Can AI Compute Become The Next Big Futures Market?

Summary

Silicon Data and CME Group are developing a futures contract tied to AI-compute prices, pending regulatory approval. The proposed market would let companies hedge the cost of renting GPUs. Prices can fluctuate as businesses decide how much computing capacity they need and providers decide how many GPUs to buy or offer.

Why hedge

AI companies spend billions of dollars on computing power to train and run models. A company expecting to need more compute could take a long futures position to protect against rising rental prices. A provider renting out GPU capacity could take a short position to protect against falling prices.

Current alternatives

Large businesses may negotiate multi-year fixed-rate agreements with cloud providers such as Amazon, Microsoft, and Google. These agreements offer cost certainty but can tie customers to a particular chip, provider, and price—potentially limiting flexibility in a fast-changing industry.

Proposed benchmark

Silicon Data tracks hourly GPU rental prices across cloud providers and marketplaces. The H100 is described as a key workhorse chip. Silicon Data says it sees more than 150,000 traded prices per day and normalizes them against an H100 baseline. It has identified more than 50 H100 configurations, reflecting differences in processors, RAM, connectivity, and data-center location.

Market participants

The market would need natural hedgers—users and providers—as well as market makers and speculators. Speculation could support price discovery, but liquidity is uncertain. The report notes that relatively few users may currently need high-end GPU access at a scale that warrants futures trading.

Regulatory and contract design

The CFTC would scrutinize the product. Key specifications include contract size, trading hours, settlement timing, and whether settlement is physical or financial. A GPU hour is less standardized than a benchmark commodity unit such as a barrel of oil, making normalization and standardization central challenges.

Outlook and cautions

Silicon Data CEO Carmen Li predicts compute could eventually become larger than oil futures, but this is her view, not an established market outcome. The contract remains subject to regulatory approval, and future compute demand, supply, and market liquidity are uncertain. The report gives no explicit investment recommendation or “not financial advice” disclaimer.

Assets, instruments, and organizations mentioned

• Assets and instruments: Proposed GPU-compute futures, GPU rental capacity, hourly GPU prices, H100 chips, oil futures
• Organizations and products: Nvidia, Amazon, Microsoft, Google, CME Group, Silicon Data, CFTC, ChatGPT

Presenters and sources

CNBC reporter Yun Li; Carmen Li, CEO of Silicon Data. CME Group and Silicon Data are identified as partners behind the proposed contract.