CME Group To Launch Futures Market For Nvidia GPU Rental Prices

Wall Street is preparing to turn the cost of renting Nvidia’s most powerful AI chips into a tradable financial product, as demand for computing power continues to surge across the technology industry.

New Contracts Tied To GPU Rental Rates

CME Group plans to introduce futures contracts on October 5, pending regulatory approval, based on the hourly rental price of Nvidia’s H100 and B200 graphics processors. These contracts will settle using benchmark pricing data tracked by Silicon Data, which monitors real-world rental costs companies pay for access to these chips. Current benchmark rates sit around $2.68 per GPU-hour for H100 chips and roughly $5.66 for the newer B200 model, with both figures fluctuating significantly over the past year.

Chip Pricing Behaves Like A Specialized Market

Much like hotel pricing varies based on brand, location, and availability, computing power costs depend on factors such as provider, contract length, and networking setup. Importantly, these futures contracts do not guarantee access to physical chips, only exposure to shifts in benchmark pricing.

History Shows Mixed Results For New Commodity Markets

Attempts to create similar tradable markets in the past, including memory chip futures and bandwidth trading during the early internet boom, largely failed due to a lack of standardized definitions and inconsistent industry exposure. Whether GPU rental futures succeed may depend on similar challenges around consistency and real world relevance.