The "son of the version" in U.S. stocks may be falling for the second time on the same wave of AI stocks. His name is Leopold Aschenbrenner, who previously researched superintelligence safety at OpenAI and later founded an investment firm. His investment approach sounds particularly solid: he doesn’t bet on which model will win, but instead buys only what all AI companies must purchase—storage, cloud computing power, and data center infrastructure. But these stocks all rely on the same pool of capital. By the end of June, Micron and SanDisk alone accounted for more than half of his publicly disclosed holdings, and he had taken on significant leverage. In the first half of the year, he netted a 439% profit—turning $1 million into roughly $5.39 million. When chip stocks dropped in July, margin calls immediately followed. He sold most of his equity positions to hedge fund Citadel. His portfolio plunged 67% that month, leaving him with just around $1.8 million from his $5.39 million peak. In September, JPMorgan Chase cut off his lending line. He didn’t stop, however—he turned to buying call options on stocks like AMD and CoreWeave, including deeply out-of-the-money options: unless the stock surges sharply in the short term, these options expire worthless. Then in November, multiple U.S. states issued red flags for data centers, triggering another wave of selling in AI infrastructure stocks; CoreWeave fell 65% from its June high. Market rumors once again spread that he had been liquidated, though no exact loss figures have been confirmed. The direction he bet on may have always been correct. Unfortunately, options have expiration dates.