One of the most talked-about funds in AI is wagering again, barely a week after it nearly blew up. Situational Awareness, the firm founded by Leopold Aschenbrenner, has returned to investing with a $400M bet.
The return is remarkable given what just happened. In July the fund suffered a catastrophic reversal, losing roughly two-thirds of its value in a single month and being forced to dump its holdings.
Aschenbrenner is not a typical fund manager. A former OpenAI researcher, he left in 2024 and published a 165-page essay, also called “Situational Awareness,” arguing that artificial general intelligence could arrive far sooner than most expected.
He turned that thesis into a portfolio. Backed by roughly $225m from names including the Collison brothers, Nat Friedman and Daniel Gross, the fund bet on the “picks and shovels” of the AI boom rather than the model makers themselves.
The strategy worked spectacularly, for a while. The fund piled into data-centre operators, memory chipmakers and power suppliers, and its assets swelled past $20bn as those bets soared, with one measure putting its return at 439%.
The holdings were a map of the boom. They included the cloud provider CoreWeave, Nebius, SK Hynix, SanDisk, Bloom Energy and crypto miners repurposed for AI, the infrastructure everyone assumed would keep climbing.
Leverage turned confidence into fragility. The fund was reportedly running at roughly four-to-one leverage, which magnifies gains on the way up and losses just as violently on the way down.
Then the correction came. When AI infrastructure stocks fell sharply, the leverage triggered margin calls, and the fund was forced to liquidate nearly its entire public equity portfolio, reportedly selling to Citadel Advisors.
The speed was the shocking part. A fund that embodied the AI trade lost most of its value in weeks, a real-world demonstration of how quickly a concentrated, leveraged position can unravel.
The lesson was not about the thesis. Aschenbrenner may well be right that AI needs vast infrastructure; what undid the fund was risk management, the decision to express a correct idea with heavy leverage and little diversification.
The human story adds to the drama. On paper the collapse erased a fortune built in barely two years, a reminder that in leveraged markets, gains that look permanent can prove astonishingly temporary.
It was not an isolated wobble. AI-linked stocks have been jittery for months, with episodes like the rout that followed China’s Kimi K3 showing how fast sentiment can turn on the sector.
The infrastructure names the fund loved are especially exposed. CoreWeave, one of its holdings, has leaned heavily on debt, and its borrowing costs swing with the market’s mood about the durability of AI demand.
The blow-up feeds a louder debate. Skeptics point to comparisons with the dot-com bubble, arguing that valuations and concentration have run ahead of what the underlying businesses can yet support.
Regulators are watching the financial plumbing too. The Bank for International Settlements has warned that an AI bust could hit credit markets as hard as 2008, precisely because so much of the boom runs on borrowed money.
Against that backdrop, the $400m bet is a statement. It signals that Aschenbrenner still believes his core thesis, even after the market delivered a brutal lesson about how to express it.
There is a logic to buying after a crash. If the underlying demand for AI infrastructure is real, then a forced, leverage-driven sell-off can leave good assets cheap, and a fund with dry powder can pounce.
The risk is that the lesson has not been learned. Betting big again so soon, in the same volatile corner of the market, invites the question of whether anything has changed beyond the size of the position.
Whether it is conviction or defiance is the open question. Returning so soon suggests the fund sees the sell-off as an opportunity rather than a verdict, a wager that the AI build-out’s stumble is temporary.
Get the TNW newsletter
Get the most important tech news in your inbox each week.