AI hedge fund Situational Awareness LP September 30 disclosure revealed a $563 million position in CoreWeave (Nasdaq: CRWV), along with a $362 million position in Core Scientific (Nasdaq: CORZ) and a $338 million position in IREN (Nasdaq: IREN). IREN and Core Scientific were the firm’s largest Bitcoin mining positions, as of September 30. All position nominal values are back-dated to September 30, the filing date with the SEC.
The firm added Bitcoin miners Cipher Mining (Nasdaq: CIFR), Hut 8 (Nasdaq: HUT), Riot (Nasdaq: RIOT), Bitfarms (Nasdaq: BITF), Galaxy Digital (Nasdaq: GLXY) and Bitdeer (Nasdaq: BTDR). Situational Awareness trimmed its position in Applied Digital (Nasdaq: APLD).
Situational Awareness was founded by former OpenAI researcher Leopold Aschenbrenner, who left the company in April 2024.
The fund also added Lumentum, Coherent, Modine, Western Digital, and Seagate in the AI supply chain sector.
Situational Awareness expanded a call option position in Intel. The fund reported a $459 million call position in May. The position increased to $678 million in the Friday filing. The fund reported $76 million in Broadcom put options.
In Aschenbrenner’s 165-page writing titled “Situational Awareness: The Decade Ahead,” the AI hedge fund’s founder wrote that he expects recursively self-improving AI systems and AGI by 2027, and to achieve that, data centers remain a bottleneck. These data center facilities could reach trillion-dollar levels of investment before the end of the decade. From the energy side, AI data center centers will consume 20% of U.S. electricity demand by 2030. He based his projections on compute, FLOPs, and model capabilities.
Daniel Reeves, cofounder of Beeminder, backs Aschenbrenner’s predictions based on stargate and hyperscale spending. The $500 billion stargate project alone would represent about half the scale of Aschenbrenner’s AI data center investment prediction.
Hyperscaler capital expenditures also rose from $239 billion in 2024 to $368 billion from a Goldman Sachs research. The investment bank also projects hyperscaler spending at $432 billion in 2026. AI-related capital expenditure could reach 3% of U.S. GDP.
