25-Year-Old AI 'Prophet' Leopold Aschenbrenner’s Fund Collapses in One Month

Written by Xiao Bing

Foreseeing leads to foresight; foresight leads to death.

July 30, 2026 is estimated to be the most unforgettable day of Leopold Aschenbrenner's life.

A month ago, his fund, Situational Awareness LP, was Wall Street’s brightest new star. According to investor letters seen by the FT, the fund delivered a net return of 439% in the first half of this year, making it one of the top-performing hedge funds of 2026, with total assets under management peaking at $45 billion in early July.

On July 24, Leopold Aschenbrenner wrote a letter to his investors, acknowledging that his fund had not escaped the recent sell-off in AI stocks. After stating this, he shifted focus elsewhere: it was one of the most attractive buying windows since early 2025. The letter ended with a postscript inviting existing investors to add capital before August 1.

That money never arrived.

Before the open on July 30, three prime brokers—Bank of America, Goldman Sachs, and JPMorgan—simultaneously issued margin calls to Situational Awareness. The fund’s publicly traded stock portfolio was leveraged approximately fourfold. That morning, the entire public stock book, including both long and short positions, was transferred in a single block trade to Ken Griffin’s Citadel.

From being fired from OpenAI’s Superalignment team, to writing a 165-page paper that shook Silicon Valley and Washington, to founding a fund that grew to $45 billion, and finally being forced to liquidate.

He completed the entire process in less than two years.

He is 25 years old this year. Five years ago, he was an undergraduate student at Columbia University.

Leopold Aschenbrenner was born in Germany to parents who are both physicians and attended the Kennedy School in Berlin. He graduated from Columbia University at age 19 with dual degrees in economics and mathematical statistics, as valedictorian.

His legendary journey began in April 2024.

At the time, he worked on OpenAI’s Superalignment Team under Jan Leike and Ilya Sutskever, researching how to ensure future superintelligent AI would not pose a threat to humanity. But he increasingly became concerned that OpenAI’s safety measures were flawed. He wrote a memo to the board highlighting serious information security vulnerabilities, particularly the risk of foreign industrial espionage.

He was then fired, with OpenAI citing "information leakage," which he denied.

Two months after leaving OpenAI, he wrote the lengthy article "Situational Awareness," which made the entire AI community remember his name.

Two months later, in June 2024, he published a 165-page paper titled "Situational Awareness: The Decade Ahead."

This paper is going viral simultaneously in Silicon Valley, Washington, and Wall Street. The core argument is simple: AGI (Artificial General Intelligence) will arrive around 2027, and may recursively self-improve into superintelligence between 2028 and 2030. The bottleneck is not algorithms, but physical infrastructure—electricity, transformers, transmission lines, cooling systems, and data center land.

U.S. President’s daughter Ivanka Trump publicly praised the paper. Stripe founders Patrick and John Collison, former GitHub CEO Nat Friedman, and former Y Combinator partner Daniel Gross pooled $225 million to help him launch a hedge fund, named Situational Awareness LP—after his paper.

He also invested nearly all of his personal net worth.

From the end of 2025 to the first half of 2026, Leopold’s fund soared like a rocket.

Starting with $2.25 billion in initial capital, assets under management peaked at $45 billion by early July 2026, expanding 200-fold.

He outlined his positioning logic in a paper: if the bottleneck is electricity, then those who control electricity, land, and cooling infrastructure are the mispriced AI infrastructure.

This reasoning led him to a place most AI investors wouldn’t go: Bitcoin miners.

After the 2024 April halving, profit margins for traditional mining have been continuously squeezed, prompting many mining companies to repurpose their data centers into托管 facilities for AI and high-performance computing. Their balance sheets hold ready-made power contracts, land, and facilities, yet the market still values them using mining-era frameworks.

As of March 31, 2026, the fund’s U.S. equity holdings were valued at $5.52 billion across 29 securities. Miners and their data center operations totaled approximately $1.38 billion, accounting for about a quarter of the portfolio, with Core Scientific being the largest position at $418.7 million, followed by IREN at $328.6 million and Applied Digital at $278 million. The top five holdings—CoreWeave, Bloom Energy, Intel, Lumentum, and Core Scientific—collectively represented approximately 68% of the equity portfolio.

These companies' stock prices surged amid the AI infrastructure boom.

The other half of the ledger is short.

The same 13F disclosed put options with a notional value of approximately $8.46 billion on VanEck Semiconductor ETF, NVIDIA, Oracle, Broadcom, AMD, Micron, and TSMC, accounting for 62% of the total $13.68 billion reported in the 13F filing.

Most people read this as: He is shorting chips.

The options values in 13F filings represent the underlying securities' value, not the actual amount spent; behind the $8.46 billion notional size, the real capital outlay may be only a few percent. 13F filings also do not disclose strike prices, expiration dates, or hedging relationships—just reviewing this document, no one can determine whether a put option is a directional short position or insurance purchased for a long position.

The movements before a crash are more telling.

According to unusual_whales' analysis, the latest 13F filing shows he is increasing positions in semiconductor stocks, establishing new positions in TSMC, Intel, ASML, NVIDIA, and Corning, while adding call options on Micron, TSMC, and SanDisk. He has reduced holdings in Core Scientific and Bloom Energy, with his true short positions concentrated in software stocks.

By the end of June 2026, this “multi-AI infrastructure, no software stocks” fund delivered a net return of 439%. Its holdings became Wall Street’s “AI infrastructure textbook,” with retail investors chasing its 13F filings and analysts treating its positions as a barometer. An analyst at a major technology fund told Fortune: “When the market gets tense, colleagues will say, ‘Leopold says it’s fine,’ and everyone stops selling.”

He became the prophet of the AI era, a 25-year-old prophet.

One month, the double strangulation has arrived.

In July, his long positions declined. SK Hynix fell nearly 40% from its June high, NVIDIA dropped below $200, and Micron and CoreWeave suffered severe losses. Meanwhile, his short positions rose, with software stocks like Adobe surging sharply on expectations of AI application adoption.

Long and short positions were simultaneously wiped out amid leveraged liquidations, causing massive losses for Leopold’s funds. Three major prime brokers stepped in to help meet margin requirements, but margin calls continued like an unstoppable avalanche.

He has tried all other methods.

The letter dated July 24 invited existing investors to increase their funding. He has spoken with lenders and also explored selling portions of the holdings directly to individual investors. Both Millennium and Jane Street evaluated the book but declined to participate; Jane Street is itself an investor in the fund.

Ultimately, all publicly traded shares were liquidated, with only one buyer: Citadel, Ken Griffin’s hedge fund. The price was not disclosed, but the market widely speculated it was a "fire sale" price—forced sales like these rarely fetch a good price.

It took only one month to go from a 439% yield to liquidating the stock.

At this point, it was just a story of a brilliant fund manager crushed by leverage and market volatility. But a timeline began circulating on X, making the entire event far more complicated.

The timeline is as follows:

On July 27, Bloomberg reported that Frank Flight, Head of Macro Strategy at Citadel Securities, predicted in a client report that the Federal Reserve would unexpectedly raise interest rates by 25 basis points at its meeting on July 29. He is one of the few major institutional strategists to make such a forecast, with the probability of a rate hike on Polymarket at only 24.3%.

Frank Flight’s argument is not about any specific inflation data, but about credibility. Kevin Warsh (Trump’s nominated new Fed chair) needs to solidify policy credibility through an unexpected rate hike and decisively bring an end to the era of forward guidance.

From July 28 to July 29, this prediction rapidly spread through the trading community. Rumors of an interest rate hike acted like a fuse, igniting an already tense market sentiment. Tech stocks led the plunge, with NVIDIA, AMD, TSMC, and SK Hynix all suffering severe losses. Leopold’s fund, already on the brink due to dual pressures, faced its final collapse as panic selling triggered by the rate hike rumors completely breached its margin requirements.

On July 30, Citadel acquired the entire publicly traded stock portfolio liquidated by Leopold.

The Federal Reserve ultimately did not raise interest rates on July 29. But the market no longer cares—the damage from the rumors had already been done, and Leopold’s position was liquidated.

This timeline has been reshared countless times on X. The core allegation is that the head of macro strategy at Citadel Securities (the market-making division) posted a rare prediction of an interest rate hike two days before the Fed’s decision; this rumor triggered a sell-off in tech stocks, leading to Leopold’s liquidation, after which Citadel (the hedge fund division) acquired Leopold’s positions at discounted prices.

Citadel Securities and Citadel hedge fund are two separate legal entities, sharing founder Ken Griffin but operating independently. Citadel Securities is a market maker that handles approximately one-quarter of all U.S. stock trading volume and 35% of retail orders, giving it visibility into real-time order flow. As the head of macro strategy at Citadel Securities, who are Frank Flight’s client reports directed toward? Can he see who in the market is being forced to sell? Could a single rumor trigger hundreds of billions in selling—and if so, is the precision of that rumor too coincidental?

These questions have no definitive answers. But discussions on X are growing increasingly heated. Some say this is a perfect hunt: first releasing signals of rate hikes to trigger market panic, forcing out leveraged positions, then buying at low prices. Others argue it’s merely coincidence—that Frank Flight’s rate hike prediction was based on his independent analysis of Warsh’s policies, and Leopold’s liquidation resulted from his own long-short strategy and excessive leverage, with the rumors simply accelerating an event that was already inevitable.

What is the truth? Coincidence, a hunt, or the harsh logic of the market itself?

Still a mystery.

After the news of Leopold's liquidation spread, the market reaction was interesting.

Stocks considered most likely to crash surged across the board: Nebius rose 26.3%, IREN increased by approximately 26%, CoreWeave jumped 21.8%, and Core Scientific gained 21%... The entire AI sector, including storage, staged a strong rebound.

CNBC host Jim Cramer provided his explanation before the open: clearing out those who are forced to sell will reveal the bottom. He added that there are not just one or two such individuals—there are many other leveraged positions piled up on the same stocks.

Market strategist Dan Niles came close, referring to this wave of forced liquidations as a cleanup and noting that prime brokers handled it with considerable care to avoid a repeat of Archegos. In 2021, Bill Hwang’s family office collapsed, causing losses of approximately $5.5 billion for Credit Suisse alone—a bank that was acquired by UBS in 2023. ZeroHedge simply dubbed the event “Archegos 2.0.”

Retail investors are celebrating on X: "The prophet has fallen; the bull market has begun."

His liquidation became a signal that Wall Street believed AI had hit bottom.

It’s ironic. He was once the most ardent evangelist of the AI narrative, yet ended up as its most tragic casualty.

But Leopold has not yet fallen, and the fund has not closed.

Only the public stock portfolio was liquidated. Situational Awareness retained all of its private holdings, the largest of which is its equity stake in Anthropic, valued at approximately $5 billion according to FT. The company will continue operating as a private investment entity. Reports suggested it is selling its Anthropic shares, but a fund spokesperson denied this to CNBC.

One detail is that Leopold’s fiancée, Avital Balwit, is the chief of staff to Dario Amodei, CEO of Anthropic.

Leveraging Anthropic, the prodigy Leopold still has the potential to return to his peak.

Looking back, the name of his fund, "Situational Awareness," has become the greatest irony.

He has situational awareness, but lacks market awareness. Or rather, he has situational awareness, but no awareness of margin calls.

Long-term correct, short-term zeroed out—in Wall Street, this is called dying right.