AI Hedge Fund Situational Awareness Sells Public Portfolio, Holds $5B Anthropic Stake

Situational Awareness, the AI-focused hedge fund founded by former OpenAI researcher Leopold Aschenbrenner, has sold the bulk of its public stock portfolio to Ken Griffin's Citadel after suffering steep losses over the past month. The sale dropped the fund's total assets from roughly $20 billion to around $10 billion. The fund is not winding down entirely: it still holds a $5 billion private stake in Anthropic, which was last valued at $965 billion in May and is expected to go public as soon as October 2026.

What happened

Fact	Detail
Fund founder	Leopold Aschenbrenner, 25, former OpenAI researcher
Fund launched	2024
Peak AUM	Up to $45 billion
AUM after Citadel purchase	~$10 billion
Return through June 2026	439% (per Financial Times)
Anthropic stake value	$5 billion (per Bloomberg)
Anthropic last valuation	$965 billion (Series H, May 2026)
Expected Anthropic IPO	As soon as October 2026

Situational Awareness sold the majority of its public equities to Citadel, the hedge fund run by Ken Griffin, according to a Wall Street Journal report published July 30. The sale came after a sharp decline in AI infrastructure stocks over the past month hit the fund hard, with several of its largest positions falling more than 30%.

The stocks most affected included memory chip producers SK Hynix and Sandisk, clean energy company Bloom Energy, and neocloud provider Nebius Group. The losses were made worse by leverage, meaning the fund had borrowed money to amplify its bets. When prices dropped, those borrowed positions had to be unwound.

Who is Leopold Aschenbrenner?

Aschenbrenner enrolled at Columbia at age 15, graduated as valedictorian at 19, and joined OpenAI’s “superalignment” team in 2023. That team, focused on AI safety at scale, was led by OpenAI co-founder Ilya Sutskever and researcher Jan Leike. Aschenbrenner was dismissed from OpenAI a year later over what the company described as an improper disclosure of internal information.

He then published a widely-read essay arguing that scaling AI models would demand a massive build-up in semiconductors, compute, memory, and energy infrastructure. That thesis became the investment framework for Situational Awareness when he launched it in 2024, with no prior trading experience. Early backers included quant-trading firm Jane Street, Stripe co-founders Patrick and John Collison, and Meta executives Daniel Gross and Nat Friedman. The fund raised several hundred million dollars at launch.

Why did the fund run into trouble?

AI infrastructure equities sold off broadly as public market investors grew concerned that enormous capital expenditure by big tech companies was not converting into near-term revenue. You can see the broader investor unease around AI capex covered in our look at Google’s $205B capex and investor reaction. Situational Awareness was heavily concentrated in exactly those infrastructure names, and the leverage amplified the damage.

On July 24, Aschenbrenner wrote to investors calling the selloff one of the best buying opportunities since early last year, and invited clients to commit new capital starting August 1. According to Bloomberg, the appeal did not generate the commitments he had hoped for.

What does Citadel’s purchase signal?

Citadel has a history of buying assets from leveraged funds that are forced to sell. Notably, its portfolio already held some of the same AI infrastructure positions before the purchase, which suggests Griffin shares Aschenbrenner’s long-term view on the sector but has the balance sheet to wait out a prolonged downturn without being forced into a fire sale.

Why it matters

The Situational Awareness story is a clean illustration of a thesis risk that many AI investors are now facing: being right about a multi-year trend but wrong about the timing. The fund’s 439% return through June looked like vindication. The selloff in the following weeks showed how quickly leverage can turn a winning position into a crisis.

For anyone watching the Anthropic IPO closely, the $5 billion stake is now the fund’s most consequential holding. If Anthropic goes public in October at a valuation above $965 billion, the proceeds could meaningfully offset the public-market losses. The fund also retains private positions in chipmaker MatX and AI data center startup Fluidstack, which was reported to be in talks in April to raise a new round at an $18 billion valuation.

This also matters for the broader AI investment ecosystem. Situational Awareness attracted serious institutional backers based on a coherent, public thesis. Its distress signals that concentrated, leveraged bets on AI infrastructure are genuinely risky, even when the underlying thesis proves directionally correct. Businesses evaluating AI integration strategies should note the gap between long-run infrastructure trends and short-run market volatility.

Our take

Aschenbrenner built a fund on a thesis that is probably right. The world will need more chips, more power, and more compute. That does not mean the stocks of companies supplying those things go up in a straight line. Leverage compounds mistakes in timing the same way it compounds gains.

The Anthropic stake is the interesting variable here. A $5 billion position in a company valued at nearly $1 trillion, with a potential IPO months away, is a meaningful lifeline. Whether it is enough to rehabilitate the fund’s reputation with investors is a different question. The people who backed Aschenbrenner at launch were betting on a 25-year-old with no trading history and a bold thesis. Some of them will stick with that bet. Others will wait to see what the Anthropic IPO actually delivers.

For context on how Anthropic fits into the broader AI competitive picture, see our recent coverage of Claude Opus 5’s capabilities and what that means for the company’s commercial trajectory.

What to do about it

1. Watch the Anthropic IPO timeline. If it prices above the $965 billion Series H valuation, Situational Awareness gets a significant boost and it validates the private-market AI thesis.
2. Treat AI infrastructure stocks as volatile. The broad selloff in SK Hynix, Sandisk, and Bloom Energy shows that even companies central to AI build-out can drop 30%+ in a month.
3. If you are a business owner considering AI tools or infrastructure spend, separate the investment-market noise from operational decisions. The underlying technology is not going away because a hedge fund had a bad month.

The fund’s trajectory is a reminder that being early and correct on a technology trend is not the same as managing the financial exposure that comes with it.

Source: TechCrunch · AI