AI Compute Futures: Brokers Prepare as CFTC Review Continues

Wall Street, accustomed to futures contracts for commodities like oil, corn, cattle, and orange juice, may soon add another to its roster: hourly rentals of Nvidia chips.

Commodity brokers are actively preparing to trade derivatives tied to the cost of artificial intelligence computing power. These new financial products would transform GPU (Graphics Processing Unit) time into an asset that traders can hedge, speculate on, and price for years into the future, pending approval from U.S. regulators.

Among the brokers involved in these preparations are ADM Investor Services, BGC Group, and Nanhua USA LLC. Their stated aim is to facilitate trading in these compute futures once the products successfully navigate the regulatory review process.

The proposed contracts will be cash-settled, meaning no physical delivery of graphics cards will occur upon expiration. Instead, financial settlements will be based on the performance of a designated price index.

The benchmark for these contracts will be the hourly rental cost of Nvidia GPUs, specifically focusing on the H100 and B200 models.

CME Group, in collaboration with Silicon Data, is developing cash-settled futures designed to track GPU rental prices. Each contract will represent 730 GPU-hours over a one-month period. This quantity is calculated to be equivalent to one GPU operating continuously for a typical month, effectively mirroring the full-time lease of a single chip.

CME is not the sole entity pursuing this initiative. ICE is also advancing contracts that reference the Ornn and NATIVX indices, while Nodal Exchange has partnered with Compute Desk.

Regulatory Scrutiny

The Commodity Futures Trading Commission (CFTC) formally requested public comment on these proposals on August 19, 2026, with the comment period scheduled to conclude on October 20, 2026.

The regulatory review for CME Group’s product launch has been extended to November 9, 2026. This places the earliest decision point a few weeks after the conclusion of the public feedback period.

Market Anticipation

In anticipation of regulated compute futures, proprietary trading firms, including Jump Trading and DRW, are actively securing capital and establishing dedicated trading desks.

Addressing Compute Uncertainty

AI developers and cloud providers currently face significant uncertainty regarding compute costs. GPU time is a fundamental input for training and operating AI models, and its price is subject to fluctuations driven by supply constraints or the influx of new hardware.

The introduction of standardized futures contracts aims to provide these businesses with a mechanism to stabilize these costs. For instance, a startup planning a large-scale AI training project for the following spring could lock in a rental rate now, mitigating the risk of future price increases. Conversely, cloud providers managing extensive GPU fleets could hedge against potential declines in rental rates, a concern given the rapid depreciation of hardware with each new chip release.

Market Growth Projections

Boston Consulting Group has projected that the AI compute market could expand dramatically, from an estimated $360 billion in 2025 to approximately $2.3 trillion by 2030.

Influence of Index Providers

CME, ICE, and Nodal Exchange are each backing distinct indices. The benchmark that garners the most liquidity early on is likely to become the de facto reference price for GPU time. This positions index providers such as Silicon Data, Ornn, NATIVX, and Compute Desk with considerable influence, as a rental price index underpinning billions in contracts effectively becomes a critical piece of financial infrastructure.

For brokers like ADM Investor Services, BGC Group, and Nanhua USA, the strategic advantage lies in being among the first to market. Should the CFTC approve these products, firms with established clearing relationships and client bases will be poised to begin routing orders immediately.

Potential Risks

Several risks warrant close attention. The reliability of index-based contracts is contingent on the quality of the underlying data, and GPU rental pricing is currently fragmented across various providers, contract durations, and chip models. Furthermore, the focus on H100 and B200 models raises a practical question: how will these contracts accommodate the emergence of newer hardware that could render current benchmark chips less relevant?

Key dates to monitor include October 20, 2026, the closing date for the CFTC comment period, and November 9, 2026, the extended deadline for the CME review.