Back to the archive

The encyclopedia · Trading & Investing · Financial decision · 2026

A $45B AI fund lost 67% in one month and sold its book to Citadel

Situational Awareness peaked at $45B on leveraged AI bets, lost about 67% in July 2026 and sold $16B of stock to Citadel at a discount to meet margin calls.

Situational Awareness · 2026-07-30

What happened

Situational Awareness, the hedge fund Leopold Aschenbrenner launched after leaving OpenAI in 2024, traded his essay thesis: AI would need an enormous build-out of chips, memory and power. The fund ran leveraged longs on AI infrastructure — SK Hynix, Nebius, Sandisk, Micron, CoreWeave — against shorts on software companies such as Adobe. Through June 2026 it returned about 439% and swelled to roughly $45 billion of assets by early July.

Then the AI trade cracked: the sector shed some $3 trillion of market value, the fund's biggest holdings each fell more than 35% in July, and the software shorts moved the wrong way. Prime brokers Bank of America, Goldman Sachs and JPMorgan issued margin calls. The fund lost about 67% in July and was forced to sell its entire public book — roughly $16 billion — which Ken Griffin's Citadel agreed to buy at a discount of about 10%, in trades arranged around 30 July 2026.

What remained was roughly $10 billion, mostly private holdings including a stake in Anthropic. Even then the fund was up about 80% on the year — July's margin call erased the compounding of a leveraged run while the market argued over whether the AI trade itself had broken, or just its financing.

Why it happened

  • One thesis sat on both sides of the book: long the hardware AI needs, short the software it replaces. When the narrative wobbled, both sides lost at once.
  • Leverage turns conviction into a timing problem — the trade may still have been right, but margin calls do not wait for it to be.
  • The longs were concentrated in less liquid AI-infrastructure names, so the only exit big enough for a $16 billion book was a block sale at a discount.
What it cost~67% lost in July; $16B sold at ~10% discountcostly

The lesson

Leverage doesn't decide whether you're right — it decides whether you stay long enough to be right. A 439% run ended in a forced sale because the fund's timeline was set by its prime brokers.

Sources

spotted an error? The club wants to know.

Comments · 0

    Sign in to join the comments.

    More like this

    Somewhere, someone solved the problem this company failed at. 2nd Opinion →