Ex-OpenAI researcher Leopold Aschenbrenner targets crypto miners in $13.6B AI expansion

Leopold Aschenbrenner is leaning deeper into AI’s infrastructure trade, shifting capital away from semiconductor leaders and toward bitcoin miners and power-backed compute providers.

The former OpenAI researcher—who previously warned about geopolitical threats to advanced AI—has significantly increased exposure to companies supplying electricity, data centers, and high-performance computing, while building sizable bearish positions against chip stocks.

His latest 13F filing with the U.S. Securities and Exchange Commission shows total disclosed equity holdings climbing from $5.5 billion at the end of 2025 to $13.67 billion as of March 31.

The bulk of his long positions centers on bitcoin miners and related infrastructure firms, including IREN, Core Scientific (CORZ), Riot Platforms (RIOT), CleanSpark (CLSK), Bitfarms (BITF), Bitdeer (BTDR), and Hive Digital (HIVE). These companies are increasingly pivoting beyond crypto mining, positioning themselves as key suppliers of energy and compute capacity for AI workloads.

Their control over power contracts and large-scale facilities has made them attractive as demand for AI data centers continues to rise.

Beyond miners, the portfolio also includes notable stakes in Bloom Energy (BE), SanDisk (SNDK), and cloud infrastructure firm CoreWeave (CRWV), reinforcing a broader conviction in the long-term growth of AI’s physical backbone.

At the same time, Aschenbrenner has taken an aggressive bearish stance on semiconductors, disclosing $7.46 billion in put options tied to chipmakers and related exchange-traded funds. Key positions include a $2.04 billion bet against the VanEck Semiconductor ETF, a $1.57 billion put on Nvidia (NVDA), and more than $1 billion in combined exposure to Oracle (ORCL) and Broadcom (AVGO).

The positioning reflects a clear strategic view: as AI scales, the advantage may shift from chip designers to companies that control the energy and infrastructure required to power increasingly compute-intensive