CME to Launch GPU Computing Power Futures on October 5, Tracking H100 and B200 Leasing Costs

As investment in AI infrastructure continues to grow, GPU computing power is gradually shifting from a basic infrastructure cost within the AI industry chain to a financial asset that can be priced, traded, and hedged.

CME Group and GPU market data provider Silicon Data announced on Tuesday that they plan to launch two computing power futures contracts on October 5, pending regulatory approval.

The two contracts are the Silicon Data H100 Rental Index Futures and the Silicon Data B200 Rental Index Futures, both to be listed on CME and subject to the rules of the New York Mercantile Exchange (NYMEX).

The contracts will track the H100 and B200 GPU rental price indexes compiled by Silicon Data, using hourly GPU rental cost as the core pricing metric.

This means that in the future, market participants may be able to hedge against AI computing price volatility through futures instruments. According to CME, the new products are designed to help traders, financial institutions, AI developers, and cloud service providers manage the price risks arising from the rapidly growing computing power market.

CME previously referred to computing power as "the new oil of the 21st century," stating that computing resources are becoming an independent and emerging asset class.

GPU Rental Price is Becoming a Key Cost Indicator for the AI Industry

For AI model training and inference companies, in addition to capital expenditures such as purchasing GPUs and building data centers, directly renting GPUs from cloud vendors or specialized computing providers is also a major way to access computing capacity.

Therefore, changes in the hourly GPU rental price can, to some extent, reflect the supply and demand situation for computing power.

Silicon Data currently publishes daily benchmark GPU rental prices covering major AI accelerators including H100, A100, B200, and AMD MI300X, based on standardized prices compiled from data provided by cloud service providers, hyperscale cloud vendors, colocation data centers, and private rental markets.

B200 is one of the core products of Nvidia’s Blackwell architecture, offering higher computational performance compared to the previous generation H100. Since Blackwell production and deployment are still expanding, B200 rental prices remain affected by both supply constraints and strong AI training demand.

According to Silicon Data, the B200 rental price index reflects the standardized hourly price from various computing power supply channels.

The Futures Market Aims to Establish a Public Forward Price Discovery Mechanism

Currently, the GPU rental market is highly fragmented, with noticeable price differences between different cloud vendors, service providers, and rental terms. Enterprises also lack mature tools, such as crude oil futures, to lock in future computing costs.

When CME and Silicon Data first announced their partnership in May, they stated that the new contracts are intended to help AI companies and cloud service providers manage the risks of computing price volatility. Silicon Data subsequently introduced a GPU forward price curve covering models such as H100, B200, and A100, with terms up to 36 months.

For computing power suppliers, if they expect GPU rental prices to fall in the future, they can use futures for hedging; for AI companies that need to rent large quantities of GPUs, futures may provide tools to lock in future computing costs.

At the same time, financial institutions and traders can also participate in price trading according to their assessments of AI demand, GPU supply, and data center construction cycles.

This also means that the financialization of the AI industry is extending beyond chips, data centers, and related stocks to encompass the underlying computing resources themselves.

If these CME products are launched smoothly and accumulate enough liquidity, H100 and B200 futures prices may become important market indicators for observing AI computing supply and demand, data center utilization, and the return on AI infrastructure investment.

However, there is a key difference between computing power and traditional commodities: GPU computing power cannot be stored like oil or gold. Once computing power is idle, its corresponding hourly service value vanishes. Therefore, the pricing mechanism, liquidity, and hedging effectiveness of GPU futures will differ significantly from conventional commodity futures.

A recent study on AI computing asset pricing also noted that because computing power cannot be stored, the arbitrage-free pricing relationship in traditional commodity futures cannot be directly applied to the computing market.

As a result, if the two contracts are launched as scheduled on October 5, their significance may lie not only in the introduction of two new futures products, but also in creating a public financial benchmark for AI computing power, which previously lacked a unified market price.