The AI Genius Who Nearly Lost a $45 Billion Portfolio Leopold Aschenbrenner entered Columbia at 15. Graduated valedictorian at 19. Joined OpenAI’s Super alignment team. Then published a 165-page thesis predicting AGI around 2027 and massive demand for chips, data centers and electricity. He may have been right about AI. Then, with no professional money-management experience, he launched a hedge fund. In less than two years: AUM: $20B+ Portfolio: roughly $45B Employees: 8 Investment professionals: 4 Reported gross leverage: roughly 4× The fund gained 439% through June 2026. Then lost 67% in July. The math is brutal: A 67% loss requires a 203% gain just to recover. The problem was not only the AI thesis. It was the structure: Long AI infrastructure. Short software. Billions of dollars of put exposure. Illiquid private investments. Short-term financing. When both sides of the trade moved against him, Goldman Sachs, JPMorgan and Bank of America demanded more collateral. Barclays had reportedly already refused to finance the fund because of its leverage. To raise liquidity, Aschenbrenner reportedly agreed to sell $3.5B of Anthropic shares then withdrew after Citadel bought most of the public portfolio. The deepest lesson: You can correctly predict the future and still lose everything before it arrives. Insight creates opportunity. Position sizing determines exposure. Liquidity determines survival. The first rule of compounding is not intelligence. It is avoiding ruin.