the end of 2025 to $13.67 billion by March 31 and piling on $7.46 billion in put options against major semiconductor companies and chip-focused ETFs.
The 13F filing shows stakes in bitcoin miners and related infrastructure names including IREN, Core Scientific, Riot Platforms, CleanSpark, Bitfarms, Bitdeer and Hive Digital, alongside large positions in Bloom Energy, SanDisk and cloud provider CoreWeave. On the bearish side, Aschenbrenner opened $7.46 billion worth of puts, with his single largest short a $2.04 billion put against the VanEck Semiconductor ETF and a $1.57 billion put against Nvidia; the filing also listed more than $1 billion in puts tied to Oracle and Broadcom.
Business Insider reported May 18 on the same filing, noting the fund disclosed more than $1.5 billion in Nvidia puts and over $2 billion in VanEck Semiconductor ETF puts, underscoring the scale and focus of the positions.
The numbers matter because they reveal a concentrated view of where value for AI workloads will come from. Aschenbrenner’s long book favors companies that supply electricity, data center capacity and compute infrastructure for AI — the kinds of firms bitcoin miners have been reshaping themselves into as they seek revenue beyond cryptocurrency. At the same time, the huge put positions amount to a broad bet that some of the most celebrated chipmakers and chip ETFs are overvalued relative to that infrastructure trade.
That portfolio tilt tracks a public arc for Aschenbrenner. He went viral in 2024 for his paper, “Situational Awareness: The Decade Ahead,” and launched his hedge fund, Situational Awareness, three months later. The filing shows the firm has moved quickly from ideas into market-sized positions: expanding equity exposure from $5.5 billion to $13.67 billion in roughly three months and layering in nearly $7.5 billion of downside protection on the semiconductor complex.
The tension in the portfolio is sharp. The same thesis that leans on power, data-center capacity and specialized cloud suppliers also requires compute — the chips produced by the companies he is betting against. Owning large stakes in firms such as CoreWeave and Bloom Energy while buying deep, multi-billion-dollar puts on Nvidia and a semiconductor ETF forces a question: can AI scale without the very chipmakers he is shorting, or does he expect a reordering in which infrastructure providers capture far more of the economics than the silicon vendors?
There is also a timing risk built into the positions. Expanding disclosed exposure from $5.5 billion to $13.67 billion by March 31 is a fast tempo for any fund, especially one that rose to prominence after a viral paper. The math in the filing — $7.46 billion in puts versus concentrated long stakes in energy and infrastructure names — reads like a two-part wager: protect against a drawdown in semiconductor equities while collecting upside from companies that supply power and capacity for AI workloads.
Read two ways, the filing is either a bold hedged long on the parts of the AI stack that are not chips, or a levered speculation that the market will reprice winners and losers in the AI supply chain. Given the size of the put positions — including the $2.04 billion VanEck Semiconductor ETF put and the $1.57 billion Nvidia put — Aschenbrenner’s moves will be watched for their potential to pressure sentiment in the semiconductor names he has targeted, even as his long bets signal confidence in AI-related infrastructure.
If the trade holds, it points to a larger thesis: the next chapter of AI returns may flow less from headline chip performance and more from the companies that provide power, space and specialized compute to run the models. Leopold Aschenbrenner has put real capital behind that view; the markets will soon test whether those convictions rest on a durable shift in where AI creates profit, or on a short-term repricing of some of the sector’s most celebrated names.
