Illustrative image generated for Smart Travel Finance. Not affiliated with JPMorgan Chase, Situational Awareness, or Citadel.
JPMorgan Cuts Lending to AI Hedge Fund After 67% Loss: How Leverage Turned an AI Bet Into a Credit Risk
A hedge fund built almost entirely around artificial intelligence bets lost roughly two-thirds of its portfolio value in a single month, after losing money on both sides of its own strategy at once. What happened next explains something bigger than one fund's bad month: how borrowed money can turn a market decline into a lending decision at one of the country's largest banks.
- JPMorgan Chase ended its lending and prime-brokerage relationship with Situational Awareness, an AI-focused hedge fund founded in 2024 by former OpenAI researcher Leopold Aschenbrenner, after the fund's July 2026 losses
- The fund's portfolio fell approximately 67% in July 2026, driven by two things happening at once: its long positions in AI-infrastructure names (including SK Hynix, Nebius, and Micron) fell, while its short positions in software names like Adobe moved against it in the opposite direction
- Independent reporting describes the fund as operating with leverage as high as roughly 4x at the time, and its assets peaked at approximately $45 billion at the start of July before the decline
- The resulting margin call forced the fund to sell roughly $16 billion of its public equity holdings to Citadel, reportedly at about a 10% discount, leaving it with around $10 billion in remaining assets
- The fund retained a private stake in Anthropic, reported at roughly $5 billion, and remained up an estimated 80% for the year overall despite the July decline
- The SEC has sent subpoenas to banks that worked with the fund regarding its trading and leverage use; that process is ongoing with no reported outcome
Situational Awareness is a hedge fund built around a specific idea: that artificial intelligence would reshape markets quickly enough to reward large, concentrated bets on the companies driving it. Founded in 2024 by Leopold Aschenbrenner, a 24-year-old former OpenAI researcher, the fund grew rapidly, reportedly gaining roughly 439% in the first half of 2026 alone and reaching an estimated $45 billion in assets by early July, according to reporting from CNBC and other outlets citing people familiar with the fund.
The strategy paired large long positions in AI infrastructure companies with short positions in software companies the fund's managers expected to underperform by comparison. In July 2026, both sides of that bet moved the wrong way at the same time: the long AI-infrastructure holdings fell as chip and AI-hardware stocks sold off broadly, while several of the shorted software names, including Adobe, rallied instead of falling. That combination, not a one-sided market move, is what independent reporting describes as driving the roughly 67% portfolio decline that month.
⚠ What This Article Does Not Claim
This is not a report that Situational Awareness is bankrupt or insolvent; multiple sources describe the fund as continuing to operate with roughly $10 billion in remaining assets and a private Anthropic stake. It is not a claim that JPMorgan's decision reflects a specific internal rationale, since JPMorgan has not disclosed one and declined to comment when asked by Reuters. It is not a claim that AI investing as a category has failed, or that any bank, chipmaker, or AI company bears fault.
The 67% Loss: What Actually Drove It
A decline of that size, concentrated in a handful of related positions, produced a margin call that forced the fund to raise cash quickly. According to reporting citing an investor letter obtained by the Financial Times, Aschenbrenner acknowledged the losses directly to investors, taking responsibility and describing them as costly but instructive setbacks for the firm going forward. He has also continued to express confidence in the long-term case behind the fund's AI-related positions, separate from the short-term financing problem the losses created.
Why the Margin Call Changed Everything
A margin call is what happens when the value of assets used as collateral for borrowed money falls far enough that the lender no longer considers that collateral sufficient. The mechanism runs in a predictable sequence.
That last step is what makes margin calls dangerous beyond the initial loss. Reporting indicates Situational Awareness had to liquidate essentially all of its public long and short positions as a result, in a single block trade completed by July 30, with roughly $16 billion of holdings sold to Citadel at an estimated 10% discount to market value.
How Leverage Magnified the Risk
Leverage means controlling a larger position than your own capital alone would allow, using borrowed money to do it. Independent reporting describes Situational Awareness as operating with gross leverage as high as roughly 4x at the time of the decline, meaning its market exposure was around four times its own capital base.
The Math, Using the Actual Reported Leverage Level
Neither the fund nor its lenders have published an exact own-capital figure, so the calculation below uses the fund's reported ~$45 billion peak asset base as a stand-in for total exposure, applied at the independently reported ~4x leverage level, purely to illustrate the arithmetic.
Implied own capital at 4x: ~$45B ÷ 4 ≈ $11.25B
A 67% decline applied to $45B in exposure ≈ $30.15B lost
That loss relative to ~$11.25B of implied own capital ≈ 268% of own capital
This is Smart Travel Finance's own illustrative arithmetic combining independently reported figures (peak AUM, leverage level, percentage decline); it is not an official accounting of the fund's actual capital structure, which has not been fully disclosed. It is shown to illustrate why 4x leverage turns a large but survivable market move into an existential one.
The core point is arithmetic, not speculation: at 1x exposure (no borrowed money), a 67% decline in assets costs you 67% of your own capital. At 4x exposure, the same underlying decline can wipe out several times your own capital, because you were exposed to four times as much market movement using a much smaller base of your own money as collateral.
Could You Survive the Drawdown? Try the Numbers Yourself
The Leverage Drawdown Simulator
Pick a starting capital amount and a leverage level, including 4x, the level independent reporting describes Situational Awareness as having used. This tool applies a hypothetical 67% decline in exposure to show how leverage changes the outcome. It is a simplified educational example, not a reproduction of the fund's actual books.
This is a simplified educational example. It does not reproduce Situational Awareness's actual capital structure or losses precisely, and figures should be treated as illustrative. This tool is not investment advice.
Why JPMorgan Cut the Lending Relationship
JPMorgan Chase was one of several prime brokers, alongside Goldman Sachs and Bank of America, working with the fund to help meet margin requirements during the decline. Following the losses, JPMorgan informed Situational Awareness it would end its lending relationship; JPMorgan declined to comment when Reuters asked. No further internal rationale has been disclosed. Goldman Sachs, Citigroup, and Bank of America reportedly continued facilitating parts of the transaction and ongoing brokerage relationship, and the fund also began working with Clear Street.
Where Citadel Fits Into the Story
Citadel, led by Ken Griffin, acquired essentially the fund's entire public book, long and short positions alike, in a single block trade completed by July 30, reportedly at around a 10% discount to prevailing market value. That reflects a forced, distressed sale, not a signal about the value of those specific positions, and this article does not treat it as an endorsement or trading signal.
The Bigger Risk: When AI Exposure Meets Leverage
It's worth being precise about what actually created the risk here. This wasn't simply "AI stocks fell." The fund lost money on its long AI-infrastructure bets and its short software bets at the same time, a combination that a purely one-directional market move wouldn't have produced. The risk emerged from several conditions overlapping: concentrated exposure to related, correlated positions on both sides of the book, leverage around 4x magnifying the effect of that decline, and a market move fast enough to trigger margin calls before the fund could adjust. This dynamic is likely to recur wherever AI-related financial risk intersects with leveraged trading, and it sits alongside a broader wave of AI-linked financing, including infrastructure players turning to corporate debt markets to fund AI buildouts, a related but distinct form of AI-linked exposure built on issued debt rather than fund leverage.
What This Means for Banks
Consumer-facing coverage of bank risk tends to focus on mortgage delinquencies or credit card losses. Banks also extend credit through hedge fund lending and prime brokerage, exposure that moves with market volatility rather than personal finances. This single case does not indicate a systemic banking problem; JPMorgan's own reported quarterly profits reflect a business with far broader lending activity than any single hedge fund relationship. What it illustrates is that AI-related market volatility is now a real factor in institutional lending decisions, alongside continued AI infrastructure financing from companies behind recent AI infrastructure capital raises.
Does This Mean AI Investing Is Broken?
No. It's useful to separate three things: an AI investment thesis, a leveraged trading strategy built around it, and the risk management applied to that strategy. This case demonstrates a risk specific to leverage and concentration in a fast-moving sector. It does not demonstrate that unleveraged AI-related investing carries the same risk profile, and the fund's continued ~80% estimated full-year gain, driven substantially by a retained private stake, is itself a reminder that the underlying thesis and the leverage that nearly broke it are two separate things.
What Investors Can Learn From This
- Leverage magnifies gains and losses in both directions. The same borrowing that boosted returns in the rally increased the size of the losses in the decline.
- A long/short strategy isn't automatically hedged. When both sides of a book move against you at once, the "hedge" can amplify losses instead of cushioning them.
- Margin calls force selling at the worst possible time, converting a paper loss into a realized one at a discount, as the roughly 10% haircut on the Citadel sale illustrates.
- Liquidity matters as much as conviction. Being right about a long-term thesis doesn't help if a fund is forced to exit positions first.
- A strong long-term view does not eliminate short-term financing risk. Those are separate risks, managed separately.
This section describes general financial concepts illustrated by this case. It is not personalized investment advice.
JPMorgan, Situational Awareness and the Credit-Risk Chain
The sequence below connects each confirmed event in this story. Click through each stage to see how it links to the next.
Visual Illustration: How a Market Decline Became a Lending Story
This original illustration is based on the themes discussed in this article. It is for educational purposes only and is not an official graphic from JPMorgan Chase, Situational Awareness, or Citadel.
Illustrative graphic created for Smart Travel Finance. Based on the themes discussed in this article; not an official graphic from JPMorgan Chase, Situational Awareness, or Citadel.
Frequently Asked Questions
Did Situational Awareness go bankrupt?
No. The fund reportedly retains around $10 billion in assets, including a private Anthropic stake, and remained up an estimated 80% for the year despite the July decline.
Why did JPMorgan end its relationship with the fund?
Reuters reports that JPMorgan ended the relationship after the fund's losses, but no specific internal reason has been disclosed. JPMorgan declined to comment.
What actually caused the 67% loss?
Independent reporting indicates the fund lost money on both its long AI-infrastructure positions and its short software positions at the same time, amplified by roughly 4x leverage, not simply a one-directional chip-stock selloff.
Does this mean AI stocks are a bad investment?
No. This case illustrates risks tied to leverage and concentration, not a verdict on AI-related investing generally.
What happened to the SEC subpoenas mentioned in this story?
The SEC sent subpoenas to banks that worked with the fund regarding its trading and leverage. That process is ongoing with no reported outcome.
Is this a sign of broader risk in the banking system?
No. This describes one lending relationship changing after one fund's losses, not a systemic banking event.
Can You Read Past the Headline Loss Number?
Answer 6 quick questions about how you'd interpret a story like this one. This is a reflection tool to help you think through the story, not a rating of any fund or bank.
This tool reflects your own reasoning back to you for educational purposes only. It does not represent JPMorgan, Situational Awareness, or Citadel.
- This is not a claim that Situational Awareness is insolvent or has filed for bankruptcy
- This is not a claim about JPMorgan's specific internal reasoning, which has not been disclosed
- This is not investment advice regarding any hedge fund, bank stock, or AI-related security
- This is not a claim that AI investing as a category is unsound
- This is not a claim of fault against JPMorgan, OpenAI, Nvidia, or any other company
- This is not a statement about systemic risk in the U.S. banking system
Final Verdict
The key lesson isn't that AI investing has failed; the fund's own retained gains for the year suggest otherwise. The more important lesson is how quickly leverage, a long/short book turning against itself, and margin requirements can convert a market decline into a financing problem, one that reaches beyond the fund itself and into the lending decisions of the banks behind it.