CME Group Integrates GPU Rental Index Into Data Feeds

The exchange is now providing forward pricing for H100 and B200 GPU rental rates to help institutional traders.

Updated on Oct. 15, 2026 in Data Centers

Bold vector editorial illustration of a modular graphics processing unit server component, representing the commoditization of compute resources. — CME Group has incorporated Silicon Data’s GPU rental index into its market data feeds to standardize pricing for institutional compute-resource traders. AI Illustration. Upload story photo >

Live Poll

Will financializing AI hardware rental capacity make essential compute infrastructure more accessible for businesses?

CME Group has added Silicon Data's GPU rental index pricing to its standard market data feeds. This integration allows institutional desks to access forward curves for H100 and B200 GPU rental rates directly within their existing workflows.

Why it matters

By providing standardized GPU rental data, the move allows financial institutions to integrate these metrics into their trading systems without needing custom infrastructure. It represents a significant step toward commoditizing compute resources as an investable asset class.

Silicon Data's index covers 95% of the neo-cloud provider market, with futures contracts sized at 730 GPU-hours each. H100 rental rates currently range from $2.53 to $2.81 per hour, while B200 units cost between $5.45 and $6.01.

The players

CME Group

CME Group is a major global derivatives marketplace that provides platforms for trading futures, options, and cash and OTC markets.

Silicon Data

Silicon Data is a financial technology firm focused on creating benchmarks and indices for the artificial intelligence and cloud computing infrastructure market.

DRW

DRW is a diversified principal trading firm that applies technology and quantitative research to identify and capture trading opportunities globally.

The details

The forward curves, which recalculate daily, reflect the cost of securing GPU capacity over various durations. The market for these rentals currently exhibits backwardation, as traders expect rates to soften as more capacity hits the market.

April 20, 2026: Silicon Data's forward curves went live.
October 5, 2026: Planned launch date for GPU futures contracts.
October 15, 2026: CME Group announced the data feed integration.
Mid-October 2026: GPU rental pricing data was recorded.

The Tech Race

This development mirrors the historical commoditization of physical resources, now shifting toward the digital realm of high-performance compute. It positions Silicon Data against traditional market indices by establishing compute power as a core institutional asset class.

For users and developers, the availability of these indices provides greater transparency into the true market cost of cloud compute resources. While institutional investors may utilize these feeds for hedging, the trend toward rate softening could eventually lower rental costs for end-users.

The takeaway

The move suggests that high-performance computing is becoming an essential commodity that is now being actively priced for financial speculation. Investors should watch how the futures contracts perform as a signal for the broader availability of AI processing power.