Aschenbrenner's Fund Was Up 439% and Closed in Four Days

On Wall Street, the clearest sign that a fund is finished is its portfolio changing hands in one block. Nobody puts out a statement. Nobody says the words "we are closing." One morning, before the opening bell, everything simply moves to a single buyer.

That is what happened on the morning of July 30. Every listed position held by Leopold Aschenbrenner's Situational Awareness LP — the longs and the shorts — went to Ken Griffin's Citadel.

Six months earlier, the same fund had closed the first half of the year up 439% after fees.

From Manifesto to Fund: The Birth of a Thesis

Leopold's story had already hardened into legend in investment circles. After leaving OpenAI in 2024, he published a long essay titled "Situational Awareness" that rested on a single claim: artificial general intelligence is far closer than the consensus believes, and its arrival will set off the largest wave of physical infrastructure investment in history.

The power of that idea was that it turned an abstract forecast into a list of things you could actually buy. If AI accelerates, you need chips. If chips are to run, you need memory. If both have to sit somewhere, you need data centers. If data centers are to operate, you need power. What came out the other end was not a think piece. It was a shopping list.

Then he put his own money behind the list. The fund launched in late 2024 with roughly $225 million. The investor roster made clear who believed in the idea: Stripe founders Patrick and John Collison, Nat Friedman, Daniel Gross and — unusually for a fund of this type — Jane Street.

Inside of two years, assets passed $20 billion. According to a person who spoke to CNBC, the fund had grown to as much as $45 billion by early July. Leopold was 25 years old at the time.

Two Fragilities: Concentration and Leverage

A 439% return does not happen by accident. The only way to reach a number like that is not simply to be right, but to press very hard on the thing you are right about. Two decisions were made here at once.

First, almost everything sat in a handful of names. In the fund's disclosed portfolio, the top five positions accounted for more than three-quarters of the total: SK Hynix, Nebius, CoreWeave, Micron, SanDisk and Bloom Energy. Six companies in different industries doing genuinely different things — but all of them different expressions of the same sentence.

Second, he did it with borrowed money. The fund carried gross exposure of roughly four times its own capital: for every $1 of equity, $4 of stock in the market.

There was a third layer as well. At the same time, the fund was short software names — the "long the chip, short the software" trade, built on the idea that AI would compress software company earnings while blowing out demand for the infrastructure underneath it. It was the most talked-about trade of 2026.

The Arithmetic of Leverage

Understanding why this structure collapsed takes only one calculation.

The real issue is not the size of the loss. It is who owns the decision to sell. In an unlevered position you decide when to exit; even if the price halves you can choose to wait, because you owe nobody anything. When you buy with borrowed money, the broker makes that decision for you — and does so at precisely the worst price, because a bad price is the reason the margin call arrived in the first place.

July: The Month the Thesis Was Tested

In July, AI infrastructure stocks sold off, and the selling stopped looking like an ordinary correction.

What the bars show is that the decline did not hit the whole market. It hit exactly where the fund was standing. The Nasdaq 100 gave back around 10% from its high while the fund's core positions lost between 27% and 54%. Korea's stock market, home to SK Hynix, fell nearly 40%.

The charts below are live: they do not show the event itself, but where those stocks trade today.

In the same week, names the fund was short, such as Adobe, moved higher. Both sides of the book lost money simultaneously.

On a portfolio carried at four times leverage, a decline of 30%-plus in the core holdings is more than enough to consume the remaining equity. The three banks financing the fund — Goldman Sachs, JPMorgan Chase and Bank of America — issued margin calls.

The Fund That Ended in Four Days

What followed, as reported by the Financial Times and CNBC, ran in the order these collapses always run: slowly at first, then all at once.

The third line is the moment the story actually ends. Once word gets out that a fund's portfolio is being shopped around by its banks, the price starts moving against it: everyone knows those shares have to be sold, and everyone bids accordingly.

The transaction price was not disclosed. Figures such as 40 cents on the dollar have circulated on social media, but none of them trace back to a verified source.

For comparison, the broader market:

Why the Buyer Was Citadel

Griffin's firm has played this role many times before. Standing on the other side of a forced seller is a business model Citadel has repeated for years. The logic is simple: the gap between the fundamental value of the positions and the seller's time pressure is the buyer's profit.

That leaves two readings of the trade, and both are defensible:

The only certainty sits between the two: the largest forced seller in the market is now gone.

What Is Left

Situational Awareness is not shutting down. Its private holdings remain, and the largest of them is a stake in Anthropic that the Financial Times values at roughly $5 billion.

That this survived is not a coincidence, and the reason is technical:

Prime brokers calculate collateral against assets that carry a price every day. A private company has no daily mark, and therefore never enters the margin calculation.

What the borrowed money killed was precisely the listed portion of the book. It repriced every day, so it entered the collateral math every day. Nothing else did.

The result is a curious structure: Situational Awareness is no longer a levered public equity fund but an investment firm holding mostly Anthropic stock. The position it believed in most, and held with the longest horizon, is still standing. The fragility that threatened it is gone.

A spokesperson for the fund denied reports that the Anthropic stake had been put up for sale. CNBC had reported that the fund was attempting to sell those private investments. On this point, the sources conflict.

The Lesson: You Do Not Choose When You Sell

Leopold's idea was not disproven this week. Whether he is right about demand for AI infrastructure is a question the next several years will answer.

What collapsed was not the thesis. It was the way the thesis was financed.

The second lesson applies even to investors who never borrow a dollar: AI stocks may represent companies doing very different things, but the market increasingly buys and sells them as a single stock.

When a memory manufacturer, a server rental company and an energy firm all fall on the same day, in the same direction, by a similar magnitude, your portfolio is not as diversified as you think. You may have bought six separate companies, but what you actually own is one sentence.

Leopold wrote the most widely read investment text of the AI era. He also sat the most expensive exam that same idea has faced.

This article is based on reporting by the Financial Times, CNBC, The Wall Street Journal and Bloomberg as of July 30, 2026. Much of that reporting relies on unnamed sources, and no party has issued an official statement. Figures such as the fund's assets under management and the transaction price differ between sources.