Ex-OpenAI executive allocates $13.6 billion to crypto miners
Leopold Aschenbrenner, the former OpenAI researcher turned hedge fund manager, has sharply expanded his bet on crypto miners and AI infrastructure, with his disclosed equity exposure rising to $13.67 billion in a new SEC filing through March 31, 2026 [1][7]. The move matters now because the portfolio shifts miners from a narrow Bitcoin revenue trade toward a broader proxy for power, land and data-center access as AI demand continues to reshape capital flows.
Overview
- Aschenbrenner’s disclosed portfolio climbed to $13.67 billion by March 31 from $5.5 billion at the end of 2025, showing a much larger allocation to public equities [1][7].
- The fund’s long book includes IREN, Core Scientific, Riot Platforms, CleanSpark, Bitfarms, Bitdeer and Hive Digital, linking the trade to miner-owned infrastructure [1][7].
- He also held positions in Bloom Energy, SanDisk and CoreWeave, indicating exposure to electricity, storage and cloud compute tied to AI buildout [1].
- The filing showed $7.46 billion in put options against semiconductor names and chip ETFs, including Nvidia, AMD and the VanEck Semiconductor ETF [1][7].
- The trade underscores a market view that miners’ electricity contracts and data-center footprints may be more valuable than their legacy Bitcoin production model [7].
- A key risk is that the thesis depends on continued AI infrastructure demand and the ability of miners to convert power assets into durable cash flow [1][7].
Subscribe to our Social Media for Exclusive Crypto News and Insights 24/7!
Aschenbrenner expands crypto miner bet
CoinDesk reported that Aschenbrenner’s latest 13F filing showed a sizeable increase in disclosed equity exposure, with some of the largest long positions concentrated in Bitcoin miners and related infrastructure companies [7]. The filing listed stakes in IREN, Core Scientific, Riot Platforms, CleanSpark, Bitfarms, Bitdeer and Hive Digital, alongside larger positions in Bloom Energy, SanDisk and CoreWeave [1][7].
The portfolio construction points to a clear theme. Market participants view the trade as a wager on physical infrastructure rather than on Bitcoin mining alone. In that framing, the miners matter because they control electricity access, data-center capacity and sites that can be redirected toward AI workloads [7].
Short semiconductors, buy the power layer
Aschenbrenner also disclosed roughly $7.46 billion in put options against semiconductor companies and related ETFs, including a $2.04 billion put on the VanEck Semiconductor ETF and a $1.57 billion put on Nvidia, according to the filing summarized by CoinDesk [1][7]. The bearish positioning extended to Oracle and Broadcom as well [1].
That contrast is important for market structure. It suggests an explicit preference for the owners of power and compute sites over the chip suppliers that sit deeper in the AI hardware chain. Analysts note that the setup reflects a view that scarce electricity and interconnection rights may command a premium as AI buildout accelerates, although the filing does not prove how long those economics will last [7].
| Long positions highlighted | Ticker | Market relevance |
|---|---|---|
| IREN | IREN | Miner with power and data-center exposure [1][7] |
| Core Scientific | CORZ | Public miner pivoting toward infrastructure [1][7] |
| Riot Platforms | RIOT | Large-scale mining and power footprint [1][7] |
| CleanSpark | CLSK | Mining operator with energy infrastructure exposure [1][7] |
| Bitfarms | BITF | Miner with site and power assets [1][7] |
| Bitdeer | BTDR | Mining and infrastructure exposure [1][7] |
| Hive Digital | HIVE | Miner with compute infrastructure exposure [1][7] |
| Short positions highlighted | Ticker | Market relevance |
|---|---|---|
| VanEck Semiconductor ETF | SMH | Broad bet against chip-sector exposure [1][7] |
| Nvidia | NVDA | Direct short on a key AI beneficiary [1][7] |
| AMD | AMD | Another major semiconductor short [1][7] |
| Oracle | ORCL | Exposure to AI-related cloud infrastructure [1][7] |
| Broadcom | AVGO | Chip and networking exposure tied to AI demand [1][7] |
Why miners are back in focus
Aschenbrenner’s trade comes as Bitcoin miners have faced pressure from weaker post-halving economics, which has pushed some operators to diversify into AI and high-performance computing. CoinDesk said his thesis centers on miners that own electricity contracts and data-center footprints needed for the next phase of AI growth [7]. Other reports on the filing described the same move as an energy-infrastructure trade wrapped in an AI thesis [1][3][4].
The investor behavior angle is straightforward. If large public-market allocators begin valuing miners on megawatts, interconnection agreements and site readiness, the group could trade more like infrastructure assets than cyclical Bitcoin producers. That would be a meaningful change for capital formation in the sector, particularly for operators that can monetize surplus power and existing buildings [4][7].
Still, the risk is material. The thesis depends on miners’ ability to convert infrastructure into stable revenue, and on continued demand from AI buyers willing to lease those assets. If AI power demand cools or financing costs stay elevated, the same companies could again be judged primarily on their mining margins rather than their optionality [1][7].
Aschenbrenner’s filing therefore lands as more than a headline-sized trade. It adds another data point to the market’s re-rating of crypto miners, not as pure Bitcoin proxies, but as controllable power platforms with potential utility in the AI supply chain. Whether that valuation shift holds will depend on execution, demand and the speed at which power-constrained AI developers keep paying for capacity [1][7].
Sources
- https://www.coindesk.com/markets/2026/05/18/ex-openai-s-leopold-aschenbrenner-bets-big-on-crypto-miners-for-his-usd13-6b-ai-play
- https://www.youtube.com/watch?v=nCbM4C3hdKk
- https://finance.yahoo.com/news/why-fired-openai-employee-betting-172639848.html
- https://www.binance.com/en/square/post/298358368095282
