CME Group Inc is working with index provider Silicon Data to launch futures contracts tied to computing power, in what could become a new tradable asset class linked to the global artificial intelligence boom, according to a report by Bloomberg.
The proposed contracts would allow investors, AI developers, cloud providers, and financial institutions to hedge and speculate on the cost of compute – the processing capacity that underpins training and running modern AI systems. The initiative is still subject to regulatory approval.
The development reflects a broader push to formalise pricing for computing resources, which have become increasingly scarce amid surging demand for AI infrastructure. Compute shortages have led to usage caps for some services and driven massive capital investment in data centres and high-performance chips.
Industry figures have increasingly described computing power as a foundational commodity of the digital economy. CME chief executive Terry Duffy said compute is emerging as a core asset class, comparing its role in the AI era to energy in previous industrial cycles.
The concept has also gained traction among major financial figures, with Larry Fink recently suggesting that futures markets tied to compute could emerge as scarcity and demand intensify across the AI ecosystem.
Futures contracts, widely used in commodities markets such as oil and metals, enable investors to lock in prices or hedge exposure to future volatility. CME’s proposal would extend this framework to GPU-based computing capacity, a resource increasingly central to AI model development and deployment.
Silicon Data, which provides benchmarking for GPU rental pricing, has developed indices tracking the cost of on-demand compute, including hourly pricing for high-performance chips used in large-scale model training. These benchmarks are expected to serve as the reference price for the new contracts.
Market participants argue that the creation of a standardised futures market could improve price transparency in a segment that has so far remained fragmented and opaque, while also providing a formal risk-management tool for AI infrastructure costs.
However, the move also raises questions about whether compute is ready to be treated as a fully financialised commodity, with some market observers warning that introducing derivatives could amplify speculative flows into already tight infrastructure markets.
Silicon Data, founded by former DRW trader Carmen Li and backed by DRW Holdings, has been building historical and real-time indices that reflect GPU rental costs across major cloud and data centre providers.
The Silicon Data H100 Rental Index, which tracks the cost of renting high-end GPUs used for AI training workloads, is expected to form a key reference point for the planned futures contracts.
