God-like in June, liquidated in July: After Leopold fell, Citadel completed an overly perfect takeover.

In June this year, Leopold Aschenbrenner was still one of Wall Street's most sought-after young fund managers.

His portfolio was widely shared, with investors scrambling to study why he had bought into storage, data centers, power equipment, and companies shifting from Bitcoin mining to AI computing power. Some turned his 13F reports into spreadsheets, while others looked for the "next AI infrastructure bull stock" based on his holdings. The media loved calling him the "AI Prophet" or the "AI Stock God."

By late July, the news suddenly changed: Leopold had been liquidated.

The Situational Awareness fund he managed faced margin calls, forcing it to sell most of its public equity portfolio to Citadel. Even more dramatically, shortly after the trade was completed, the Federal Reserve did not raise interest rates as Citadel had previously predicted. A batch of AI infrastructure stocks that were forced to be sold subsequently rebounded sharply.

From being forced to hand over positions to the market reversing, only a few hours may have passed.

Who is he?

Leopold was born in Germany. He entered Columbia University at age 15 to study economics and mathematical statistics, graduating at 19 with the highest academic honors for his undergraduate class.

His career has been short, but almost every step has landed on the era's windfall. He participated in long-term economic growth research at Oxford University, worked at FTX's Future Fund, and later joined OpenAI's superalignment team to study how to supervise and control AI systems capable of surpassing humans.

In 2024, Leopold was fired by OpenAI for sharing internal materials with external researchers. Both sides took different stands: OpenAI claimed he was involved in leaking secrets, while Leopold believed the real reason he angered management was pointing out to the board that OpenAI's security measures were insufficient to prevent foreign powers from stealing model weights and algorithmic secrets.

Two months after leaving OpenAI, he published a 165-page paper titled "Situational Awareness." The article predicted that AGI might arrive around 2027, triggering an unprecedented industrial construction cycle. What will truly be scarce in the future is not just GPUs, but also storage, power, data centers, transformers, cooling systems, and land that can be quickly connected to the grid.

This article spread rapidly through Silicon Valley and Washington. Subsequently, he founded the Situational Awareness fund under the same name. Early investors included Stripe co-founders Patrick and John Collison, former GitHub CEO Nat Friedman, and investor Daniel Gross.

Thus, Leopold transformed from a young researcher fired by OpenAI into a fund manager overseeing billions of dollars.

He got AI right, but made his positions too heavy

Leopold's investment logic was not complex.

If AI capabilities continue to improve rapidly, tech companies will inevitably invest heavily in building computing power. So, he bought heavily into storage, data centers, power, energy, and AI cloud service companies, while betting that some Bitcoin mining companies would use existing land, power, and grid connection resources to transition into AI computing suppliers.

This approach was once very successful.

According to investor letters, as of the end of June 2026, Situational Awareness achieved a net return of 439% for the year. The fund's scale and equity exposure expanded rapidly. The Financial Times later disclosed that its public equity book reached approximately $16 billion at one point.

However, a 439% return itself is also a warning sign.

For such a massive fund to achieve several times returns in half a year, relying solely on ordinary stock holdings is difficult. Behind this likely lies highly concentrated positions, borrowing leverage, options, or total return swaps.

A bigger problem is that although Leopold appeared to buy many different companies, what he was actually betting on might have been only one theme: high capital expenditure, high valuation, and high financing demand AI infrastructure.

When the market rises, these stocks can rise together; once interest rate expectations rise and funds start deleveraging, they will also fall together. So-called diversification quickly becomes different stock codes for the same trade.

In July, the AI infrastructure sector suddenly corrected sharply, with core targets like Nebius falling by tens of percentage points. Leverage began to bite back, collateral shrank, and lenders demanded additional margin. To raise cash, the fund was forced to sell stocks, which further depressed prices, triggering a new round of margin calls.

At this stage, the decision-making power was no longer in Leopold's hands.

What did Citadel say before taking over?

If the story were merely "a young fund manager over-leveraged and eventually liquidated," it wouldn't be particularly complex.

What truly raises suspicion is Citadel's continuous amplification of the risk of Fed rate hikes over the preceding month, and the timing of its final takeover.

Here, two companies need to be distinguished.

Citadel is the hedge fund under Ken Griffin; Citadel Securities is the market-making and trading institution he founded. Legally and compliantly, they are separate entities, theoretically requiring strict information barriers. There is no evidence that the two companies 违规 exchanged client information in this incident.

But their relationship still makes the entire timeline appear very sensitive.

On June 8, Bloomberg reported that Citadel Securities began warning that the Fed might soon be forced to raise rates, and investors needed to prepare for tighter financial conditions.

The reasons given included overheating AI investment, tight energy markets, strong employment, and rising inflation again. This narrative happens to hit hard against high-valuation, financing-dependent, and interest-rate-sensitive AI infrastructure stocks.

By July 29, on the day of the FOMC meeting, Citadel Securities went further, betting against mainstream market expectations that the Fed would unexpectedly hike rates by 25 basis points.

According to CoinDesk's report at the time, Citadel's argument was not primarily based on inflation and employment data.

Its macro strategy head, Frank Flight, argued that the Fed should use this opportunity to hike rates suddenly, ending the era of forward guidance where "every policy action requires advance hinting," and bringing about a so-called "cleansing event" to the market.

This logic reads rather deliberate.

Normal rate hike judgments should be built on persistently above-target inflation, expanding wage pressures, or obviously overheated demand. But Citadel's reasoning seemed more like: since a rate hike might be needed in the future, better to act suddenly now, because the surprise itself can deter the market.

CoinDesk summarized it directly: this prediction had little to do with where economic data landed; it was mainly a "tactical issue."

However, "cleaning the market" is not the Fed's responsibility. Central banks have no need to deliberately cause asset price volatility via unexpected rate hikes just to prove their independence. If a September rate hike were truly very likely, waiting six weeks to observe more data would clearly be more conventional than suddenly changing policy.

This report alone might not have caused the market drop, but Citadel Securities is not an ordinary analysis firm. As one of the world's most important market makers, its public prediction of an unexpected rate hike easily leads investors to suspect: Does Citadel possess information others haven't seen yet?

Institutions don't necessarily need to believe the Fed will definitely hike rates. As long as they worry other funds will reduce positions early, they have reasons to lower risk first. Panic thereby gains the conditions for self-reinforcement.

An overly perfect trade

What happened next developed exactly along this risk narrative.

Around July 29, Leopold's fund lost the ability to keep waiting due to margin calls. Situational Awareness was forced to sell most of its public equity portfolio, with Citadel ultimately taking a large portion of it.

Business Insider reported that Millennium and Jane Street also participated in the bidding. This indicates that, at least superficially, there was competition in the asset sale, rather than Citadel unilaterally setting the price.

But Citadel undoubtedly held the most crucial advantage: it knew the seller had to close immediately.

Leopold could not wait for the Fed's results, nor could he demand an ideal price. Prime brokers didn't care if these stocks would rebound the next day, but whether the day's collateral was sufficient, whether the fund violated financing terms, and whether holding on would cause greater losses for the lender.

Citadel was completely different. It had a sufficiently strong balance sheet and could hedge short-term risks via index futures, sector ETFs, individual stock shorts, and interest rate products. Leopold had to sell at the worst price, while Citadel had the capacity to wait.

A few hours later, the Fed ultimately did not raise rates. Although three officials supported a hike, maintaining unchanged rates remained the final decision.

With policy risks temporarily lifted, and the massive forced-selling pressure from Situational Awareness gone, multiple AI infrastructure stocks rebounded quickly, with some targets gaining around 20% in a single day.

The whole process formed a very glaring closed loop:

Citadel Securities continuously emphasized the risks of rate hikes and tightening financial conditions; Leopold's high-leverage AI portfolio subsequently came under immense pressure; Citadel took over assets when he lost bargaining power; the Fed did not raise rates, and related stocks rebounded immediately.

Although Citadel's previous macro judgment was wrong, it may have become the biggest beneficiary of this erroneous judgment.

What it truly bet on might not have been "the Fed will definitely hike rates," but rather "will the panic over rate hikes force someone who must sell their positions?"

Leopold ultimately became that person.

How much might Citadel have earned?

It is currently impossible to calculate Citadel's actual profit from this trade.

The outside world doesn't know exactly how many stocks it took over, the discount on the block trade, or whether it simultaneously retained hedges on indices, interest rates, or individual stocks. Any precise number would only be speculation.

But a rough range can be estimated.

If Citadel actually took over $8 billion to $12 billion in stocks, and the portfolio subsequently rebounded by 10% to 20%, the corresponding increase in book value would be $800 million to $2.4 billion.

This does not yet account for the block trade discount.

Leopold was a seller who had to close quickly. For an asset package of such huge scale, so concentrated in holdings, and with falling prices, buyers usually demand a certain discount. If the average discount on $10 billion in assets was 5%, Citadel already secured a $500 million safety cushion at the time of the deal; if the discount reached 10%, it was $1 billion.

Therefore, it is not hard to imagine Citadel obtaining ten-billion-dollar-level book profits in a very short time. Whether it reached $2-3 billion depends on the actual takeover size, rebound magnitude, transaction discount, and hedging costs.

These are merely scenario calculations, not disclosed profit figures.

Leopold might really have been just hours away

Based on the subsequent gains of related stocks, Leopold might indeed have been just hours away from a rebound that could significantly alleviate the margin crisis.

If the fund still had 5x effective directional leverage on the edge of liquidation, then a 20% portfolio rebound would theoretically be enough to double the remaining net value. If the leverage were higher, the net value elasticity would be even greater.

This does not mean that rebound would necessarily save the fund entirely. Situational Awareness might have already violated financing terms, and banks might have already withdrawn credit lines. Even if book prices recovered, prime brokers might not be willing to re-recognize the original collateral value.

But it is certain that once the asset transfer was complete, the subsequent rebound no longer belonged to Leopold.

What Citadel bought was not just stocks, but the waiting time Leopold had already lost.

Was this a precise hunt?

Starting from existing public information, we cannot directly conclude that Citadel manufactured Leopold's liquidation.

Leopold's own high leverage, concentrated holdings, and liquidity mismatch were the root causes of the crisis. If the fund hadn't pushed positions to a level unable to withstand short-term volatility, any macro report wouldn't have brought it down.

The market also wasn't completely excluding a rate hike. Futures prices once gave a probability of about 35.8% for a rate hike, and finally, three Fed officials supported taking action. Although Citadel's judgment was radical, it cannot be deemed malicious rumor-mongering solely based on a wrong prediction.

What truly needs investigation are other issues:

When did Citadel know Leopold was approaching the margin critical point? Was it the financier, derivatives counterparty, or prime broker for Situational Awareness? Before and after publishing its rate hike views, did it short Leopold's core holdings? Was the information barrier between Citadel Securities and the Citadel fund complete? How much of a discount was the final takeover price relative to the market price?

There is currently no public evidence to answer these questions, nor evidence proving Citadel engaged in market manipulation.

But describing it as a coincidental passerby, simply providing liquidity to the market, is equally naive.

A more realistic description might be: Leopold exposed his weakness using excessive leverage, and Citadel likely saw this weakness earlier than others, turning it into an extremely favorable trade when he lacked time and bargaining power the most.

Leopold might have gotten the decade-long AI industrial trend right, but failed to manage the margins for the next few days.

He might have been just hours away from turning things around.

And what Citadel ultimately bought was precisely those few hours.

$Microsoft(MSFT.US) $Sandisk(SNDK.US) $Micron Tech(MU.US) $SK Hynix(SKHY.US)