The Bucket
The single largest bucket in the Q4 2025 13F at ~25 % of reportable value, and the most distinctive one — almost every name began life as a Bitcoin miner and is now repositioning as an AI/HPC hosting business. The fund owns nearly the entire complex.
Thesis
Bitcoin miners turn out to be an unusually good supply-side option for the AI buildout for three structural reasons:
- They already control power. A typical sub-GW substation interconnect with utility-grade transformers takes 4–7 years to build today. Miners hold ~30 GW of energized US capacity that AI workloads cannot replicate fast enough through greenfield builds.
- Their power costs are low. Many sites were sited specifically for cheap, stranded, or curtailed energy (West Texas wind, Pacific NW hydro, Appalachian gas, Quebec hydro). Those characteristics translate directly into HPC operating-cost advantages.
- Their balance sheets are catastrophically undervalued vs. the AI replacement cost of the asset base they sit on. When CoreWeave, OpenAI, or a hyperscaler pays for energized power capacity, the implied EV/MW is multiples of where these stocks trade as miners.
The “HPC pivot” trade is therefore: buy stranded power capacity at miner multiples, hope it gets re-rated at data-center multiples as each name signs an HPC hosting contract. Core Scientific’s $3.5 B+ contract with CoreWeave is the template every other name in the bucket is trying to replicate.
Constituents
Sorted by Q4 2025 value.
| Company | Ticker | Q4 2025 value | % of 13F | Pivot stage |
|---|---|---|---|---|
| Core Scientific | CORZ | $418.7 M | 7.59% | Multi-billion CoreWeave hosting contract signed; pure-play HPC trajectory |
| IREN Limited | IREN | $328.6 M | 5.96% | Texas/BC HPC sites under construction; AI cloud product launched |
| Applied Digital | APLD | $278.0 M | 5.04% | North Dakota HPC campus; CoreWeave 250 MW lease signed |
| Cipher Mining | CIFR | $154.5 M | 2.80% | Black Pearl Texas site convertible to HPC; talks with hyperscalers |
| Riot Platforms | RIOT | $78.1 M | 1.42% | Corsicana TX site; HPC conversion plans announced |
| Hut 8 | HUT | $39.5 M | 0.72% | Pivoted hardest — stood up dedicated AI subsidiary (Highrise AI) |
| WhiteFiber | WYFI | $27.8 M | 0.50% | Bit Digital spin-off, pure-play AI/HPC infra |
| Bitdeer | BTDR | $20.0 M | 0.36% | Norway/Bhutan/Texas; AI cloud business launching |
| CleanSpark | CLSK | $16.6 M | 0.30% | Mostly still mining; smaller HPC pilot |
| Bitfarms | BITF | $16.2 M | 0.29% | New leadership executing HPC pivot at PA site |
| Bucket total | $1,378.2 M | 24.97% |
Position History
The bucket built in two waves. Q1–Q2 2025 established the high-conviction names (CORZ, IREN, APLD). Q3 2025 widened the basket with mid/small-cap miners (CIFR, RIOT, HUT). Q4 2025 completed the “buy the whole bucket” trade by adding the smallest names (BITF, BTDR, CLSK, WYFI). Core Scientific position grew from ~6 M sh on initial entry to ~28.8 M sh by Q4 — clear signal that it’s the bucket’s anchor.
Risk Model
The bucket has three failure modes the fund is explicitly accepting:
- Bitcoin price collapse while the HPC pivot is incomplete — most names still derive >50% of revenue from mining.
- Power-contract counterparty risk — many sites have power-purchase agreements that don’t legally permit a non-mining tenant without renegotiation.
- Construction risk — HPC retrofits require liquid cooling, redundant fiber, and tier-III/IV uptime characteristics that miner sites typically don’t have.
Buying the entire bucket diversifies the binary “which miner signs the next CoreWeave-style deal” question.