From $45B Fantasy to a Citadel Fire Sale: How Aschenbrenner's AI Fund Crashed in Days

Margin calls, not thesis debates, drove the collapse

This was a liquidity event first and an AI-thesis event second. Situational Awareness went from $45 billion in assets earlier this month to around $10 billion after Aschenbrenner was forced to unwind public positions to Ken Griffin's Citadel at a discount. What unfolded was less a slow repricing of an AI theme than a rapid forced sale.

Aschenbrenner's authority did not come from a long public investing record. It came from vision. Two years ago, he wrote a 165-page essay that helped establish him as a prominent voice in Silicon Valley's AI conversation. That context matters. When a manager is seen mainly as a prophet, investors can become more tolerant of risk than the structure actually justifies.

Once the AI trade worked, recency bias took hold. Strong performance made the portfolio look like proof of genius; when semiconductor stocks reversed, the same investors overshot in the other direction and treated him as if he had no edge at all. The deeper issue, though, was structural rather than ideological.

The mechanics accelerated the collapse. The fund posted large losses from declines in AI infrastructure investments such as SK Hynix, and all of his leveraged stock bets were offloaded in the unwind. A leverage problem can usually be argued about. A margin call usually cannot.

How a vision became a leveraged trade

The 400% leverage figure changed the story

The key number was not only the headline drawdown. It was reported leverage of up to 400%. That is the detail that turned a bold AI thesis into a fragile structure. When losses mounted, the market stopped reading the fund as a portfolio and started reading it as a source of distressed inventory.

Momentum invited copying, and copying invited crowding

Aschenbrenner's fame helped attract attention, but momentum did the rest. After a run of outperformance, investors began treating the fund less as a thesis and more as a shortcut. In that kind of setup, losses are often reframed as temporary setbacks and margin pressure is read as bad timing rather than bad structure.

• The bias chain: early wins boost confidence → confidence attracts copycats → copycats reinforce belief → belief masks fragility.

That is why leverage can stay elevated long after a more conventional risk review would call for de-risking.

The unwind began as a liquidity squeeze

The forced selling did not start with a valuation debate. It started with funding pressure. Prime brokers were rushing to raise cash to meet margin requirements, and the pressure hit hardest in the fund's most visible positions, including AI infrastructure names such as SK Hynix. At the same time, short positions in software companies such as Adobe moved sharply against it, so the squeeze was not confined to one part of the book.

That changed the dynamic from disagreement to distress.

• Before the break: investors assumed Aschenbrenner knew where AI capex was heading.
• During the break: every seller looked like someone exiting the same crowded trade.
• After the break: price reflected who had to sell first, not who had the better view.

The outcome was a fire sale. All of his leveraged stock bets were offloaded to Citadel at a discount, including SK Hynix and CoreWeave. Social media reaction captured the speed of the reversal in sentiment, with observers mocking him and others describing the event as a liquidation of one of the top AI funds.

Why the episode matters beyond one fund

The lesson is mechanical, not moral. A popular story can remain popular even as the positions behind it become unsound. When leverage is high and price action has drawn copiers into the trade, a portfolio can move from hero worship to forced selling before valuation ever gets a fair chance to catch up.

That is why the more useful question now is not who still believes the AI thesis. It is which forced sellers are still in the system, and what that means for the stocks they were pressured into dumping.