Citadel Sold the $4 Billion Book It Took Over in Three Weeks

The Letter That Arrived on Friday

Ken Griffin wrote to Citadel investors on Friday, August 21. The letter described a single piece of work: more than 80% of the aggregate risk in the equity portfolio the firm had acquired from Situational Awareness in late July had been unwound.

The method was this: roughly 100 block trades, more than $4 billion in market value, and the largest single-session block sales of the year in ten separate names. Total elapsed time, three weeks. Reuters put the trade count at "nearly 100" and CNBC at "more than 100"; both agree on the $4 billion figure.

The sentence quoted from the letter describes the whole business:

"Our ability to distribute this risk was central to our investment thesis."

That is not a boast. It is a job description. Citadel did not buy the portfolio because it was cheap; it bought the portfolio because it could sell it. The difference between those two things is the subject of this piece.

What Was Acquired

Situational Awareness was the fund Leopold Aschenbrenner founded in 2024. It reported a net return of 439% for the first half of 2026, peaked at roughly $45 billion in assets in early July, and ran leverage of as much as four times. When AI hardware stocks sold off sharply in the last week of July, margin calls followed, and on the morning of July 30 the fund disposed of its entire listed equity book in a single transaction. Citadel was the buyer. We covered that collapse itself in a separate piece.

The 13F filed on August 14 showed what had changed hands. As of June 30 the fund's disclosed U.S. long positions came to roughly $20 billion, and two names accounted for more than 56% of the total.

That distribution explains why the transfer was difficult. A single $5.5 billion position in a name like SanDisk or Micron sits far above those stocks' daily traded volume. A package of that size cannot simply be handed to the market at the prevailing price.

The Economics of a Block Trade

The difference between working a large position through the market over time and handing it to one buyer at once comes down to who absorbs the market impact.

For the seller, an outright transfer buys certainty: the price is settled today and tomorrow does not matter. For the buyer, the reverse. The moment the package is taken on, the buyer is exposed both to the direction of prices and to the pressure its own selling will create. The discount it pays is the up-front price of those two risks.

Profit is the difference between the two, and all of it depends on shrinking the second. The arithmetic makes this concrete.

None of these inputs were disclosed; the percentages above are assumptions chosen to show the arithmetic. Citadel's purchase price was never made public, and the Wall Street Journal did not report a price in its account of the deal. The structure holds regardless: the discount is fixed, market impact varies with the seller's skill, and profit lives entirely in the second term.

The three-week schedule and the roughly 100 separate blocks exist for exactly this reason. Pushing the same $4 billion into a single session would have multiplied the impact. Breaking it into pieces, placing each with a different set of buyers, and spreading the work over time is the standard way to reduce the cost of finding a bid. The line Griffin devoted to the banks follows from the same logic:

"A transaction of this magnitude could not have been completed without the extraordinary cooperation of the trading and prime brokerage teams at the banks serving both firms."

The Year's Largest Block, in Ten Names

One detail in the letter conveys the scale on its own: Citadel executed the year's largest single-session block sale in ten separate stocks. That is not a list of records. It is a description of a constraint.

The ceiling on how much of a stock can be sold in one session is set by that stock's daily volume. When there are not enough buyers on the other side, the price falls faster than the quantity sold would imply. Large packages are therefore usually spread across days. Printing the largest block of the year means going to the edge of what the market could absorb in that name that day.

Doing it in ten different names is a function of how concentrated the book was. The weight sat in a handful of stocks, where the position ran to multiples of an ordinary day's volume. Working in small pieces would have turned three weeks into months, and the price risk would have stayed with the buyer the whole time. Three weeks was not a display of speed but a choice between those two costs: accept slightly more impact, hold the risk for less time.

What Happened on Transfer Day

The effect of removing a forced seller showed up directly on July 30. The transferred names rallied far above the indices that day.

The move was attributed not to any new company news but to the change of hands. The market had known for days that a forced seller was working in these stocks; once the seller was gone, so was the pressure. The same logic runs in reverse: knowing a buyer still has inventory to place leaves an overhang on the price.

The chart shows only the general path of the period. It cannot separate which moves came from block sales, for a simple reason: which stocks were sold on which days was never disclosed.

Who Sold Friday's Decline

Micron, SanDisk, Bloom Energy and Nebius all closed lower on the Friday the letter was published. Linking the two is easy and wrong.

A letter is a disclosure, not a transaction. The sales it described had already been executed over three weeks. CNBC's account is explicit on the point: Citadel did not disclose which stocks it sold or when, so its role in the declines is unclear. Bond yields were rising the same day and the market was waiting on Jackson Hole.

The index is here to show how much of the move in individual names belongs to the market as a whole; it carries no trace of the block sales.

The distinction that matters for risk management is this: the day a piece of news is announced is not the day the event occurred. Prices move on the second. Headlines print on the first.

The Two Sides of the Trade

The July outcome row needs care. How much of that 5.94% came from this trade and how much from the firm's other strategies was never broken out. The fund is up 12% for the year and managed roughly $71 billion as of July 1; there is no way from the outside to calculate how many points a single transaction added to a pool that size.

There is also a side that does not show. The acquired book was not only equities; the fund's first-quarter filing showed billions of dollars in put option positions. The block trades describe the equity leg, and how the derivatives leg was closed was never disclosed. Unwinding 80% of a portfolio's risk is not the same as selling 80% of its positions; hedges can stay on while the underlying stays on the books. The figure in Griffin's sentence is 80% of aggregate risk, not of positions.

The Remaining 20%

The number Griffin reported is more than 80%, not all of it. The remaining risk is still on Citadel's books, and it could be there for either of two reasons: the least liquid positions were left for last, or the firm chose to keep that piece. Which one was not disclosed.

The distinction matters. The first means distribution is still under way, and that those stocks have a seller in them for a while yet. The second means Citadel is no longer an intermediary in those names but an owner of them. From the outside the two look identical. They do not end the same way.

This piece draws on Reuters and CNBC reporting based on Griffin's investor letter, the 13F filed on August 14, and secondary accounts describing the structure of the block trades. Citadel's purchase price and the dates of its sales have not been made public; the percentages in the mechanism section are assumptions chosen to illustrate the arithmetic. The discrepancy between sources on the trade count is noted in the text.