Situational Awareness LP Exits Its Entire Public-Equity Book After Steep AI-Rout Losses

UPDATE — JULY 30: The buyer is Ken Griffin’s Citadel, which acquired the bulk of Situational Awareness’s roughly $16 billion public-equity book in a single trade — rival Millennium also bid, per the Wall Street Journal. The unwind followed margin calls from the fund’s three prime brokers — Goldman Sachs, JPMorgan and Bank of America — as roughly 4x leverage turned the AI drawdown into a forced liquidation that ended a 1,000%+ return streak. Situational Awareness retains its ~$5 billion Anthropic stake and will operate going forward as a private investment vehicle. For scale: the Philadelphia Semiconductor Index fell 28.6% from its June 22 peak, and the Morgan Stanley Momentum TMT index dropped 53.5%. (Ken Griffin’s Citadel per FT/Bloomberg/WSJ.)

CNBC’s David Faber reports the ~$20–24B fund sold its entire public book — longs and shorts — to a single buyer after steep losses in this week’s AI selloff; the private holdings, including a reported Anthropic stake, remain untouched.

What Happened

CNBC’s David Faber reported on July 30 that Situational Awareness LP — Leopold Aschenbrenner’s AI-focused hedge fund, with roughly $20–24 billion in assets under management — has exited its entire public-equity book, both its long and short positions, through a single transaction with one undisclosed buyer. Sources cited by CNBC characterized the move as a forced unwind driven by steep losses in this week’s AI market selloff, and said the fund may be forced to liquidate further. The Financial Times and Bloomberg separately reported that the fund is seeking fresh capital. Those hedges matter from the start: “forced to unwind” and “may be forced to liquidate” are the sources’ characterizations, not statements from the fund itself, which has not publicly confirmed the sale or its terms.

Equally important is what this transaction does not cover. The fund’s private holdings — approximately one-third of its book, reported to include a significant stake in Anthropic — are not part of this public-equity sale and remain in place. And a single-buyer block trade covering an entire public portfolio can represent an orderly de-risking as readily as a distressed fire sale; the framing will sharpen as more is confirmed. This piece corrects our earlier reporting, which noted the fund was “raising capital into the rout” with net performance unknown: the picture is now unambiguous in direction — steep losses, and a full exit of the public book.

The fund’s largest disclosed public positions at the end of Q1 — Nebius, SanDisk, Micron, and CoreWeave — are each down more than 35% this month. Public equities constituted roughly two-thirds of the portfolio, making this exit a structural event, not a trim. The scale of the drawdown is striking against the backdrop of what was, through June, one of the strongest performance records in the market: up approximately 439% net, per CNBC’s sources. A steep drawdown from an extraordinary peak is not the same as a fund below its inception value — that distinction matters for reading what happens next.

The Structural Read

Read through the Map of AI Redrawn and Beyond NVIDIA’s Moat frameworks, and what happened at Situational Awareness reads as a precise case study in the reflexivity of the AI trade — now resolved on the downside, at least in its public-equity expression.

The fund’s thesis was structurally coherent: AGI-scale ambition would demand a vast, sustained expansion of AI infrastructure — GPU clusters, HBM memory, data centers, power generation. It went long the names that would benefit from that buildout: Nebius, CoreWeave, Bloom Energy. Its skepticism on the chipmakers was expressed through shorts. This is the trade that delivered a reported 439% net return through June. It is also the trade that collapsed when the KOSPI-led chip and memory selloff hit this week — because the “AI infrastructure” names and the “AI chip” names moved together in the rout, regardless of which side of the book they sat on.

This is the market-side mirror of the capex-versus-returns stress test the week delivered broadly. As we noted in our coverage of vendor-financed AI buildout dynamics, the financing structure of the AI infrastructure cycle contains embedded fragility: when confidence in near-term returns wavers, the reflexivity runs in reverse. The fund that was most sophisticated about the AI trade’s long-run logic was also maximally exposed when the market’s short-run confidence broke. Leverage amplified both the ascent and the descent.

Three Implications