Jane Street loses $15bn as leveraged AI bets collapse

Jane Street lost roughly $15 billion in July after the collapse of Situational Awareness, the AI-focused hedge fund it backed, according to the Financial Times.

The disclosure surfaced through a debt refinancing that moves the trading firm's borrowing out of public markets and into the hands of private lenders including Pimco, the FT reported on August 14. The wider transaction, a $14.6 billion private debt deal led by JPMorgan, combined a rollover of existing public debt with additional borrowings, and the arrival of new lenders triggered the disclosure obligation. Both Jane Street and Pimco declined to comment.

The fund, run by former OpenAI researcher Leopold Aschenbrenner, was leveraged roughly four times over when Goldman Sachs, JPMorgan and Bank of America issued margin calls on July 30-31. Holdings fell from a peak of $45 billion to around $10 billion, with SK Hynix, CoreWeave, Nebius, SanDisk and Micron each shedding more than 35 percent during the period.

The loss lands in Jane Street's third quarter and wipes out a third of what the firm earned across all of 2025. The episode raises questions about the financing conditions underpinning the AI infrastructure buildout, which now funds a measurable share of platform capital expenditure and advertising revenue.

Jane Street, established in 2000 by a group including Robert Granieri, recorded $16.1 billion in net trading revenues in the first quarter of 2026, outstripping comparable hauls at Wall Street banks. Full-year 2025 revenues came in at roughly $40 billion. Even after absorbing the July loss, net trading revenues for the year to Friday exceeded $40 billion, according to a person familiar with the matter cited by the FT.

The fund at the center

Situational Awareness was founded in 2024 by Aschenbrenner, then in his early 20s, after he left OpenAI's Superalignment team over an alleged information leak, which he disputes. On June 4, 2024, he published a 165-page essay arguing that continued scaling of compute and algorithmic efficiency made artificial general intelligence around 2027 plausible.

The fund gained 47 percent in the first half of 2025 against 6 percent for the S&P 500, managing about $1.5 billion at that point. Full-year 2025 finished at roughly 200 percent. By June 2026, assets under management passed $20 billion, cumulative net gains since launch exceeded 1,000 percent, and the fund co-led a $500 million round in the AI chip startup MatX. Jane Street joined as a backer in the same period.

On July 24, 2026, Aschenbrenner wrote to investors reporting a 439 percent net return for the first half of the year. Six days later, the position broke. A subsequent investor letter admitted a single-month loss of approximately 67 percent. Aschenbrenner attributed part of the decline to short sellers, said all leverage had been removed, and took full responsibility.

The fund was not liquidated. It continues as a mixed vehicle holding unlevered public positions alongside a private book that came through intact, including stakes in Anthropic, Fluidstack and MatX. Between August 5 and August 10, the fund invested $400 million in Source Foundry, a chip-manufacturing startup.

What the unwind means for AI infrastructure

The episode matters beyond the trading floor because the AI infrastructure buildout now funds a measurable share of platform capital expenditure and advertising revenue. Alphabet raised its 2026 capital expenditure guidance to between $195 billion and $205 billion on July 22, the second increase of the year, and reported negative free cash flow of $5.9 billion for the quarter. Meta reported quarterly capital expenditures of $31.08 billion on July 30, against $17.01 billion a year earlier. Microsoft capital expenditures rose 70 percent to $41.0 billion in its June quarter.

S&P Global Ratings cut Oracle to BBB- in July 2026, citing a forecast fiscal 2027 free operating cash flow deficit of nearly $42 billion. The same concentration question has been raised about Google Cloud's projected reliance on OpenAI and Anthropic revenue.

Situational Awareness was, in effect, a leveraged bet on that spending continuing. When the equity market repriced the trade in July, the leverage did the rest. The episode says nothing definitive about whether the underlying capital expenditure forecasts are correct. It does say something about how quickly financing conditions for that thesis can change.

For marketers, the practical relevance is narrow but real. A meaningful share of 2026 advertising growth is attached to companies whose funding depends on capital markets remaining receptive to AI infrastructure risk. Jane Street's own portfolio illustrates the overlap: the firm holds positions in Anthropic and CoreWeave, whose compute serves the models behind a growing share of ad tech tooling.

This article is for informational purposes only and does not constitute investment advice.