Leopold Aschenbrenner’s hedge fund Situational Awareness has been forced to sell its entire public stock portfolio to Ken Griffin’s Citadel after a margin call wiped out its leveraged positions. The fund that peaked at $45 billion in early July is now essentially an Anthropic holding company with a hedge fund attached.
🔍 THE BOTTOM LINE
The 25-year-old former OpenAI researcher who wrote the defining essay of the AI infrastructure boom — a 165-page treatise declaring “the AGI race has begun” — got margin-called on the very trade he told everyone else to make. Heavy leverage on AI chip stocks turned a bad month into a crisis. The thesis was not wrong that AI needs more chips. The leverage was the problem.
The Trade That Broke
The fund made one big, leveraged bet: that the AI build-out would keep lifting companies supplying chips, memory, and power. That trade unravelled fast. Positions in memory maker Micron, SK Hynix, Sandisk, Nebius, and CoreWeave collapsed in the AI chip sell-off that deepened through late July.
The damage was brutal. Nebius, where the fund disclosed a multi-billion-dollar stake in May, fell about 48% from its peak, erasing roughly $35 billion in market value. Sandisk dropped 56% in barely a month. The fund had also bet against software — shorts in names like Adobe moved the wrong way, squeezing it from both sides.
It was, in effect, the long-chips, short-software trade that defined this year’s market, taken to an extreme with borrowed money. Leverage turned a bad month into a crisis. Bank of America, Goldman Sachs, and JPMorgan were left marketing its positions, according to The Next Web.
From $225 Million to a Fire Sale
The speed of the rise makes the fall more striking. Aschenbrenner launched the fund in 2024 with about $225 million, backed by Stripe founders, Nat Friedman, Daniel Gross, and the trading firm Jane Street. It grew past $20 billion, and CNBC reported it swelled to as much as $45 billion at the start of July. The fund was up 439% in the first half of the year.
It ran on a skeleton crew. According to The Verge, the fund had eight employees, only four of them investment professionals. Aschenbrenner had no prior trading experience before launching it. He does have a following: more than 250,000 people on X.
The tone did not shift as the losses mounted. In a July 24 letter to investors seen by the Financial Times, Aschenbrenner said the fund had “not been immune” to the sell-off. He then called it one of the best buying windows since early 2025. A postscript invited clients to add fresh cash on August 1.
What Citadel Bought
Ken Griffin’s Citadel acquired the entire public book in one trade. The Wall Street Journal reported that Millennium had also bid for the portfolio. The collapse was first reported by CNBC’s David Faber, who said the fund’s prime brokers had been scrambling to raise cash to meet margin calls.
Roughly two-thirds of the fund’s assets were public equities, held long and shorted. The whole public book went in one enormous trade. Bloomberg reported that the fund’s borrowing magnified returns — a strategy that amplifies losses just as fast.
There is a wrinkle worth noting. The very stocks the fund had to dump — SK Hynix and the rest — jumped sharply the next day. Some market participants read that as a sign the forced selling marked a short-term bottom rather than a verdict on the AI trade itself.
The Anthropic Stake Survives
One large bet is still standing. Situational Awareness keeps its private holdings, and the biggest is a stake in Anthropic, which the Financial Times valued at about $5 billion. The firm will carry on as a private investment vehicle.
Reports that the fund was shopping the Anthropic stake are “not accurate,” a spokesman told CNBC. Aschenbrenner is engaged to Avital Balwit, chief of staff to Anthropic CEO Dario Amodei — a personal connection that has drawn attention but does not appear to have played a role in the fund’s investment decisions.
The Prophet on Trial
This is a personal blow to one of the most watched figures in the AI trade. Aschenbrenner, now 25, was a Columbia valedictorian at 19 and worked on OpenAI’s Superalignment team before the company fired him in 2024. His essay “Situational Awareness” gave the fund its name and the market its script. He forecast trillion-dollar compute clusters and hundreds of millions of humming GPUs. It became the intellectual playbook for the entire infrastructure bet.
The essay was not wrong that AI needs more chips. We tracked his power-bottleneck thesis in May, when his $878 million Bloom Energy position was worth $2.7 billion and rising. The fund’s undoing was leverage, not the thesis.
The comparisons wrote themselves. ZeroHedge dubbed it “Archegos 2.0,” after the family office that imploded on hidden leverage. Others reached for Three Arrows, the crypto fund that believed in its own supercycle. Commentators could not resist the obvious point: a firm named Situational Awareness had missed the risk in its own book.
NZ Angle
The collapse carries a signal for NZ investors who have exposure to AI infrastructure through international funds or ETFs. The AI chip sell-off that broke Aschenbrenner’s fund has gone global, and forced deleveraging tends to overshoot on the way down. For a market as small as New Zealand’s, the lesson is not about hedge fund leverage — it is about concentration risk. A fund with eight people, 4x leverage, and a single thematic bet was always one bad month away from a margin call. Diversification is not a hedge fund strategy. It is a survival strategy.
❓ FAQ
What was Situational Awareness’s main trade? Long AI infrastructure stocks (Micron, SK Hynix, Sandisk, Nebius, CoreWeave) and short software stocks (Adobe and others), amplified with heavy borrowing. When chip stocks fell and software rose, both sides of the trade moved against the fund simultaneously.
How much money did the fund lose? The fund peaked at approximately $45 billion in early July. The exact loss is not public, but the entire public equity portfolio was sold to Citadel. The private holdings, including a $5 billion Anthropic stake, remain.
Who is Leopold Aschenbrenner? A 25-year-old former OpenAI Superalignment team member who wrote a widely circulated 165-page essay called “Situational Awareness” arguing that the AGI race had begun and forecasting massive compute infrastructure spending. He launched his hedge fund in 2024.
Why did Citadel buy the portfolio? Citadel acquired the entire public book in a single trade, reportedly at a discount to market value. Millennium also bid. This type of block acquisition is common when a fund faces forced liquidation — the buyer gets a portfolio at a discount, the seller gets immediate liquidity.
Does this mean the AI infrastructure thesis is wrong? Not necessarily. The stocks the fund was forced to sell bounced the next day, and the underlying demand for AI compute has not disappeared. The fund’s failure was about leverage and concentration, not about whether AI needs more chips.
🔍 THE BOTTOM LINE
A 25-year-old with no trading experience, an eight-person team, and 4x leverage made the most aggressive version of the AI infrastructure trade. When the market turned, the leverage turned a correction into a margin call. The essay that named the fund was not wrong about the direction of AI compute. It was silent about the cost of getting the timing wrong.
📰 Sources
- The Next Web
- Bloomberg
- CNBC
- Wall Street Journal
- Financial Times
— CJ Murden, editor of Singularity.Kiwi. Former digital technologies teacher, author of AI-focused books. Writing with a New Zealand focus.
