Leopold Aschenbrenner, the former OpenAI researcher who now runs a hedge fund, has made one of the most provocative bets in recent market history. His fund, Situational Awareness LP, held $13.67 billion in equity exposure as of March 31, 2026 — nearly triple the $5.52 billion recorded at the end of 2025. The central thesis: former Bitcoin miners with vast energy infrastructure holdings represent the true bottleneck play for scaling AI.

Aschenbrenner is best known for 'Situational Awareness,' a 2024 essay arguing that artificial general intelligence could arrive by the end of the decade and that compute buildouts would be the decisive factor in reaching it — a worldview that directly frames the fund's positioning. The fund itself made headlines when its disclosure filings became public, offering a rare window into how a large, thesis-driven fund is trading the AI infrastructure theme.

The 24-year-old's top holdings resemble a roster of companies that once operated warehouses full of ASIC mining rigs. IREN, Core Scientific, Riot Platforms, and CleanSpark all feature prominently in the portfolio, alongside energy firm Bloom Energy and compute company CoreWeave.

Power over chips

Aschenbrenner's investing philosophy turns the conventional AI narrative on its head. His fund is wagering that chips are no longer the truly scarce resource. In his view, the real constraints on AI scaling are power and land — securing grid connections, building substations, and obtaining site permits.

The bet comes amid a documented scramble for electricity across the AI sector, where utilities and grid operators in the United States have reported surging interconnection requests from data center developers, and new facilities are increasingly sited around available power rather than fiber connectivity.

Former Bitcoin miners happen to own exactly those assets. Over the years they negotiated power purchase agreements, locked down industrial sites near cheap electricity, and built out the cooling and electrical infrastructure required to run thousands of machines at full capacity. By comparison, swapping mining rigs for GPU clusters is the relatively easy part.

Core Scientific has already begun that transition, announcing AI hosting arrangements that include multi-year contracts expected to generate substantial revenue. In under four years, the company has gone from a late-2022 bankruptcy filing to becoming one of the most talked-about AI infrastructure plays. The pivot mirrors a wider pattern across the sector: Hut 8, TeraWulf, and other former miners have likewise signed AI hosting deals or repurposed energy assets for high-performance computing, moves that have at times driven sharp revaluations of their publicly traded shares.

The semiconductor hedge

What makes Aschenbrenner's positioning especially notable is the other side of the trade. Alongside his long positions in miners and energy, the fund disclosed roughly $7.46 billion in put options against semiconductor stocks.

The logic reflects the same thesis. If power and physical infrastructure are the genuine bottleneck, the market may be overpricing chip supply while underpricing everything downstream: the land, the energy, the permits, and the cooling systems. A GPU sitting in a box generates no revenue; a GPU plugged into a rack inside a facility with 200 megawatts of guaranteed power does.

From crypto to compute: the broader trend

CoreWeave, another of the fund's key positions, represents the pure-play expression of this thesis. The company began as a crypto mining operation before transforming into a GPU cloud provider, and it has since attracted billions in financing to build out AI-focused data centers.

The inclusion of Bloom Energy adds a further layer. Bloom manufactures solid oxide fuel cells capable of on-site power generation — a potential solution for data centers that cannot wait years for utility grid upgrades. This 'behind-the-meter' approach has drawn growing interest in the data center industry, where grid interconnection queues can stretch for years in key markets.

The drawdown and what comes next

By mid-2026, Situational Awareness LP suffered sharp losses amid broader volatility in AI-linked stocks. The drawdown was significant enough to prompt a major liquidation of public equities through a transaction with Citadel. The episode is a reminder that concentrated, thesis-driven positioning cuts both ways: the same leverage to AI infrastructure demand that powered the fund's rapid growth also amplified its losses when sentiment turned.

For those evaluating the miners in Aschenbrenner's portfolio, the key metric is no longer hash rate. It is megawatt capacity under contract for AI workloads, and the margin differential between mining a Bitcoin and renting a rack to an AI company. Watch for future disclosures on contract terms, tenant credit quality, and whether planned AI capacity is actually energized on schedule — the same factors that will test whether the power-over-chips thesis holds.