Prediction markets company Kalshi has added the ability to gamble on the compute per hour cost of Nvidia GPUs.

The outcome is verified based on data from Ornn AI, a startup that hopes to build a compute derivatives market that turns compute into a tradeable commodity.

"You can now trade AI compute prices on a CFTC-regulated exchange," Ornn co-founder Kush Bavaria said. "From day one, Ornn has been built for regulated markets, and Kalshi is the only CFTC-regulated prediction market exchange in the US. We believe compute is becoming a commodity class on par with energy and metals, and it deserves the same institutional rigor."

The company tracks live traded spot prices across hardware, including H100, H200, B200, and RTX 5090, to create an index of average compute per hour prices, with H100s currently at $1.70. Kalshi offers the ability to gamble on the price of different GPU compute offerings by the end of March.

While technically Commodity Futures Trading Commission (CFTC) regulated, prediction markets like Kalshi and Polymarket operate in a legal grey area after surging in popularity.

“It is time for clear rules and a clear understanding that the CFTC supports lawful innovation in these markets,” CFTC chairman Michael Selig said in January.

“Consistent with my commitment to fostering responsible innovation in crypto asset markets, I will continue to support the responsible development of event contract markets.”

The markets have faced criticism over concerns about insider trading, gambling addiction, and the risks of turning every action into a gamified and monetizable bet. Kalshi says it bans insiders, and has fined those found to be guilty, including a MrBeast editor.

Ornn raised $5.7 million in late October in a seed funding round led by Crucible Ventures and Vine Ventures, with participation from angel investors who have worked at OpenAI, Palantir, Blackstone, and Coinbase.

The company hopes that, by making compute a tradeable commodity, it will allow AI businesses to lock in training and inference costs through long hedges, data center companies to presell capacity and stabilize revenue through short hedges, and investors to hedge against GPU depreciation.