What Compute Futures Are
Compute futures are financial contracts whose value derives from the future price of GPU-hours. Launched on CME (May 2026) and ICE (mid-2026), they let market participants hedge or speculate on the cost of AI compute over time.
What they are
A compute futures contract is a derivative whose value is tied to a reference price for compute capacity at a specific future date. Buyers and sellers agree to a price now for compute that will be delivered or financially settled later.
The contracts have specific dimensions:
- Reference asset: A specific GPU type (H100, H200, B200, RTX 5090) or a composite index.
- Time horizon: Settlement at a specific future month (March, June, September, December, etc.).
- Quantity: A standardized contract size (e.g., 100 GPU-hours, 1000 GPU-hours).
- Settlement type: Financial (cash difference) or physical (actual compute delivery).
Futures vs spot
Spot pricing for GPU compute exists today — Vast, RunPod, CoreWeave, etc. all publish hourly rates. The futures markets extend price discovery to forward dates.
The relationship between spot and futures is the futures curve. If the market expects spot prices to fall (because Blackwell supply increases), futures for later dates will be priced lower than spot. If supply is expected to tighten, futures will be priced higher.
Settlement
Both financial and physical settlement structures exist:
- Financial settlement: At expiry, the contract settles to cash based on the reference index price. The party who bet the right direction pays / receives the difference vs the contract price. No actual compute changes hands.
- Physical settlement: At expiry, the seller delivers the agreed compute capacity to the buyer. Requires coordination with actual compute providers.
Financial settlement is far more common in commodity futures markets. Physical settlement is also offered for specific use cases.
Reference indices
The contract values reference a published price index. Index methodologies aggregate prices from multiple providers and produce a representative number.
- Silicon Data Index methodology (CME contracts).
- Ornn Compute Price Index (OCPI, ICE contracts).
- Future entrants likely to publish competing indices.
Index methodology is non-trivial. The index has to be transparent, manipulation-resistant, and reflect actual transaction prices accurately.
Contract specifications
Each exchange publishes specifications:
- Tick size (minimum price movement).
- Contract size (standardized unit).
- Margin requirements.
- Position limits.
- Settlement procedures.
- Delivery rules for physical contracts.
The specifications mirror the conventions of established commodity futures (oil, electricity, etc.) adapted for compute's specific properties.
Why now (2026)
Compute futures became viable in 2026 because several conditions came together:
- Spot markets matured enough to support reference indices.
- Trading volumes reached levels that justify futures infrastructure.
- Customer hedging needs grew with AI budgets.
- Capital structures at neoclouds created demand for hedging instruments.
- Regulatory frameworks adapted enough to allow the launches.
The 2026 timing is the result of years of underlying market development.
Takeaway
Compute futures are real, live, and structurally important. The next chapter examines how the commodity analogy works and where it breaks.