A ChainCatcher article by 贩财局 describes a sharp reversal in a leveraged AI stock trade, arguing that the real vulnerability was not the investment theme itself but a combination of high leverage, transparent positioning and weak liquidity.
The article identifies the central figure as Leopold, a 24-year-old German. It says he entered university at 15, graduated first in his class from Columbia University at 19, and later worked on AI safety at OpenAI. After being fired in 2024, he wrote a 165-page essay predicting that AGI would arrive in 2027. The piece says that narrative helped him raise $45 billion with an eight-person team, then deploy the capital with 4x leverage into AI hardware-related trades.
Small-cap AI names became the core of the trade
According to the article, Leopold began building large positions in Nebius, CoreWeave and SanDisk from April this year. The report says these were relatively small-cap names that had not drawn broad market attention before his buying accelerated.
With $45 billion and 4x leverage, the article says, the purchases helped drive those stocks higher. It puts SA’s first-half return at 439% and says some early investors posted gains of 2,000%.
The same article says a number of Wall Street firms had been short those stocks because they viewed the valuations as excessive. As prices climbed, those short sellers were forced to cover, which added more fuel to the move. It also says Leopold publicly shorted traditional software stocks including Adobe, putting him directly at odds with more established investors.
What Wall Street saw: visible holdings, high leverage, limited liquidity
The article argues that large firms did not need to move quickly because SA’s weakness was already visible. In its telling, market participants knew what the fund owned, knew it was running 4x leverage, and knew the names in the portfolio were not especially liquid.
That mattered because once prices started to fall, exiting would be difficult. The report frames the setup as a fund that had effectively shown the market its cards while relying on leverage to keep the trade working.
Pressure built in early July
The article says discussion began circulating in early July that AI valuations were too high and that a bubble might be close to breaking. After that, it says, Wall Street money started selling the names that SA owned in size. The initial declines were not dramatic, but the report stresses that with 4x leverage, a 25% drop in the underlying would wipe out the equity.
At the same time, the traditional software stocks Leopold had shorted started to rise. That left the fund under pressure on both sides of the book. Margin calls followed, according to the article, as banks focused on whether the collateral remained sufficient.
The report also says Citadel’s macro team publicly stated that an unexpected Federal Reserve rate increase was under discussion and shifted its base case from no hike to a hike. Bloomberg and the Financial Times were among the outlets that covered that view, the article says.
In the article’s account, that added to market tension and pushed SA’s leveraged positions closer to collapse.
Forced selling on July 29 and Citadel’s reported purchase
The piece says SA was hit with margin calls on July 29 and had to sell. Leopold then sought help in the market, looking to unload more than $10 billion in stock, according to the article.
It says Millennium reviewed the positions and declined, while Jane Street, described as one of SA’s investors, was also unwilling to pay a high price.
Citadel then stepped in, the report claims, and bought roughly $16 billion of SA’s entire disclosed public holdings at a discount. The article presents that moment as the turning point: once the market understood the seller was under pressure, bargaining power shifted heavily to the buyer.
Rates stayed unchanged on July 30, and the stocks rebounded
According to the article, the Federal Reserve left rates unchanged on July 30, one day after the forced sale. That removed the immediate rate-hike overhang described earlier in the piece.
The report says the stocks Leopold had owned rose 15% to 29% that same day, while short sellers moved to cover. On that basis, the article argues that if SA had been able to hold on for another 24 hours, the outcome could have looked very different.
At the same time, the article says Citadel did not rely on fabricated claims but publicly stated its own view. It contrasts a firm with $71 billion under management and more than 30 years of operating history with an eight-person fund that is less than two years old, running 4x leverage and holding positions the market could clearly identify.
The article’s conclusion centers on leverage, not the AI call itself
贩财局 concludes that Wall Street was not “killing AI” or even Leopold’s core thesis. In its view, what got punished was the structure of the trade. The direction may have been right, the article says, because the stocks rallied the day after the liquidation, but being right on direction was not enough if the portfolio could not survive the drawdown.
The piece adds that Citadel, led by Ken Griffin, had $71 billion in assets under management and enough cash to wait for forced sellers, while Leopold’s fund had existed for less than two years and had not gone through a full bull-bear cycle.
It closes by summarizing the sequence this way: SA exited, Citadel took the other side, and the market rebounded. The article also points to a valuation comparison between SK Hynix and Coca-Cola to argue that, in a margin-driven selloff, fundamentals do not necessarily set the price in the moment. Liquidity and pricing power do.