The New Oil: Efforts to Turn AI Computing Power into a Tradable Commodity
Artificial intelligence is turning computing capacity into one of the most important inputs in the global digital economy. As demand for training and running advanced AI models expands, exchanges and data providers are moving to create financial contracts that could make compute power more transparent, measurable and tradable.
The shift is still at an early stage, but it marks a significant development in the financialization of AI infrastructure. What was once negotiated through private cloud contracts and bespoke capacity arrangements is now moving toward benchmark pricing, futures contracts and risk management tools similar to those used in energy and other commodity markets.
CME and Silicon Data move first
On 12 May 2026, CME Group and Silicon Data announced plans to launch a first in class compute futures market later in 2026, pending regulatory review. CME described Silicon Data as a GPU market intelligence and benchmarking company backed by global trading firm DRW.
The planned contracts will be based on Silicon Data’s indices, which CME described as the world’s first daily GPU benchmarks for on demand rental rates. The objective is to give traders, financial institutions, AI builders and cloud service providers a way to manage volatility and price risk in what CME called the multi trillion dollar compute market.
CME Group Chairman and Chief Executive Officer Terry Duffy described compute as the new oil of the 21st century, arguing that compute is becoming an emerging asset class in its own right. Silicon Data Chief Executive Officer Carmen Li said compute markets remain fragmented, with pricing that can vary across providers, regions and contract structures, and that standardized benchmarks can help turn compute from an opaque operating cost into a more mature financial market.
Why compute is becoming a market
The reason is structural. Large language models, generative AI services and inference workloads require large volumes of GPU capacity. For companies building AI products, compute is not only a technical input but also a major operating cost.
Daily GPU rental benchmarks can therefore serve a similar function to reference prices in energy or metals markets. They do not eliminate scarcity or volatility, but they create a common pricing language that companies and investors can use for budgeting, hedging and valuation.
The comparison with oil is not exact. Oil is a physical commodity with established grades, storage systems and delivery points. Compute is a service based on hardware type, location, power availability, cloud integration, contract duration and performance. That makes standardization more complex, but also explains why transparent benchmarks are becoming more valuable.
Competition expands beyond CME
CME is not the only exchange moving in this direction. On 19 May 2026, Intercontinental Exchange and Ornn announced plans to launch U.S. dollar denominated, cash settled GPU compute futures contracts based on Ornn’s Compute Price Index, subject to regulatory approval. ICE said the index tracks live traded spot prices for GPU compute across major hardware types.
This suggests that compute derivatives may develop through competing benchmarks, different hardware references and different market structures. As with other new asset classes, liquidity will depend on trust in the index methodology, the depth of underlying market data and the willingness of commercial users to hedge real exposure.
China explores a token based model
China is also exploring AI linked futures, but with a different approach. Reuters reported in late May that the Shanghai Futures Exchange is in early stage design work on futures tied to AI tokens, the smallest units of information processed by AI models. Unlike GPU hour contracts, token based instruments would focus more directly on the output and usage of AI services.
The scale of Chinese AI usage helps explain the interest. China Daily reported that daily token consumption in China had exceeded 140 trillion by March 2026, up from 100 billion at the start of 2024 and 100 trillion by the end of 2025, citing National Data Administration head Liu Liehong.
Challenges before compute becomes a true commodity
The opportunity is significant, but the challenges are substantial. GPU markets are not uniform. Prices can vary by chip type, region, provider, availability, power cost, network performance and contract structure. A futures contract will need a benchmark that market participants view as robust, representative and difficult to manipulate.
There is also uncertainty around regulation, settlement design and adoption. Futures markets succeed when they attract both commercial hedgers and financial liquidity providers. For compute, that means cloud providers, AI developers, data centre operators, investors and trading firms must see the contract as useful for managing real economic exposure.
Outlook
Compute futures will not instantly turn AI infrastructure into a mature commodity market, but they could be an important step toward greater transparency and risk transfer. If successful, they would allow AI companies to manage one of their largest and most volatile costs more systematically, while giving investors a new way to express views on demand for AI infrastructure.
The broader message is clear: compute is no longer only a technology resource. It is becoming an economic input with price risk, supply constraints and financial relevance. As AI adoption grows, the market infrastructure around compute may become as important as the chips and data centres themselves.
Sources: CME Group; Intercontinental Exchange; Reuters; China Daily.
