The fund's book grew +148% QoQ to $13.68B notional. Roughly 62% of that is bearish put exposure on the most crowded semiconductor names — Nvidia, Broadcom, AMD, Oracle, and the SMH ETF. The long book leans into physical AI: power generators, crypto miners pivoting to GPU hosting, and memory.
AI compute requires gigawatts of continuous power. Leopold is long the unsexy backbone: Bloom Energy (on-site fuel cells), crypto miners pivoting to GPU hosting (CoreWeave, IREN, Core Scientific, Applied Digital, Riot, CleanSpark), and memory (SanDisk, Micron calls). Thesis: the bottleneck is physical, not algorithmic.
The puts target names where consensus multiples have run too far: Nvidia, Broadcom, AMD, ASML, Oracle, TSM, and the broader SMH ETF. The bet is not that AI fails — it's that these specific stocks have priced in too much, and a single missed quarter or hyperscaler capex pause re-rates them violently.
| Ticker | Company | Type | Value | % book | Shares / Contracts | Status | Why it's there |
|---|
Data on this page is from the Q1 2026 13F filing (period ending 3/31/2026, filed 5/18/2026, accession 0002045724-26-000008). Verify and refresh from:
To update: edit data.js (or the inline DATA array in this file) when the next 13F drops — roughly mid-August 2026 for Q2 2026.