Did Citadel Manipulate the AI Selloff? What the Situational Awareness Timeline Shows

Background: How a $45 Billion Fund Reached Zero Public Exposure in Six Days

Situational Awareness LP was founded in late 2024 by Leopold Aschenbrenner, a former OpenAI researcher who had published a 165-page essay of the same name arguing that advances in artificial intelligence would create enormous demand for semiconductors, memory, data centers, and electrical power. The fund launched with roughly $225 million. Backers included Stripe co-founders Patrick and John Collison, former GitHub chief executive Nat Friedman, investor Daniel Gross, and — unusually for a firm that rarely allocates externally — Jane Street.

The thesis worked, and then some. The fund returned approximately 439 percent net of fees through the end of June 2026 and reached roughly $45 billion in assets at the start of July, according to reporting by CNBC and the Financial Times.

In simple terms: the fund bet heavily that AI infrastructure companies would keep rising and that traditional software companies would fall, and it borrowed money to make both bets larger. Reported leverage ran as high as four times the fund’s capital.

July reversed both sides at once. Long positions in AI infrastructure names — Nebius Group, SanDisk, Micron, CoreWeave, and South Korea’s SK Hynix among them — fell sharply, with several core holdings down between roughly 27 and 54 percent on the month. Short positions in software companies including Adobe moved against the fund as those shares rose. The Nasdaq Composite ended one late-July session more than 10 percent below its all-time high. South Korea’s Kospi triggered a circuit breaker after an 8 percent single-day plunge, with SK Hynix falling nearly 13 percent.

Prime brokers Bank of America, Goldman Sachs, and JPMorgan Chase issued margin demands. On July 30, before the U.S. market opened, Situational Awareness sold its entire public book — longs and shorts together, valued at approximately $16 billion by the Financial Times — to Citadel in a single block trade. Goldman Sachs, JPMorgan, Bank of America, and Citigroup arranged the transfer. Millennium Management and Jane Street reviewed the portfolio and declined to participate, Bloomberg reported. The fund retained roughly $10 billion in remaining stock and private positions, including a stake in the AI company Anthropic, and did not close.

The Contested Four Days

The disputed portion of the timeline is narrow and can be stated precisely.

July 24 — In an investor letter reported by the Financial Times, Aschenbrenner described the ongoing selloff as producing some of the most attractive entry points since early 2025, and the fund was understood to be open to fresh capital.
July 27 — Frank Flight, head of macro strategy at Citadel Securities, published a note arguing the Federal Reserve could deliver a surprise quarter-point increase, describing such a move as a potential “cleansing event” for market pricing.
July 29 — The Federal Open Market Committee held its target range at 3.5 to 3.75 percent. Three of the twelve voting members preferred an increase.
July 30 — Citadel acquired the bulk of Situational Awareness’s public equity portfolio in a pre-market block trade assembled within roughly 24 hours.
July 31 — Several of the same AI-linked shares rebounded sharply once the threat of continued forced selling was removed from the market.

Everything in the allegation rests on the relationship between the second item and the fourth.

The Legal Standard: What Manipulation Actually Requires

Under Section 9(a)(2) of the Securities Exchange Act of 1934 and Rule 10b-5 promulgated under Section 10(b), market manipulation is not simply conduct that moves prices in a direction that happens to benefit the actor. Federal securities law requires a manipulative act or deceptive device, combined with scienter — the intent to deceive, manipulate, or defraud, or at minimum recklessness amounting to the same.

In simple terms: publishing a research opinion that turns out to be wrong, or that turns out to be profitable for the firm publishing it, is not by itself illegal. What is illegal is publishing a view the firm knows to be false, or does not actually hold, in order to move prices for the benefit of its own positions — a practice regulators sometimes call scalping.

A second legal structure matters here. Citadel Securities, the market-making and research business that published the July 27 note, and Citadel, the hedge fund that purchased the portfolio, are separately organized and separately regulated firms. Both were founded by Griffin. Broker-dealers are required under FINRA and SEC rules to maintain information barriers between research, trading, and affiliated advisory operations. Establishing manipulation across those two entities would require evidence that information or coordination crossed that barrier — not merely evidence of common ownership.

The Case That Something Improper Occurred

Shay Boloor, chief market strategist at Futurum Equities, articulated the theory most prominently. Writing on X on July 30, Boloor argued that Citadel had been advancing surprise rate-hike fears in the days immediately before the AI trade broke, and had then acquired most of the resulting distressed portfolio at significantly reduced prices — a sequence he characterized as “absolutely ruthless.” Similar claims spread rapidly through replies to market commentary accounts, some framing the episode in blunter terms as a deliberate flush of a rival.

Three factual observations lend the theory its force.

First, the timing is unusually tight. Four business days separate a public argument for a hawkish surprise from a transaction that required a distressed seller to exist.

Second, the target was visible. Situational Awareness’s holdings were disclosed in public 13F filings, and its investment thesis was famous. Any sophisticated trading desk could reconstruct the book with reasonable accuracy — and therefore knew which specific securities a wounded, four-times-levered fund would be compelled to sell. That is an information asymmetry that accelerates any margin spiral, whether or not anyone intends it to.

Third, neither the purchase price nor the discount has been disclosed, and Citadel’s own hedges are not public. The transaction’s economics cannot currently be evaluated by outside observers.

The Case That Nothing Improper Occurred

The counterargument is equally grounded in the record.

The rate-hike view was defensible on its merits. Three of the twelve FOMC voters did in fact prefer an increase at the July 29 meeting. A macro call that a quarter of the voting committee agreed with is difficult to characterize as a knowingly false statement, which is the element federal law requires. Under Chair Kevin Warsh, the Fed had already shifted hawkish: the June dot plot showed roughly half of officials projecting at least one 2026 increase, up from none in March, against inflation running at 4.2 percent in May.

The selloff was also overdetermined. July’s semiconductor decline was driven by a cluster of company-specific catalysts — Meta’s announcement of Meta Compute, offering excess AI capacity for sale; Intel’s 18A manufacturing delay alongside AMD overtaking it in data-center revenue; and Samsung’s record profits being met with selling rather than buying. Goldman Sachs estimated that Asia-focused fundamental long-short funds lost an average of 18.6 percent in July through the 28th. Situational Awareness was the largest casualty of a sector-wide event, not an isolated one.

Major financial outlets have not connected the two. Reporting by The Wall Street Journal, the Financial Times, Reuters, and Bloomberg documented the transaction without linking the Citadel Securities note to it or alleging that the note was used to depress AI shares. No regulatory inquiry into the sequence has been publicly reported.

Finally, the transaction fits an established business line rather than an anomaly. Citadel and JPMorgan absorbed the energy portfolio of Amaranth Advisors in 2006 after its natural gas positions collapsed; Citadel took on Sowood Capital’s book in 2007. Acquiring a distressed portfolio in a single trade removes a forced seller from the open market, which is materially different from the 2021 Archegos unwind, when banks sold competing blocks and several absorbed heavy losses.

Impact: The Leverage Question Underneath the Conspiracy Question

Whatever conclusion readers reach about intent, the mechanical story is not in dispute. A fund carrying roughly four times leverage, in a crowded and publicly documented trade, encountered a drawdown it could not fund. Prime brokers demanded collateral. The fund did not have it. The book was sold.

That failure mode has a long history and does not require a villain. Archegos Capital Management collapsed in 2021 under the same arithmetic; its founder, Bill Hwang, was subsequently convicted. The distinguishing feature in 2026 is speed — the position moved from record performance to complete liquidation of public exposure in six trading days — and the degree to which the fund’s own transparency became its vulnerability.

It also raises a question that regulators have not resolved since Archegos: prime brokers extended the leverage that made the collapse possible, then executed the liquidation that resolved it. Those incentives have not meaningfully changed.

Analysis: What Would Have to Be Shown

For the manipulation claim to move from allegation to finding, several specific things would need to surface. None currently has.

Evidence that the Citadel Securities macro view was not sincerely held — internal communications contradicting the published note, for example.
Evidence that information or direction crossed the information barrier between Citadel Securities and Citadel the hedge fund.
Evidence of the hedge fund’s positioning in the affected names before July 27, which would establish whether it stood to gain from a decline it was allegedly encouraging.
Disclosure of the block trade’s pricing relative to prevailing market levels, and of Citadel’s associated hedges.
An enforcement referral, SEC inquiry, or FINRA information-barrier examination arising from the sequence.

Absent those elements, what the public record contains is a strikingly timed coincidence within a sector-wide selloff that had multiple independent causes — and a firm that has, three times across twenty years, been the buyer of last resort when a leveraged fund ran out of collateral.

Conclusion

The four-day sequence is real, and readers are entitled to weigh it. A firm’s research arm argued for a hawkish surprise; the surprise did not come; the firm’s investment arm bought a forced seller’s entire book three days later at an undisclosed price. Reasonable people can look at that and conclude the incentives were too aligned to be comfortable.

Equally, reasonable people can note that the Fed genuinely was closer to a hike than markets assumed, that the AI selloff had causes that had nothing to do with Chicago, that a four-times-levered fund in a publicly mapped trade was going to break on its own, and that no reporting or regulatory action has connected the note to the transaction.

NexfinityNews has requested comment from Citadel, Citadel Securities, and Situational Awareness LP, and will update this article with any response. The question of whether the sequence was engineered or coincidental remains open. This publication does not claim to have answered it.

Key Takeaways

On July 27, a Citadel Securities strategist publicly argued the Federal Reserve could deliver a surprise quarter-point increase, calling it a potential cleansing event for markets.
The Fed held rates at 3.5 to 3.75 percent on July 29, with three of twelve voters preferring an increase.
On July 30, Citadel — a separate firm from Citadel Securities, though both founded by Ken Griffin — bought roughly $16 billion of Situational Awareness LP’s public equity book in one pre-market block trade.
Situational Awareness had returned about 439 percent in the first half of 2026 and peaked near $45 billion before July losses on roughly four-times leverage forced liquidation.
Traders including Futurum Equities strategist Shay Boloor have alleged the sequence amounted to manipulation. No filing, regulatory action, or major-outlet report has substantiated that claim.
July’s AI selloff had multiple independent causes, including Meta Compute, Intel’s 18A delay, and a hawkish Fed under Chair Kevin Warsh, with inflation at 4.2 percent.
Federal manipulation claims require proof of scienter and of information crossing the barrier between the two Citadel entities — neither of which is present in the public record.

Sources

CNBC, “Leopold Aschenbrenner’s Situational Awareness fund: $45B to fire sale,” July 31, 2026.

CNBC, “AI investor Leopold Aschenbrenner forced to unwind all public stock positions after steep losses,” July 30, 2026.

CNBC, “Why Leopold Aschenbrenner’s Situational Awareness hedge fund imploded,” July 31, 2026.

The Wall Street Journal, “Citadel Buys Situational Awareness’s Stock Portfolio After Big Losses in AI,” July 30, 2026.

Financial Times, reporting on Situational Awareness H1 investor letter and public-equity holdings, July 24 and July 30, 2026.

Reuters, “Citadel buys most of Situational’s stock holdings after AI share rout,” July 30, 2026.

Bloomberg, “Citadel Snaps Up AI Stocks From Situational Awareness Amid Rout,” July 30, 2026.

Citadel Securities Global Macro Strategy, Frank Flight, “Fed Views: The Case for July,” July 27, 2026.

Federal Reserve, FOMC statement and implementation note, July 29, 2026.

Disruption Banking, “Did Citadel Buy the Dip in Situational Awareness’s Leveraged AI Unwind?” July 30, 2026.

SpotGamma, “Anatomy of a Margin Call: How Situational Awareness LP Unwound a $20 Billion AI Book in One Trade,” July 2026.

CNN Business, “The market’s big AI doubts are exposing the riskiest players,” July 31, 2026.

Securities Exchange Act of 1934, Sections 9(a)(2) and 10(b); SEC Rule 10b-5.

Shay Boloor (@StockSavvyShay), posts on X, July 30, 2026.