The AI Trade's Loudest Believer Just Got a 92% Haircut
Leopold Aschenbrenner's Situational Awareness fund went from $20.24 billion in reported holdings to $1.58 billion. Most of what's left is a fresh billion-dollar options bet on AI that has already slid. The thesis and the trade were never the same thing.
The number that doesn't blink
$20.24 billion in June. $1.58 billion now.
That's not a drawdown. That's a deletion, and it belongs to Leopold Aschenbrenner, the former OpenAI researcher who turned a viral essay about superintelligence into an actual hedge fund with actual outside money in it.
The fund is called Situational Awareness, same as the 2024 essay that made him briefly famous in AI circles. The pitch was straightforward enough. If agents and AGI are coming, own the companies that supply the compute, the power, and the chips. Proponents of that thesis had a very good 2024.
Then the bill arrived.
What the filing actually says
Aschenbrenner's latest 13F shows $1.58 billion in US stocks and options. The June filing showed $20.24 billion. That's a 92% collapse in reported value over a few short months.
Here's the part that matters more than the headline. Most of what's left is a fresh bet. Roughly $1 billion of it sits in AI options positions, and those have slid since the filing date. So the fund didn't just shrink. It got smaller and more concentrated at the exact same time.
Admittedly, 13F filings are a lousy window into any fund. They capture long US-listed positions and certain options. Shorts, cash, offshore holdings, and anything private stay invisible. A 92% drop in reported longs tells you about a book, not necessarily about an entire portfolio.
But let's not be too generous here. When a filing goes from $20 billion to $1.6 billion, something broke. Options amplify moves in both directions. That's the whole point of them, and that's the whole problem with them.
Conviction isn't a strategy
The question worth asking: if the most credentialed believers in the AI trade can be down 92% on it, what's the actual edge?
My answer is uncomfortable for the true believers. Aschenbrenner's thesis about compute demand is probably right. The trade built on top of that thesis was wrong, or at minimum badly sized. Those are different things, and the market just spent a quarter separating them in public.
Look at the pattern. The narrative that AI infrastructure names only go up has been tested repeatedly this year. Concentration plus options plus conviction equals a portfolio with no floor underneath it. You don't need a crash in the underlying companies to lose most of your money. You need a drawdown and a stack of borrowed exposure.
I'm not entirely convinced the fund is as broken as the number suggests. Granted, the disclosure rules have real holes in them. And a manager willing to reload with a billion dollars in options is either making the trade of the decade or confirming a pattern. History suggests otherwise on the second one.
Who wins? The skeptics, mostly. Not because AI is a bubble, but because "AI is real" and "this fund is safe" were never the same claim, even though a lot of people treated them that way. Who loses? Anyone who read the essay and assumed the returns would follow the argument.
What to watch
Two things. First, the next 13F. If the options position gets rolled forward, conviction is intact and risk tolerance hasn't budged an inch. If it gets cut, you're watching a manager recalibrate in public, which is a different and more interesting story.
Second, watch the language. Aschenbrenner built his reputation on being early and loud. The real test isn't whether he was right about AGI. It's whether he can run money when the trade stops agreeing with him.
Time will tell, though the paperwork will probably tell us first.
Related Articles
Explore More
Key Terms Explained
Taking a position that offsets potential losses in another investment.
Contracts giving the right, but not obligation, to buy (call) or sell (put) an asset at a set price before expiration.
Your collection of investments across different assets.
Shares representing partial ownership in a company.