CME Group, the exchange behind everything from oil futures to Bitcoin derivatives, is making its move into the AI economy. The company announced a partnership with Silicon Data to launch what would be the first-ever compute futures market, pending regulatory approval, with a target debut later in 2025.

The logic is straightforward. GPU rental prices for AI workloads fluctuate wildly, and nobody currently has a standardized way to hedge that risk. CME wants to change that by turning compute power into a tradeable commodity, right alongside crude oil and corn.

How GPU rentals became a commodity play

The futures contracts will be built on Silicon Data’s daily benchmarks for on-demand GPU rental rates, described as the world’s first standardized pricing index for compute resources. Silicon Data tracks what it costs to rent GPUs every day, and CME will use those numbers as the foundation for tradeable contracts.

The intended audience spans traders, financial institutions, AI builders, and cloud-service providers. Basically anyone who either pays for compute, sells compute, or wants to speculate on where compute prices are headed.

CME described the target as the “multi-trillion-dollar compute market” in its press release.

Why this matters now

For AI startups and enterprises alike, compute costs represent a significant and unpredictable expense. A futures market would, in theory, let them budget more effectively. An AI company planning a major training run six months from now could lock in today’s GPU rental rates instead of praying prices don’t spike.

On the other side of that trade, GPU cloud providers could use futures to guarantee revenue at current price levels, protecting themselves if the market softens.

CME’s press release stated that “the products will allow market participants to manage volatility and price risk associated with the multi-trillion-dollar compute market.”

The regulatory and structural hurdles

CME operates under the oversight of the Commodity Futures Trading Commission, and any new futures product needs to clear that bar. The CFTC will need to evaluate whether Silicon Data’s benchmarks are robust enough to serve as the underlying index for a regulated derivatives market.

Silicon Data’s daily benchmarks aim to solve this by tracking on-demand GPU rental rates in a standardized way. But the compute rental market is still fragmented across multiple providers with different pricing structures, contract terms, and hardware configurations.

There’s also the question of market depth. Futures markets need enough participants on both sides to function. The early days of Bitcoin futures at CME saw relatively thin volumes before institutional adoption ramped up. Compute futures could face a similar cold-start problem, especially if AI builders aren’t yet accustomed to using derivatives for cost management.