From Silicon Valley Manifesto to a $45 Billion Fund
Two years ago, Leopold Aschenbrenner was best known for a sprawling essay, not a trading book. The former OpenAI researcher — who had graduated with honours from Columbia University at just 19 and later joined OpenAI's superalignment team — published a 165-page manifesto titled "Situational Awareness" in June 2024, arguing that the world was underestimating how fast artificial general intelligence was approaching. The essay became required reading across Silicon Valley.
Aschenbrenner then did something less conventional for an AI researcher: he turned that thesis into a hedge fund, naming it Situational Awareness after the essay itself. The fund's core bet was straightforward in concept — that the AI infrastructure buildout would keep driving demand for chips, memory, data centres, and power, and that concentrated, leveraged exposure to this theme would generate outsized returns.
For a long stretch, it worked spectacularly. The fund reportedly finished 2025 up around 2,000%, and returned a further 439% after fees through the end of June 2026. Assets under management swelled to as much as $45 billion by early July 2026 — remarkable for a fund reportedly run by a skeleton crew of just eight employees, only four of them on the investment side.
What the Fund Actually Held
Regulatory filings show that Situational Awareness's largest publicly disclosed holdings included Nebius Group, SanDisk, Micron, CoreWeave, and SK Hynix — companies squarely in the AI infrastructure supply chain (chips, memory, and cloud compute). Roughly two-thirds of the fund's assets were held in public equities, both long and short. On the short side, the fund had also placed bets against software companies like Adobe, on the thesis that AI tools would squeeze traditional software firms' market position.
The fund additionally held a stake in Anthropic, the privately held AI company — a position reportedly separate from the public portfolio that was later sold off.
What Went Wrong
The unwind was triggered by a sharp downturn in AI infrastructure stocks. Several of the fund's largest holdings — Nebius, SanDisk, Micron, and CoreWeave among them — reportedly lost more than 35% of their market value within a single month. The sell-off reportedly coincided with SK Hynix's US IPO, and a broader unwinding of leveraged positions across South Korean stocks added further pressure. Meanwhile, the short bet against Adobe moved against the fund as well, with Adobe's stock rising more than 30% over the same period — the opposite of what the position needed.
Compounding the losses was leverage reportedly as high as 400%. With that much borrowed money amplifying the downside, falling stock prices quickly triggered margin calls from the fund's prime brokers, reported to include Bank of America, Goldman Sachs, and JPMorgan Chase, who began working with the fund to either raise cash or unwind positions in an orderly manner.
The Forced Sale to Citadel
As losses mounted, Aschenbrenner reportedly approached existing investors and lenders for fresh capital, with some investors even offered the chance to buy portfolio assets directly. That wasn't enough. The fund's public stock portfolio was ultimately sold in one large block trade to Ken Griffin's Citadel, with rival firm Millennium also reported to have bid for the book.
The scale of the reversal is striking: from $45 billion at the start of July to roughly $10 billion by the time the dust settled — though the precise size of the drawdown hasn't been officially disclosed, and the current net asset value remains unconfirmed as reporting continues to develop. It's also unclear, based on current reporting, whether the fund secured any fresh capital from lenders or investors after its initial approaches, or what becomes of its remaining Anthropic stake.
Why This Story Has Captured So Much Attention
Part of what makes this episode notable is the contrast: a researcher who built his public reputation on forecasting the future of AI with unusual clarity ran into one of the most conventional risks in markets — concentrated, leveraged exposure meeting a sudden reversal in sentiment. Some market commentary has framed the broader backdrop as a convergence of financing uncertainty, rising corporate AI capital expenditure, and pressure on big technology companies' free cash flow — three themes that map closely onto the exact thesis Situational Awareness was built around, which may help explain why the fund faced outsized pain once sentiment shifted.
The episode has also drawn renewed scrutiny of Aschenbrenner's background, including his limited traditional money-management experience before launching the fund and his past association with FTX, the collapsed crypto exchange.
The Investor Takeaway
Whatever one thinks of the underlying AI thesis, the mechanics of this collapse are a familiar cautionary tale in markets:
- Concentration risk cuts both ways. A small number of high-conviction positions can deliver spectacular returns on the way up — and equally spectacular losses on the way down, especially when several of those positions are correlated (as AI infrastructure stocks tend to be).
- Leverage amplifies both outcomes. Reported leverage of up to 400% meant that even a moderate correction in the underlying stocks could — and did — wipe out a large share of the fund's value once margin calls forced selling at unfavourable prices.
- A strong long-term thesis doesn't protect against short-term timing risk. Even if the broader AI buildout narrative proves directionally correct over time, a fund built on concentrated, leveraged bets can still be forced into a fire sale by a sharp, near-term move against it.
Popular Blogs
