Leopold Aschenbrenner, who departed OpenAI in April 2024 after challenging the lab's security protocols, filed his first 13F as manager of Uncorrelated Ventures in March 2025. The portfolio holds north of $500 million in US-listed equities, heavily weighted toward companies that supply electrical infrastructure and data center capacity to Bitcoin miners and AI compute operators. The filing marks a clean break from the policy-focused AGI safety work that defined his tenure at OpenAI and places him among a narrow cohort of former frontier-lab researchers now allocating institutional capital in public markets.
The disclosed positions center on power generation and cooling infrastructure providers serving Bitcoin mining operations. Aschenbrenner's largest holdings include Riot Platforms, Marathon Digital, and CleanSpark, three miners that collectively operate more than 1.2 gigawatts of nameplate capacity across North America. Secondary positions in Core Scientific and Cipher Mining add exposure to firms that lease hashrate to third-party operators, a business model that insulates revenue from spot Bitcoin volatility but ties returns to network difficulty adjustments and electricity spread. The portfolio avoids direct Bitcoin exposure and holds no spot ETF shares, suggesting a thesis on the pick-and-shovel layer rather than the asset itself.
The structure tells allocators two things. First, Aschenbrenner is betting that compute infrastructure—whether for training runs or proof-of-work consensus—becomes the scarce input as both AI labs and Bitcoin miners compete for stranded energy and grid capacity. The overlap is not theoretical. Crusoe Energy, a private operator backed by Founders Fund and Valor, already runs GPU clusters and mining rigs on the same flared gas in the Permian Basin, arbitraging workload by power price. Second, the public-equity wrapper implies Aschenbrenner expects liquidity events or M&A in this subsector within 18 to 24 months. Marathon acquired two smaller miners in Q4 2024, and Riot has publicly discussed consolidation as hashrate difficulty climbs past 120 exahashes per second. A concentrated book in a narrow industry subset rarely survives without near-term catalysts.
Allocators should watch three follow-on events. One, whether Uncorrelated Ventures raises a private vehicle to take pre-IPO stakes in companies like Applied Digital or Hut 8, which straddle mining and AI hosting. Two, how the portfolio adjusts if Bitcoin breaches $120,000 and miners shift from infrastructure reinvestment to balance-sheet accumulation, a move that would compress equipment demand. Three, whether Aschenbrenner publishes research or white papers on compute scarcity, a signal that he is building intellectual capital to justify a multi-strategy fund rather than a single-thesis vehicle. The first follow-on filing will clarify whether this is a long-duration infrastructure bet or a tactical play on the current mining cycle.
The Q2 13F, due in mid-May, will show whether Aschenbrenner added to positions after Bitcoin's March pullback or began trimming into strength. His public commentary on AGI timelines—he has argued for 2027 to 2028 arrival—suggests he views compute access as the binding constraint for the next 36 months. If the portfolio holds or expands, it prices that view into public equities before private markets catch up.