Leopold Aschenbrenner’s Hedge Was the Same Bet Twice

Leopold Aschenbrenner lost an estimated 67% in July even though the S&P 500 barely moved.

His portfolio appeared hedged but behaved as if it carried the same AI narrative twice: long the builders and short the expected losers. When that relationship reversed, both sides lost.

The exact July book is not public, so this is an economic interpretation, not an audited attribution. Still, the investor letter said its positive long-short spread reversed violently, while public reporting identified falling SK Hynix investments and adverse Adobe shorts.

The Same Thesis Twice

A portfolio can be hedged by direction yet concentrated by meaning. Long AI infrastructure and short incumbent software have opposite signs, but both depend on the same relative-value relationship continuing.

The March Form 13F confirms that the fund employed hedges, including semiconductor puts. But it omits short stock, written options, private investments, financing, and option terms; by July it was four months stale. Outsiders cannot reconstruct the loss. We know only that the central relationship failed when needed most.

The Market Did Not Need to Crash

The S&P 500 declined only 0.13% in July. The Nasdaq-100 fell 6.61%, while semiconductors fell 20.61% with a 26.67% maximum closing drawdown. The VIXEQ-minus-VIX gap reached 34.14 points, the widest in their overlapping history since 2014.

The index looked calm because constituent moves partly canceled. Risk lived in the portfolio's coordinates, not in the market as an abstraction.

Return, Drawdown, and Liquidity

The fund estimated a 67% July loss but remained up approximately 80% for the year. A reported 439% gain through June followed by a 67% loss compounds to roughly 78%. Yet recovering requires approximately 203%.

The loss became a financing problem. A block transaction removed leverage, closed shorts, and preserved private positions. In stress, investors sell what they can sell. The public book became the liquidity valve.

A Hedge Should Have a Different Failure Mode

July did not disprove AI. It showed that diversification by ticker is not diversification by thesis. At Runtime, we therefore treat market beta, thematic exposure, position size, liquidity, and financing as separate risks.

A hedge should have a different failure mode from the position it protects. Otherwise, it is the same bet twice.

Engineered compounding,
Deniz Erkan