Leopold Ash and Brenner's leveraged AI bet

Leopold Ash and Brenner, a 24-year-old promoted by a viral 165-page essay titled 'situational awareness' and a 4.5-hour podcast, raised tech money and built a $45 billion trading book that lost about two-thirds of its value (~67%) in weeks after a July market wobble. He funded the portfolio with roughly $225 million from tech insiders, then "borrowed a great deal more from Wall Street prime brokers" (reportedly 3–4x additional leverage from firms including Goldman Sachs and JP Morgan).

His strategy was a single-theme, directional "AI will change everything" bet: long perceived AI winners (SK Hynix, Sandisk, Bloom Energy) and short firms he thought would lose (Adobe). That positioning did not hedge market risk; as Matt Levine phrased it, the book 'rotated' risk rather than removing it. When AI-exposure sold off and software rallied, both legs lost simultaneously, triggering margin calls during his wedding weekend.

The mathematical failure is volatility drag amplified by leverage: using the speaker's example, an unlevered basket with 15% expected return and 40% volatility compounds at ~+7% after an 8% drag; levered 4x the expected return becomes 60% but drag scales with volatility squared and jumps from 8% to 128%, turning typical compound returns to about -68% per year. Cliff Asness summarized the cultural lesson: 'You never go full Kelly.' Prime brokers and rival funds smelled distress; rival bidders (Jane Street, Millennium) and Citadel (Ken Griffin's $71 billion firm) competed at auction and Citadel bought the public book.

A large illiquid private stake—most notably in Anthropic, where his bride is chief of staff to the CEO—couldn't be margin-called and saved part of the business; Citadel closed first on the public assets and the fund emerged smaller (reported at $8–10 billion). Leopold said he 'removed all of the leverage', offered to call investors, and faced Silicon Valley investors clamoring to redeploy capital (he wired $400 million into a Sequoia-backed startup days later, atop $100 million earlier). The takeaways: concentrated directional leverage on a single thesis amplifies volatility drag, collapses the median outcome even if the mean is high, and multiplies correlated exposure for Silicon Valley limited partners already tied to the same AI bet.