A 4x-leveraged AI hedge fund went from $45 billion to $10 billion in six weeks

Situational Awareness, the AI-focused hedge fund run by 24-year-old former OpenAI researcher Leopold Aschenbrenner, has been forced to sell its entire public equities portfolio to Ken Griffin's Citadel after a leveraged bet unwound in under six weeks, cutting the fund's assets from a peak near $45 billion to roughly $10 billion. The fund will continue operating as a private investment vehicle, retaining its roughly $5 billion stake in Anthropic, but its run as one of the most closely watched names in AI-adjacent public markets investing has effectively ended.

The strategy that built the fund

Situational Awareness launched in late 2024 with a few hundred million dollars from backers including Stripe co-founders Patrick and John Collison, former GitHub CEO Nat Friedman, and investor Daniel Gross. Aschenbrenner structured the fund around a specific thesis: increasingly capable AI systems would drive enormous, sustained demand for the physical infrastructure underneath them — semiconductors, memory, data centers, and power generation. The trade was long AI infrastructure names like SK Hynix, CoreWeave, Nebius, Micron, and Bloom Energy, financed against short positions in software incumbents such as Adobe, all levered at roughly 4x. The fund had returned more than 1,000% since inception, including 439% in the first half of 2026 alone, making it one of the best-performing hedge funds on Wall Street heading into the summer.

How six weeks erased the gains

From June 22 through July 30, the trade reversed sharply on both legs simultaneously. Marvell Technology, one of the fund's largest infrastructure holdings, fell 40.44% over that span, while other AI infrastructure names in the portfolio — SK Hynix, CoreWeave, Nebius, Micron, and Bloom Energy — dropped between 35% and 47%. On the short side, Adobe rallied 27.19% to close at $247.90, helped by record Q2 revenue of $6.62 billion and AI-first annual recurring revenue that tripled year-over-year past $500 million. At 4x leverage, a long book down roughly 40% combined with a short book up roughly 27% is a mathematical wipeout — there was no scenario in that combination of moves where the position survived intact.

The unwind's market impact extended well beyond the fund itself: Morgan Stanley's sector-neutral Momentum Index tumbled 17.4% over just four trading days, its worst such decline on record — a steeper momentum reversal than the dot-com crash, the pandemic shock, or the 2022 bear market produced. That scale of dislocation suggests other funds running similar AI-infrastructure-long, software-short trades were caught in the same unwind, not just Situational Awareness alone.

What this means for AI-adjacent investing

The collapse is a pointed case study in what happens when a highly-leveraged, highly-concentrated thesis meets a sudden reversal in relative performance between two correlated trades — the same structural risk that has ended prior leveraged funds, just compressed into six weeks instead of months. It also raises a broader question the AI investing world will likely keep returning to: how much of the AI infrastructure buildout's stock performance reflects durable demand versus a crowded trade that multiple large funds were running simultaneously, amplifying moves in both directions. Aschenbrenner's decision to retain the fund's Anthropic stake while liquidating everything else signals he's drawing a distinction between direct AI-lab exposure and the infrastructure-adjacent public equities that got hit hardest in the reversal.

As reported by CNBC, with additional detail from Tech Times.