Jane Street posted a loss of roughly $15bn for July after the collapse of Situational Awareness, the AI-focused hedge fund it had backed, according to the Financial Times, which reported the figure on August 14, 2026. The disclosure surfaced not through a public filing but through a debt refinancing that moves the trading firm's borrowing out of public markets and into the hands of a small group of private lenders.

The number is unusual for two reasons. It is large enough to wipe out a third of what the firm earned across all of 2025, and it became known only because a financing transaction forced it into the open.

According to the Financial Times, the New York-based firm disclosed the July figure to lenders as part of a deal to shift its roughly $11bn public debt pile to private investors including Pimco. Both Jane Street and Pimco declined to comment. The wider transaction, a $14.6bn 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. One consequence of the switch is structural: moving from the public debt market to a private lender group narrows the set of parties that receive regular financial reporting from the firm.

What Jane Street is and why the loss stands out

Jane Street was established in 2000 by a small group of founders including Robert Granieri, who previously worked at the Pennsylvania-based firm Susquehanna, according to the Financial Times. It uses technology to make markets across equities, bonds, exchange traded funds and other assets. Market making is a high-turnover, spread-capture business, and for most of its history the firm's public profile has been close to nonexistent.

Its recent numbers have not been. In the first quarter of 2026 the firm recorded net trading revenues of $16.1bn, a figure that outstripped comparable hauls at Wall Street banks. Full-year 2025 revenues came in at roughly $40bn. Even after absorbing the July loss, net trading revenues for the year to Friday exceeded $40bn, which is more than the firm produced across the whole of 2025, according to a person familiar with the matter cited by the Financial Times. The July shortfall lands in the third quarter, which was still running at the time of the report.

What makes the exposure notable is that Jane Street trades its own capital. An investment in an outside hedge fund sits outside that model. The Financial Times reported that a former Jane Street employee worked at Situational Awareness and that Granieri attended the wedding of the fund's founder, Leopold Aschenbrenner, in California in recent weeks. The firm has also taken direct stakes in private AI-adjacent companies, including the AI lab Anthropic and the data centre operator CoreWeave.

The fund at the centre of it

Situational Awareness was founded in 2024 by German-born Leopold Aschenbrenner, then in his early 20s and a former OpenAI employee, according to CNN. The firm's site states that it was built on the premise that "AI will be the dominant driver of global market returns over the next decade."

The intellectual groundwork came first. Aschenbrenner left OpenAI's Superalignment team in April 2024 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. According to CNN, the essay argued that governments, businesses and investors underestimated how quickly AI would transform the economy.

The fund followed later that year, with Aschenbrenner as chief investment officer and managing partner. Early backers included Stripe founders Patrick and John Collison, former GitHub chief executive Nat Friedman, and Daniel Gross. The thesis was mechanical rather than thematic: the AI race would pull trillions of dollars into semiconductors, networking hardware, data centres and power, and the fund would be long those names while shorting industries it expected to be left behind.

The returns arrived quickly. The fund gained 47% in the first half of 2025 against 6% for the S&P 500, managing about $1.5bn at that point. Full-year 2025 finished at roughly 200%. A first-quarter 2026 regulatory filing showed $13.7bn in portfolio value, with the book rotating out of traditional semiconductor names into AI and high-performance computing, energy and digital infrastructure holdings including Riot, Bitdeer, IREN and Keel. By June 2026 assets under management passed $20bn, cumulative net gains since launch exceeded 1,000%, and the fund co-led a $500m 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% net return for the first half of the year. The letter carried a postscript suggesting the moment was a good one to add funds.

How the unwind happened

Six days later the position broke. On July 30 and 31, Goldman Sachs, JPMorgan and Bank of America issued margin calls on a portfolio leveraged roughly four times over. The entire public equity book went to Citadel, the hedge fund run by Ken Griffin, sold at a steep discount. CNN reported that the fund was forced to sell the bulk of its public holdings on Thursday, July 30, after many of its investments went south.

The mechanism was leverage, not stock picking alone. Borrowing to buy equities amplifies gains and amplifies losses at the same ratio, and a concentrated book compounds the effect because there is no offsetting position to cushion a drawdown. Holdings fell from a peak that some reporting places at $45bn to around $10bn. Individual names took heavy damage: SK Hynix, CoreWeave, Nebius, SanDisk and Micron each shed more than 35% during the period.

A subsequent investor letter admitted a single-month loss of approximately 67%. Aschenbrenner attributed part of the decline to short sellers, said all leverage had been removed, and took full responsibility. According to CNN, he reportedly likened the episode to a "bank run" in that letter. CNN also cited Wall Street Journal reporting that the fund's remaining positions are not financed by borrowing. Situational Awareness did not respond to CNN's request for comment.

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, 2026, the fund invested $400m in Source Foundry, a chip-manufacturing startup founded by the Stanford researchers Abdulmalik Obaid and Joe Burg, taking its total stake in the company to $500m.

Contested figures

Not every number in the public account agrees. The peak-holdings figure of $45bn appears in some reporting, while other accounts put the top nearer $20bn, a gap wide enough that the top-line drawdown should be treated as unsettled rather than established. Assets under management, portfolio value disclosed in a quarterly filing, and gross long exposure created by leverage are three different quantities, and coverage has not always separated them cleanly. The 67% monthly loss and the $15bn Jane Street figure come from different sources and measure different things: one is a fund return, the other a proprietary trading firm's monthly profit and loss.

Why an AI hedge fund matters to advertising

The connection between a levered equity book in New York and a media plan is not obvious at first glance. It runs through capital expenditure.

The buildout that ad platforms are financing

Every large advertising platform is now also an infrastructure borrower. Alphabet raised its 2026 capital expenditure guidance to between $195bn and $205bn on July 22, 2026, the second increase of the year, and reported negative free cash flow of $5.9bn for the quarter alongside $98.2bn in long-term debt. In June the company raised approximately $85bn in equity to fund infrastructure it described as supply-constrained, combining a private placement to Berkshire Hathaway, an oversubscribed public offering and an at-the-market programme. Meta reported quarterly capital expenditures of $31.08bn on July 30, 2026, against $17.01bn a year earlier, with free cash flow down to $784m. Microsoft capital expenditures rose 70% to $41.0bn in its June quarter.

Credit markets have already registered the strain elsewhere. S&P Global Ratings cut Oracle to BBB- in July 2026, citing a forecast fiscal 2027 free operating cash flow deficit of nearly $42bn and noting that the unit economics of the AI infrastructure business remain opaque. 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. Its short book was a bet against the companies it expected to lose from it. 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.

The advertising revenue that AI money buys

There is a second, more direct channel. AI-native companies buy advertising, and the infrastructure race itself has become an advertising revenue line. WPP Media's June 2026 forecast attributed advertising's growth to a single force, pointing to an investment cycle in which AI companies spend on advertising to reach developers, enterprise buyers and consumers. That forecast also found that more than 80% of AI advertising in 2026 will run adjacent to AI content, such as search ads beside AI Overviews, rather than inside chatbots, a ratio that EMARKETER's separate 2030 projection reached independently.

Budgets funded by venture capital and infrastructure equity behave differently from budgets funded by operating cash flow. They expand faster and they contract faster.

The tooling layer sits in the same position. Agencies have begun treating AI token consumption as a margin business, negotiating wholesale rates with model providers and cloud vendors and reselling that consumption to clients, a construction that mirrors principal media buying. Ruben Schreurs, chief executive of Ebiquity, described the underlying pressure by noting that agencies cannot afford to keep subsidising those costs. That model prices off the same infrastructure economics now under scrutiny. On the demand side, KPMG research published in August 2026 found that 49% of organisations cut AI agent rollouts when costs outran value, while TransUnion research from August 5, 2026 recorded only 53% of senior US marketing leaders reporting meaningful ROI from AI against 89% expecting to increase AI spending.

The bubble question, again

The comparison to earlier asset cycles is not new to the industry. Jeff Bezos described the current AI period as an industrial bubble at Italian Tech Week on October 1, 2025, distinguishing between market valuations and business fundamentals. A Yale study catalogued classic bubble indicators in the AI market as early as 2024, including overleveraged investments and the substitution of conventional valuation metrics with model and benchmark measures.

What the Situational Awareness unwind adds is a data point about mechanics rather than valuation. A fund that had returned more than 1,000% since launch lost roughly two-thirds of its value in a single month without any change in the published capital expenditure guidance of the companies it was betting on. The move came from positioning and financing, not from a revision to the AI spending forecasts themselves.

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, which committed in February 2026 to keeping its Claude assistant free of advertising while running Super Bowl commercials, and in CoreWeave, whose compute serves the models behind a growing share of ad tech tooling. The same infrastructure that supports frontier model training also runs programmatic bidding, creative generation and retail media assistants.

Nick Whittaker's argument, covered on August 13, 2026, framed the pressure as a double compression: regulation squeezing monetisation per user while capital expenditure explodes, with Bank of America describing the shift as a generational transfer of free cash flow from big technology companies to chipmakers. A $15bn monthly loss at a market maker that had been printing record revenues is one more reading on how much risk has accumulated in the transfer.

Timeline

Summary

Who: Jane Street, the New York proprietary trading firm founded in 2000 by a group including Robert Granieri, and Situational Awareness, the AI-focused hedge fund founded in 2024 by Leopold Aschenbrenner. Citadel, run by Ken Griffin, bought the fund's public equity book. Pimco is among the private lenders taking on Jane Street debt in a $14.6bn deal led by JPMorgan.

What: A loss of roughly $15bn for July at Jane Street, disclosed to lenders during a debt refinancing, following margin calls that forced Situational Awareness to sell its entire public equity book. The fund reported a single-month loss of approximately 67% after a first half in which it returned 439%.

When: The forced sale occurred on July 30 and 31, 2026. CNN published its explainer on July 31. The Financial Times reported the Jane Street figure on August 14, 2026. The loss falls into Jane Street's third quarter.

Where: New York, where both firms are based, with holdings spanning US and international semiconductor, data centre, energy and digital infrastructure names including SK Hynix, CoreWeave, Nebius, SanDisk and Micron.

Why: The episode matters to advertising because the AI infrastructure buildout now funds a measurable share of platform capital expenditure and advertising revenue. Alphabet guided to $195bn to $205bn in 2026 capital expenditure with negative quarterly free cash flow, Meta doubled quarterly capital expenditure year over year, and S&P downgraded Oracle over a forecast $42bn cash flow deficit. A leveraged fund built on that thesis lost two-thirds of its value in a month without any change to the published spending forecasts, which locates the risk in financing conditions rather than in the forecasts themselves.