Kalshi, the CFTC-regulated prediction market, launched GPU compute forward curves on July 14 to help data centers and cloud providers hedge against volatile AI infrastructure costs. The curves track projected hourly rental prices for Nvidia’s B200, H200, and A100 GPUs, essentially creating a financial benchmark for the price of raw AI horsepower.
What Kalshi actually built
Kalshi’s forward curves aren’t traditional futures contracts in the way most traders would understand them. They function more like reference prices, aggregating data from various CFTC-regulated event contracts to produce market-derived predictions of where GPU rental costs are heading.
Kalshi positioned the offering squarely at institutional buyers: hyperscalers, enterprise AI teams, and the growing ecosystem of GPU cloud brokers who sit between Nvidia’s supply chain and end users. The platform’s status as a CFTC-designated contract market gave the product a veneer of regulatory legitimacy that pure crypto-native exchanges couldn’t match.
The bigger race for compute derivatives
Kalshi isn’t the only exchange trying to financialize AI infrastructure. CME Group has been developing its own rental-index futures tied to Nvidia H100 and B200 GPUs, with a planned launch date of October 5 pending regulatory approval. ICE has also been exploring compute futures products.
The CFTC is currently reviewing public comments on compute derivatives, with a comment period running through October 20.
