CME Group will launch two AI compute futures contracts on October 5, pending regulatory approval, in partnership with Silicon Data. The exchange announced the move on August 11 as institutional markets begin pricing exposure to GPU rental capacity as a hedgeable asset class.
The contracts will track rental rates for Nvidia H100 and B200 processors, the two chips driving the majority of large-scale model training and inference workloads. CME offers the financial instrument while a trusted third party maintains the pricing methodology and daily rate feeds.
Silicon Data, which operates GPU rental marketplaces and has published pricing indices on Nvidia compute, will serve as the data provider and underlying reference for contract settlement. Until now, GPU rental rates have moved with spot supply and demand signals, leaving enterprises with no mechanism to lock in future capacity prices or transfer that risk to financial counterparties.
CME already lists futures on energy, metals, and agricultural commodities. Compute futures would sit alongside equity index contracts and bond futures in a product lineup designed for hedging and speculation across asset classes. The October 5 launch date places the contracts within a window when AI infrastructure costs remain high and enterprises are actively negotiating long-term GPU procurement deals.
Regulatory approval is not guaranteed. The Commodity Futures Trading Commission must determine whether AI compute capacity qualifies as a commodity under the Commodity Exchange Act, a question that hinges on whether the underlying GPU rentals meet legal definitions of fungible, standardized goods. CFTC staff have not previously opined on whether compute should be regulated as a commodity futures contract.
Enterprises with committed workload volumes have begun hedging their marginal compute costs through bilateral swap agreements with large cloud providers. A standardized futures contract would allow smaller players and new AI startups to access similar risk management tools.
Silicon Data's involvement means CME intends to build the contract around verifiable market data. If H100 and B200 rental indices become liquid enough to support futures trading, other compute types and processor families could follow. The contract specifications, including notional size and minimum price increments, have not been disclosed.
CME's ability to launch and sustain order flow in compute futures depends on participation from GPU owners seeking to lock in rental income and from AI operators looking to fix future costs. If the contracts achieve sufficient open interest by the first contract expiration, CME and Silicon Data would establish a precedent for financializing infrastructure capacity at scale.
