Turning AI computing power into a tradable commodity: new oil?

Industry Context: Volatility in AI Computing Costs Spurs New Risk Management Tools

Over the past decade, companies have managed price risks for fuel, agricultural products, and metals through futures markets. Now, the explosive growth of AI infrastructure has made computing power a critical input for enterprises. Since most companies do not directly own high-end GPUs but instead rent computing power from cloud services or "new cloud" providers, GPU rental prices fluctuate dramatically with supply and demand, creating significant budget uncertainty. CME Group, in collaboration with pricing data company Silicon Data, plans to launch the world's first set of futures contracts tied to AI computing power, aimed at helping companies lock in future computing costs. The contracts are pending regulatory approval.

Market Impact: From Hedging Tool to Asset Class

Within days of the announcement, asset management firms ProShares and Rex Shares submitted applications for ETFs based on these futures, including leveraged and inverse products, indicating that investors view AI computing power as a tradable asset class. Silicon Data founder and CEO Carmen Li believes the market could eventually surpass oil futures, as AI-related energy demand will exceed the sum of all other energy uses. If approved, the contracts will profoundly impact AI companies' financial planning—companies can lock in costs in advance, while suppliers with large GPU capacity can hedge against price declines. Additionally, SpaceX has already cited Silicon Data's GPU rental price index in its IPO prospectus, showing that its data is being adopted by mainstream enterprises.

Competitive Landscape: Winners and Losers

• Winners: GPU manufacturers like NVIDIA will gain more stable demand signals; cloud service providers (such as AWS, Azure, Google Cloud) and "new cloud" providers can manage their capacity risks; financial institutions (e.g., hedge funds) can access new profit sources through speculation.
• Losers: Small AI startups may face greater financial pressure from increased computing cost volatility, but the futures market can also provide hedging tools; traditional commodity exchanges (e.g., ICE) need to accelerate their AI-related product offerings.
• Potential Followers: Other data providers (such as Luxury, Banana Dev) may launch competing indices, promoting market maturity.

Enterprise Implications: Enterprises Need to Rethink Computing Procurement StrategiesEnterprises should focus on the following three points: 1. Hedging Strategy: Over the next 12-24 months, large AI companies can assess whether to use futures to lock in training and inference costs, similar to how airlines manage fuel costs. 2. Benchmark Reliability: With over 50 GPU configurations (including multiple variants of the NVIDIA H100 alone), Silicon Data constructs an index by standardizing and converting them into "benchmark H100 cases". Enterprises need to understand the index compilation methodology to ensure it reflects the type of computing power they use. 3. Regulatory Dynamics: The CFTC will strictly review contract specifications, settlement mechanisms, and benchmark transparency. Enterprises should track compliance progress and adjust internal financial models in advance.