Bloom Energy

Leopold Aschenbrenner’s AI-focused hedge fund has surged to $20 billion in assets under management with 270% gains through May 2026, driven by a concentrated bet on Anthropic that now represents roughly one-fifth of the portfolio. The strategy hinges on a thesis that infrastructure constraints, electricity and computing power, will determine winners in AI, not software innovation alone, a bet that faces significant concentration risk if AI spending slows or market conditions shift.

• Situational Awareness fund gained approximately 270% after fees in 2025-2026, turning $1 invested in January into nearly $4 by May.
• Anthropic position grew from $60 billion valuation in February 2025 to $965 billion by May 2026, now representing roughly 20% of fund assets.
• Fund maintains over $1.5 billion in short bets against Nvidia and $2 billion against chip sector, betting infrastructure scarcity outpaces semiconductor supply.

• $20B Current assets under management, up from startup capital in 2024
• 270% Year-to-date returns after fees through May 2026 period
• $965B Anthropic valuation in May 2026, up from $60 billion in February 2025

Situational Awareness, the hedge fund launched by 24-year-old Leopold Aschenbrenner, a former OpenAI researcher, has emerged as one of Wall Street’s most successful AI-focused investment vehicles, managing approximately $20 billion and delivering outsized returns that have attracted attention from institutional investors including Jane Street, the secretive trading giant.

The fund’s 270% gain through May 2026, meaning a dollar invested in January would have grown to nearly four dollars by spring, reflects a concentrated thesis about artificial intelligence infrastructure that diverges sharply from consensus market positioning.

Aschenbrenner’s central argument, articulated in a 165-page viral essay published in 2024, holds that physical constraints around electricity and computing capacity, rather than algorithmic innovation, will ultimately determine which companies capture value in the AI era.

Aschenbrenner’s Infrastructure-First Thesis Bets Against Nvidia’s Dominance

The fund’s strategy inverts conventional AI-sector positioning by simultaneously going long on power and computing infrastructure providers while taking substantial short positions against semiconductor leaders.

Aschenbrenner has deployed more than $1.5 billion in short bets specifically targeting Nvidia, the dominant chipmaker, and approximately $2 billion against a broader basket of semiconductor stocks.

His reasoning rests on a simple but contrarian claim: market prices for chips already reflect overly optimistic scenarios where everything proceeds smoothly, while the actual bottleneck will shift to electricity and raw computing capacity rather than silicon availability.

This infrastructure-first positioning has led the fund to accumulate holdings in unconventional AI plays. Bloom Energy, a fuel cell manufacturer that generates power on-site, represents his largest single public stock position. CoreWeave, a private company that rents out AI computing capacity to enterprises, also appears prominently in the portfolio.

More creatively, several former Bitcoin mining data centers that have repurposed their operations to run AI workloads feature among his positions, a bet that former miners’ existing power infrastructure and real estate make them natural transitions to AI compute providers.

The short bets against chip stocks reflect Aschenbrenner’s belief that the semiconductor consensus has priced in permanent supply shortages that will not materialize. Instead, he argues, capital will flow toward whatever truly constrains scaling, which he identifies as reliable, cost-effective power delivery.

That thesis has proven profitable so far, as the fund’s overall returns have compounded alongside rising expectations that power availability, not chip availability, represents AI’s binding constraint.

Anthropic Position Balloons to 20% of Fund as Valuation Skyrockets

The fund’s single largest holding is not a publicly traded stock but rather a private equity stake in Anthropic, the AI company behind Claude, one of the leading consumer-facing large language models.

Aschenbrenner purchased that position in February 2025 when Anthropic’s valuation stood at approximately $60 billion, a significant but not exceptional valuation for a leading AI laboratory at that time.

Between February 2025 and May 2026, Anthropic’s valuation exploded to $965 billion following a new funding round, a roughly 16-fold increase that transformed Aschenbrenner’s initial position into the fund’s most valuable single asset.

The Anthropic stake now comprises roughly one-fifth of Situational Awareness’ total assets under management, concentrating the fund’s returns on a single company whose valuation continues to move sharply. That concentration risk is amplified by Anthropic’s transition toward a confidential initial public offering, a process that could unlock liquidity but also introduce new volatility.

Aschenbrenner’s firm appears among Anthropic’s formally listed investors, a signal of both the investment’s magnitude and the portfolio company’s institutional standing in the AI ecosystem.

The Anthropic investment also reflects a subtle refinement of Aschenbrenner’s original thesis. While his infrastructure argument emphasizes power and compute as the binding constraints, Anthropic represents a pure-play bet on frontier AI capability and the commercial value of models trained on massive compute clusters.

In other words, the fund is not simply betting on the picks-and-shovels vendors; it is also holding a significant stake in a major consumer of those resources, creating dual exposure to both infrastructure scarcity and the value capture of leading AI systems.

Concentration Risk Intensifies as Fund Becomes Dependent on Anthropic and AI Spending Momentum

The fund’s extraordinary returns mask a structural vulnerability inherent in concentrated investment theses. If companies materially slow their AI capital expenditures, if power becomes sufficiently abundant to ease infrastructure constraints, or if Anthropic’s private valuation fails to sustain its recent run, the fund could experience equally dramatic reversals.

The same concentration risk applies across Bitcoin miner AI stocks more broadly, a sector that has experienced sharp moves both upward and downward as market sentiment shifts.

Aschenbrenner’s thesis depends on several linked assumptions holding: that the AI industry will continue spending capital at accelerating rates, that electricity and compute capacity will remain the actual binding constraints, that infrastructure providers will capture meaningful margins, and that Anthropic’s valuation reflects genuine value rather than speculative excess.

Should any of these prove incorrect, the fund’s dominant positions, particularly the 20% Anthropic allocation, would face immediate repricing pressure.

The coming months will show whether shovels really do beat gold.

Fund positioning and statement of thesis

Aschenbrenner’s next major test arrives with Anthropic’s confidential IPO process, which will determine whether the $965 billion private valuation translates to comparable public market pricing and whether his concentrated bet generates exit liquidity or faces valuation compression upon listing. Simultaneously, actual AI capital spending by major cloud providers in the second half of 2026 will reveal whether growth remains on the trajectory his thesis requires or whether spending moderates, forcing a reassessment of infrastructure constraint severity.