The Commodity Futures Trading Commission is asking the public what it thinks about trading derivatives tied to computing capacity. With AI model training consuming GPU resources at a pace that makes electricity grids nervous, US regulators are preparing to treat computing power the way they’ve long treated crude oil, natural gas, and soybeans: as a commodity worthy of its own futures market.
The CFTC’s request for comment, submitted to the White House for review and made publicly available on August 17, signals that US regulators are preparing to treat computing power the way they’ve long treated crude oil, natural gas, and soybeans: as a commodity worthy of its own futures market.
From server racks to trading floors
Back in May, CME Group and data analytics firm Silicon Data announced plans to launch futures contracts linked to GPU rental price benchmarks. The target launch date is October 5, assuming the CFTC gives its blessing.
These contracts would reference daily and hourly benchmarks for GPU rental prices, creating a standardized mechanism for price discovery in a market that has, until now, operated without one. Carmen Li, CEO of Silicon Data, has pointed out that GPU rental markets historically lack standardized reference pricing.
The products are designed to help three distinct groups: AI developers trying to budget for model training runs, cloud providers managing capacity and revenue, and investors looking to gain exposure to computing infrastructure without buying actual hardware.
What the futures market could change
If approved, GPU futures could reshape how the AI industry manages its single largest variable cost. A functioning futures market would allow AI companies to hedge that risk, locking in compute costs the way manufacturers lock in raw material prices. It would also generate transparent, market-driven pricing data, something the GPU rental market currently lacks.
For investors, the contracts open a new avenue for exposure to AI infrastructure. Rather than buying shares in chipmakers or data center operators, traders could take positions directly on the price of compute itself.
The public comment period also gives stakeholders a chance to raise concerns. How do you ensure benchmark integrity when GPU rental pricing varies across providers and geographies? What prevents manipulation in a market this nascent? With an October target for the CME-Silicon Data contracts, the CFTC’s timeline for processing public comments and issuing guidance is tight.