News · 2026-08-14

An AGI-thesis fund fell 67 percent and took a market maker with it

Situational Awareness, the investment firm founded by Leopold Aschenbrenner explicitly around a thesis about artificial general intelligence, fell 67 percent in July, sold the bulk of its stock portfolio to Citadel, and raised cash to meet margin calls from lenders. The Financial Times reports that Jane Street, one of the world's largest trading firms, took a hit of roughly 15 billion dollars connected to trouble at the fund. This is the first time the AGI investment thesis has produced a loss large enough to matter to the wider market.

Key facts

• The anchor number: down 67 percent in July, per Reuters, followed by a forced sale of most of the stock book to Citadel.
• Who: Situational Awareness LP, a San Francisco investment advisor founded by Leopold Aschenbrenner, with anchor investments from Patrick Collison, John Collison, Nat Friedman and Daniel Gross.
• The reported knock-on: approximately 15 billion dollars at Jane Street, per the Financial Times.
• What is not established: the mechanism by which Jane Street was exposed.

The firm is real, registered, and open about what it does. Its own description says it invests in "companies involved in AI development" and in companies "that AI may help or hurt," forming a macro view of AI development through technological and economic research and then looking for both winners and disrupted incumbents. Aschenbrenner's public biography confirms he "recently founded an investment firm focused on AGI," with backing from a recognizable set of technology investors. A regulatory 13F filing confirms the entity exists and reports public securities holdings, though the filing alone does not reveal the full portfolio composition.

The July event was not a soft drawdown. Reuters reported that the fund was down 67 percent after heavy losses on AI stocks, that it sold the bulk of its stock portfolio to Citadel, and that it had to gather cash to meet margin calls from lenders. Each of those elements points the same direction: this was a leveraged position on public equities being liquidated under pressure, not a fund quietly marking down its book.

Margin is the mechanism that turns a bad month into a forced one. An investor who borrows against a portfolio must maintain a minimum level of equity. When prices fall, the lender demands more cash, and if the investor cannot produce it, positions are sold at whatever price the market offers rather than at a time of their choosing. That is why the sale to Citadel is the most informative detail in the story: selling the bulk of a book in one block to a single buyer is what a firm does when it needs cash now, not when it is repositioning.

What makes this specific rather than generic is the thesis. Plenty of funds hold technology stocks. This one was constructed around a view that AI development follows a particular trajectory, and expressed that view with leverage in public markets. Reuters reported that private holdings, including a stake in Anthropic, survived the unwind. So the losses came from the liquid, leveraged part of the book, which is exactly the part that a sharp drawdown in AI-linked equities would hit hardest and fastest.

The Jane Street figure is the part that turns a fund story into a market story, and it is also the part to handle with the most care. The FT headline reports a roughly 15 billion dollar hit connected to the trouble at Situational Awareness. The article body is paywalled, and no public filing establishes what Jane Street's relationship to the fund actually was. Investor, lender, trading counterparty, seed backer, or some combination are all consistent with the reporting available, and journalistic summaries attribute the loss partly to the fund and partly to broader weakness in AI stocks. A number that large deserves a precise mechanism, and there is not one on the public record yet.

The analogy worth holding is not a stock picker being wrong. It is a leveraged bet on a timeline. If a thesis says a technology will transform the economy on a particular schedule, and you borrow to express that view, you are not only betting on whether the thesis is right. You are betting on whether it is right before your lenders lose patience. Being early and being wrong produce identical margin calls.

The honest caveat is that a single fund's blowup proves nothing about the underlying thesis. AI capability has continued advancing regardless of what happened to any portfolio in July, and markets have historically punished correct long-term views on short-term schedules many times over. What this episode does establish is narrower and still important: AI-linked leverage has grown large enough that its unwinding is now visible on the balance sheet of a major market maker. That is a structural fact about the financial system, not a verdict on the technology.

Related on Ground Truth: Nvidia lining up six financiers to mobilize 500 billion dollars.

Primary source, verified: read the paper →

Ground Truth. (2026, August 14). An AGI-thesis fund fell 67 percent and took a market maker with it. Ground Truth. https://groundtruth.day/news/an-agi-thesis-fund-fell-67-percent-and-took-a-market-maker-with-it.html

@misc{groundtruth:an-agi-thesis-fund-fell-67-percent-and-took-a-market-maker-with-it,
  title  = {An AGI-thesis fund fell 67 percent and took a market maker with it},
  author = {{Ground Truth}},
  year   = {2026},
  month  = {aug},
  url    = {https://groundtruth.day/news/an-agi-thesis-fund-fell-67-percent-and-took-a-market-maker-with-it.html}
}

Topics: industry · finance · markets · ai-investment · risk