Two of America’s derivatives exchanges are building futures markets for computing power, and a third has said it will offer a regulated contract of its own. Kalshi and CBOE are racing to establish products that let buyers and sellers lock in the future price and availability of AI compute, while CME has announced plans for the first regulated compute futures contract. The forward market, proponents say, could eventually be worth $100 trillion.
The ambition is stated plainly by the people building it. Tarek Mansour, the chief executive of Kalshi, has said computing power will surpass oil as the world’s largest commodity, and that a futures market is the natural next step in its financialization. Oil has a century of futures trading behind it; compute, the argument goes, is where oil was in the 1960s, with demand exploding and no way to hedge the price.
The mechanics are still being worked out. A futures contract needs a standardized underlying asset, and computing power is famously heterogeneous: a data center in Texas is not the same as one in Virginia, and the chips that run AI models differ in ways that affect their price and availability. The exchanges are proposing different solutions, from contracts tied to the price of renting standard configurations of computing capacity to indexes that track the cost of AI compute across providers.
The demand for such contracts is real. Companies that build AI models and run AI services face a cost structure dominated by computing, and their ability to plan depends on what that computing will cost months from now. Cloud providers, which are spending billions on data centers, want to hedge against a drop in prices; their customers want to hedge against a rise. A futures market would let both sides lock in terms.
The exchange race is also a race for legitimacy. CME, the world’s largest derivatives exchange, entering the space gives the products institutional credibility that a startup cannot match, and its announcement has forced Kalshi and CBOE to move faster. Each exchange believes the first to establish a liquid market will set the standards, the contracts, and the pricing benchmarks that the whole industry will use.
The exchanges are also competing for the data that will underpin the contracts. A futures market needs reliable price discovery, and the price of computing is not published anywhere: cloud providers quote bespoke rates to individual customers, and the terms are confidential. The exchange that can assemble the best picture of what compute actually costs, and persuade participants to accept it as a benchmark, will have an advantage that is hard to dislodge.
The stakes extend beyond the exchanges. If compute futures succeed, the pricing power in AI infrastructure would shift from the chipmakers and cloud providers that set prices today to the exchanges and traders who set futures prices. That is the pattern of every commodity market: the financial layer ends up influencing, and sometimes determining, the prices that physical producers charge.
Skeptics question whether the commodity is ready. Computing is not a single thing, the quality of capacity varies, and the industry has not yet agreed on what a standard unit of compute is. Futures markets have failed before when the underlying asset was too heterogeneous, and the exchanges’ early contracts could be thinly traded if buyers and sellers cannot agree on what they are pricing.
The history of commodity markets offers lessons in both directions. Electricity futures, once dismissed as impractical because power cannot be stored, became a mainstream market; weather derivatives, similarly exotic when introduced, found a stable niche. The difference is usually a single thing: a benchmark that all parties trust. Whether computing gets one depends on the exchanges’ ability to define a standard unit that reflects what buyers and sellers actually transact.
The regulators will have a say. Compute futures fall under the same rules as other derivatives, and both the CFTC and the exchanges’ own risk committees will need to sign off on the contracts. The novelty of the asset, and the concentration of AI infrastructure among a handful of companies, raises questions about manipulation and about whether a few large players could distort the benchmark prices.
The potential size of the market is what draws the exchanges in. If computing becomes the world’s largest commodity, with demand measured in trillions of dollars, then even a small share of the derivatives volume would be a significant business. The exchanges’ costs are low, once the infrastructure is built, and the margins on liquid derivatives markets are high. That math, more than any single contract design, explains the race.
For the AI industry, the development is a sign of maturity. Every major technology cycle has produced its own financial instruments, from the options markets that followed the PC boom to the futures that emerged around electricity trading. A market for computing power would mean that AI has moved from a technology story to an economic one, with prices set not by the firms that sell capacity but by the traders who buy and sell its future. The exchanges are betting that day comes, and that they will be the ones holding the contracts.
