Peter Wildeford, who spent two years at OpenAI working on alignment research, filed a 13F this week showing the Quantified Uncertainty Research Institute Fund holds positions across four Bitcoin mining operations totaling approximately $47 million. The largest position: $22.8 million in Marathon Digital Holdings, followed by $13.1 million in Riot Platforms, $7.4 million in CleanSpark, and $3.7 million in Cipher Mining. The filing marks the first time a fund led by a former frontier AI researcher has disclosed material exposure to the energy-intensive mining sector.

The positions were built during Q4 2024, a period when Bitcoin climbed from $63,000 to above $95,000. Marathon traded between $16 and $29 during the quarter; Riot ranged $9 to $18. Wildeford's fund entered near the middle of those bands, avoiding both the September lows and the December peaks. The 13F shows no options overlay, no short positions—just equity exposure. The fund's total disclosed AUM sits at $183 million, meaning the Bitcoin miner basket represents roughly 26% of the portfolio. That concentration level typically signals conviction, not exploration.

The timing carries second-order weight. Wildeford left OpenAI in early 2023, before the ChatGPT explosion reshaped compute economics. He spent the next eighteen months building QURI as a quantitative research operation focused on forecasting and decision theory. The Bitcoin miner positions arrive as institutional players begin separating mining from trading—treating it as an infrastructure bet on proof-of-work security rather than a leveraged bet on coin price. Marathon operates 40 exahash of capacity; Riot runs 28 exahash. Both vertically integrated in 2024, acquiring power contracts and ASIC inventory to control the full stack. Wildeford's entry price implies a view that hashrate growth and operational efficiency will outrun coin volatility over the next twelve to eighteen months.

The filing also surfaces a structural question: whether AI researchers rotating into Bitcoin mining signals a broader recognition that compute-intensive businesses share risk profiles. OpenAI's training runs consume tens of megawatts; a mid-sized mining operation pulls similar loads. Both face stranded power risk, both optimize for uptime, both live inside regulatory uncertainty. Wildeford's experience managing inference costs at OpenAI likely informed his analysis of mining unit economics—kilowatt-hours per terahash, ASIC depreciation curves, power purchase agreement structures. The overlap is tighter than most allocators assume.

Operators should watch three follow-on events. First, whether QURI adds to the position in Q1 2025 or holds flat—the next 13F deadline is May 15. Second, whether other AI-adjacent funds follow the pattern; several former DeepMind and Anthropic researchers now run capital, and their filings lag by one quarter. Third, the April Bitcoin halving will cut miner revenue by 50%, forcing operators to either improve efficiency or raise capital. Marathon and Riot both have $500 million+ in cash, but CleanSpark and Cipher operate closer to the margin. If Wildeford trims the smaller names and concentrates into Marathon, that's the tell.

The 13F lists no sales during Q4. The positions remain open. Wildeford has not commented publicly, and QURI's website carries no investor letters. The filing stands alone—a former alignment researcher now long 26% of his fund in machines that solve useless equations to secure a distributed ledger. The irony writes itself, but the position sizing suggests he checked the math.