Compute like Oil: GPU Futures Arrive on Wall Street

Intercontinental Exchange, the owner of the New York Stock Exchange, has announced its intention to launch a series of futures contracts on GPU compute, developed in collaboration with the startup Ornn. The announcement comes a week after a similar agreement signed by competitor CME Group with Silicon Data, marking the simultaneous entry of the two largest U.S. derivatives operators into the computing capacity market for artificial intelligence.

What Are ICE and Ornn Developing?

The contracts will be denominated in dollars and cash-settled, referencing the Ornn Compute Price Index (OCPI), which tracks spot prices traded in the GPU market for key types of hardware. The index currently covers Nvidia chips H100, H200, B200, and RTX 5090, with the possibility of extending to other models as the market develops. The launch is still subject to approval by U.S. regulatory authorities.

OCPI has an important methodological feature: it is built exclusively on actual executed transactions, not on surveys or indicative quotes, and is already distributed on the Bloomberg Terminal. It will become the benchmark rate for cleared GPU derivatives. Ornn presents itself as a financial infrastructure for the compute market and simultaneously operates its own exchange platform for risk transfer on computing capacity. According to the company, the first swap transaction on compute with cleared prices was executed in December 2025.

The Parallel Race of CME and Silicon Data

On May 12, CME Group, the largest derivatives market in the world, announced a parallel initiative in partnership with Silicon Data, a company specialized in GPU benchmarking and backed by Don Wilson's trading firm DRW. CME contracts will reference the Silicon Data H100 Rental Index, which tracks the hourly rental cost of chips used in AI training loads. The CME launch is also subject to regulatory approval, with activation expected by the end of 2026.

The close timing of the two announcements signals that the two main American operators now consider the institutional compute market to be mature. The historical parallel is the 1980s, when ICE Brent and CME WTI competed for the reference price of oil: the exchange that captures the most liquidity in the early months will likely set the industry benchmark.

Competition is not limited to the two giants. In January, Architect Financial Technologies launched perpetual futures traded on GPU and RAM through the AX platform, also using Ornn data, and the prediction market Kalshi has long offered bets on Nvidia compute prices. However, ICE and CME bring something that the new entrants lack: deep institutional liquidity, regulatory credibility, and the clearing infrastructure that large GPU-as-a-Service providers will require.

Compute as a Flow Commodity and Asian Settlement

The design of the contracts faces a challenge that traditional commodities do not have. Unlike oil, which can be stored in tanks, computing capacity is what traders call a flow commodity: it is consumed in real-time and cannot be accumulated. A futures contract settled on the price on the expiration day, as is the case for ICE Brent, would not reflect the actual economic exposure of those renting GPUs.

The Ornn contracts therefore adopt an Asian settlement mechanism: the final price is calculated based on the average of the daily index values over the entire duration of the contract, rather than on the price of a single day. This is the same mechanism used for electricity futures, which is also non-storable. For an AI company planning a thirty-day training cycle, the payoff thus replicates the actual average cost incurred.

The volatility that the market seeks to cover is significant. The spot rental price of Nvidia Blackwell chips, according to the Ornn index cited by the primary source, rose by 48% between mid-February and mid-April 2026, jumping from $2.75 to $4.08 per GPU-hour. For those financing a data center or planning a training budget in the tens of millions of dollars, fluctuations of this magnitude radically alter the economic balance of the project.

The actual launch of both projects remains conditional on the approval of U.S. regulatory authorities. CME has indicated a target activation date by the end of 2026; ICE has not communicated a specific timeline.