CME Group will introduce two futures contracts tracking GPU computing capacity on October 5, pending regulatory approval, in partnership with data infrastructure firm Silicon Data. The move marks the first standardized exchange-traded vehicle for AI hardware rental costs.
The announcement comes as demand for GPU capacity has outpaced supply across the AI sector. The contracts will track indexes covering H100 and B200 GPU rentals, allowing operators, cloud providers, and financial firms to hedge exposure to compute cost fluctuations. Specific contract specifications including tick size, notional value, and index methodology have not been disclosed.
Silicon Data aggregates real-time pricing data from GPU rental markets across multiple platforms and providers. The firm built the underlying indexes that CME will use to settle the two contract types. GPU capacity constraints have persisted since late 2023, when AI model training and inference demand accelerated faster than hardware supply could match. Pricing for H100 clusters has fluctuated between 2 and 4 dollars per hour depending on contract length and geography, while B200 rentals remain scarce enough that many spot prices are quoted on request.
CME operates the largest derivatives exchange globally by open interest. The firm added bitcoin and ethereum futures starting in 2017 and 2021, respectively, and has since expanded into equity index options, commodity spreads, and energy contracts. The compute futures launch extends this portfolio into infrastructure-layer digital assets.
Microsoft reported in its latest quarterly filing that Azure GPU capacity approached 95 percent in certain regions through mid-2026. No exchange currently offers standardized derivatives on compute capacity, leaving enterprises and data center operators without direct hedging tools for their largest variable cost.
The regulatory timeline for the two contracts remains open. CME typically receives SEC and CFTC sign-off within 45 to 90 days for commodity futures tied to published spot indexes. The October 5 launch date means approval must occur by early September for trading to begin on schedule.
If CME obtains regulatory clearance by early September, compute futures could absorb trading volume currently confined to bilateral negotiation and small-lot spot markets. The two new contracts would represent CME's first direct exposure to the infrastructure layer of the AI value chain, distinct from its existing equity index derivatives tied to semiconductor manufacturers.