Read first

Start Here

Eight-chapter deep dive on the brand-new compute futures market. CME + Silicon Data launched May 2026; ICE + Ornn followed. Treating GPU-time as a tradeable commodity changes hedging, capital structures, and the long-term shape of the industry.

Scope & audience

What compute futures are, how the two major exchanges (CME + Silicon Data, ICE + Ornn) are structured, how contracts work mechanically, who uses them (hedgers vs speculators), the risks and open questions, and the longer-term implications for the neocloud industry.

Key framings to carry

  1. Compute futures are a 2026 phenomenon. The CME announcement on May 12, 2026 is the watershed. Anything written before then about compute as a financial commodity is speculative; from now on it's a thing.
  2. The commodity analogy is imperfect. GPUs depreciate, generations roll over (H100 → H200 → B200 → Rubin), and there's no spot price aggregator with the depth of oil or wheat markets. The contract design is genuinely novel.
  3. The biggest impact is on neoclouds, not traders. Hedging revenue and locking forward compute changes capital structure for neoclouds and procurement strategy for AI buyers. The financial-market participation is downstream of those use cases.

Reading order

01 (what they are) → 02 (commodity analogy) → 03–04 (CME and ICE) → 05 (mechanics) → 06 (use cases) → 07 (risks) → 08 (future).