First, the death certificate. This fund did not die because of a bad thesis. It died because of leverage. In just six trading days, one of the world’s best-performing funds lost control of all its public positions, which were sold off entirely before the market even opened. Dramatic collapses usually require a villain, but this event needs no villain—just one of the oldest numbers in finance. To understand that number, we must first trace what happened chronologically.
Who Was the Deceased?
Leopold Aschenbrenner. 24 years old, German, former OpenAI researcher. He became famous in Silicon Valley in 2024 with an AI manifesto titled “Situational Awareness.” That memo—arguing AGI is coming soon—became a defining investment thesis of our era, and he founded a hedge fund of the same name in July 2024. Tech heavyweights like former GitHub CEO Nat Friedman and Collison brothers of Stripe backed him. At the same time, criticism followed: his lack of formal money management experience and past ties to FTX came under scrutiny.
Judging by performance alone, the criticism seemed irrelevant. Starting with around $225 million, the fund generated cumulative returns exceeding 1,000% since inception, rising 439% net in the first half of 2026 alone. Assets under management once swelled to roughly $45 billion (figures vary across reports, with early estimates around $20 billion). A team of about 12 people was managing capital at a scale that takes decades to build, doing so in just two years. Most of it was concentrated in stocks tied to the AI data center boom.
Six Days of Collapse
Mid-July — The Peak. In a first-half investor letter reported by the Financial Times, the fund touted a 439% net gain. The portfolio was heavily long on AI infrastructure and semiconductors. Names tied to NVIDIA’s supply chain and data centers—Nebius, SanDisk, Micron, CoreWeave, SK Hynix, IREN, and others—were paired on the short side against software stocks like Adobe. Public equities made up about two-thirds of the book, with the rest in private holdings.
Late July — The Cracks. A sharp sell-off in AI infrastructure in July hit the book head-on. The Philadelphia Semiconductor Index fell 28.6% from its June 22 peak, and the Morgan Stanley Momentum TMT Index collapsed 53.5%. The fund’s core long positions dropped 40% to 50% in a single month. What proved brutal was the short side: the software stocks held short rallied sharply. The worst-case combination unfolded: getting hit simultaneously on both longs and shorts.
July 29 — Peak Margin Calls. Prime brokers Goldman Sachs, JPMorgan, and Bank of America issued margin calls. Facing a roughly 4x leveraged portfolio, there were only two options: raise cash or sell everything. Aschenbrenner frantically called institutions across Wall Street. Interestingly, just six days prior, he had urged investors to treat the July dip as “the most attractive opportunity since early 2025,” requesting capital calls by August 1. That capital never arrived.
Thursday, July 30, Before Dawn — Liquidation. Before the opening bell, the fund transferred its entire public equity portfolio—longs and shorts alike—to Ken Griffin’s Citadel in a single block trade. The sale price was discounted roughly 10% against market value. Reports put the block size at between $10 billion and $16 billion. With that, the fund, once valued at around $45 billion, shrank to roughly $10 billion, logging a 67% loss for the month of July alone. It took just six trading days for the world’s top-performing fund to become a forced seller of everything.
[A Quick Sidebar] The Math of 4x Leverage
Why did it happen so fast? The answer is simple arithmetic. Applying 4x leverage to 1 unit of equity yields a total position size of 4. In this scenario, if the position declines by just 25%, the loss equals 100% of the equity. In practice, long before reaching that point, margin calls arrive at the maintenance threshold set by brokers. Margin calls then force sales at the worst possible moment—compelling positions to be liquidated at bottom-barrel prices. In a bull market, leverage multiplies gains; in a downturn, it accelerates bankruptcy at the exact same multiplier. Whether a thesis is right or wrong, this math makes no exceptions.
A Brutal Twist: He Wasn’t Wrong
This is the most bitter chapter of the story. In the days immediately following the block deal transfer to Citadel, the very stocks that were forcibly liquidated rebounded across the board. CoreWeave surged 41% from its July 29 low, IREN rose 36%, and SanDisk jumped 27%. In other words, his stock picks were not fundamentally wrong. He was simply early and over-leveraged—and the market punishes “being early with leverage” exactly the same as being dead wrong.
Some analysts note that had he used no leverage or a moderate amount, he would be smiling with top 1% performance for the year. Forced liquidation essentially dragged him out of a correct position at the worst possible price. Aschenbrenner stated he takes “full responsibility” for the losses, calling it a “costly wound, but a valuable lesson.” What remains of the fund is a private portfolio centered on a roughly $5 billion stake in Anthropic, which will operate as a private strategy going forward.
Lessons for Individual Investors
Why this story isn’t just about hedge funds is because the lessons apply equally regardless of scale.
First, leverage does not care about the size of your conviction. No matter how correct the thesis, a position built on borrowed money can be forcibly closed during even a brief move against you. If you are ejected from the game before your correctness materializes, being right means nothing.
Second, concentration amplifies both the upside and the downside. A book all-in on a single theme posts top returns when that theme rises, but the moment that theme drops 30%, it collapses with equal force.
Third, avoid the curse of forced selling. The worst losses stem not from flawed judgment, but from being forced to sell at the worst possible timing. Only those who can endure capture the rebound. What allows you to endure, ultimately, is a position that uses no leverage or only a manageable amount of it.
The author of the world’s premier AI thesis tripped over the oldest number in finance. The name of that number is leverage.
⚠️ This article is an event commentary based on public reporting and does not constitute investment advice for specific stocks or strategies. Figures cited (AUM, deal size, returns, etc.) vary across reports and may be updated; please verify precise details directly from original sources (FT, WSJ, CNBC, Bloomberg, etc.).
Frequently Asked Questions
How does the math of 4x leverage work?
Using 4x leverage on 1 unit of equity creates a total position size of 4. In this case, if the position declines by just 25%, the loss equals 100% of equity, and margin calls arrive even earlier at the maintenance threshold set by brokers.
Was his investment judgment inherently wrong?
No. In the days following the block trade, CoreWeave rebounded 41% from its low, IREN 36%, and SanDisk 27%. He simply had to exit the game before his thesis played out.
What lesson can individual investors take away?
That leverage is indifferent to your degree of conviction. No matter how solid the thesis, positions magnified by borrowed money can be forcibly liquidated during even a temporary drawdown.
Leave a Reply