The fund that lost 67% in July is still up 80% for the year
Leopold Aschenbrenner had three of them.
His fund, Situational Awareness, collapsed last week under the weight of illiquidity, leverage and concentration.
No wonder failed managers find it easy to raise fresh capital.
A fund that lost 67% in the month of July is still up 80% for the year. Both numbers belong to Situational Awareness, the hedge fund built around AI infrastructure names, and both come from the letter it sent its investors.
'We let you down this month': Aschenbrenner's letter to investors (Business Insider, 2026-07-31)In the letter Aschenbrenner wrote that the fund "came closer to permanent capital impairment than is acceptable to us." The same letter says a single block transaction stripped out all of the borrowed money, and that every short position was closed.
Why this fund and not another
Marc Rubinstein explains the episode through three rules Steve Cohen set out in 2021. Liquidity, leverage, concentration. One of them you may survive, two is trouble, and all three at once is whistling past the graveyard. Take them one at a time. Liquidity is whether you can sell without cutting the price much. Leverage is how much borrowed money sits on top of your own, and concentration is how few names hold all of it. When the three overlap, a small drop forces a sale, the sale pushes the price down further, and because it all sits in a handful of names the loss lands in one place. Rubinstein's account of July runs like this. The 29 US-listed long positions fell 21% on average over the month, while shorts such as Adobe rose 26%. With three to four times leverage on top of that, the fund ended the month down 67%. Strip out unlisted holdings such as Anthropic and he puts the public book down around 85%.
Quartz reports the fund ran borrowed money as high as 400%, and that SK hynix, CoreWeave, Nebius, SanDisk and Micron each fell more than 35% in July. The size of those falls matters less than the fact that they were not five separate companies so much as five ways of owning the same story. That is what concentration means here.
How did the market find out
The letter also describes how the market worked out the fund's position. "As these moves proceeded, we started to see increasingly adverse trading in names publicly associated with us." Aschenbrenner compared it to a bank run: vulnerability begetting more vulnerability. Where that list of names came from is not spelled out. In the US, an institution managing more than $100 million in listed equities has to disclose its holdings every quarter on a form called the 13F.
The July book cannot be verified from disclosure. The list the market called "this fund's names" came from reporting and inference, not from a filing.
Retrieved 2026-08-03 · SEC EDGAR (Situational Awareness LP, CIK 0002045724)
Without a confirmed list, the market is likely to guess broadly. The broader the guess, the wider the set of names that gets sold ahead of the seller.
Why blown-up managers raise money again
What Rubinstein's piece is really about is not the collapse but what comes after it. The trader John Arnold used to hold that "the optimal number of past blow ups was one" when hiring. Someone who has been broken once, in that view, beats someone who never has. The record points the same way. Citadel's Ken Griffin was down 55% in 2008 and TCI's Chris Hohn 43% in the same year, and both survived. After LTCM failed in 1998, the fund John Meriwether started next grew to $2.7 billion.
No one hands money back to a manager who lost 67% in a month. A hit like that ends a career.
The record shows managers who blew up large raised fresh capital with relative ease. The market counts the experience as something closer to a credential.
The split is whether you read the same event as a failure or as tuition paid. The precedent Rubinstein calls the closest fit, though, is not a comeback story. It is Amaranth Advisors, twenty years ago. That fund was also shopped to several buyers before Citadel took it on, and its star trader was two years into the job. A month before the collapse, the Wall Street Journal sent a reporter to his home for a profile. Aschenbrenner was profiled by the same paper in June. The fund that Amaranth's trader started next drew $800 million of commitments and never got off the ground.
So what settles it
The question comes down to one thing. Whether this was one manager's accident, or a structural problem with all the money crowded into AI infrastructure names. What settles it is the record without leverage. The fund says it will now run its public book on a fully-paid-for basis. If the same names hold up from here, July's fall was made by borrowed money. If they slide anyway, the problem was the selection. 67% and 80% are the same fund in the same year. Losing two thirds of the capital in a month and still sitting above where the year started tells you how much it had been making until then. Borrowed money multiplies the rate of gain and the rate of loss by the same factor.
- Situational Awareness, a hedge fund concentrated in AI infrastructure names, lost 67% in the month of July. The same fund is still up 80% for the year.
- On the surface it reads as a verdict on AI investing. Marc Rubinstein reads it instead as what happens when illiquidity, leverage and concentration land in one place.
- What his piece is really about is not the collapse but what follows it. Managers who have blown up once have a long record of raising money again.