CME Group (Chicago Mercantile Exchange) announced that its first compute futures contract will be listed on October 5, 2026, still pending regulatory approval. This product will provide a public market reference price for the input resource used by any artificial intelligence application worldwide.
For Bitcoin miners who have spent the past year busy reinventing themselves and transforming into AI data center compute suppliers, this means their role will no longer be just sellers of hashpower, but also producers of a tradable commodity.
The challenge facing the author of a mining company's business plan is quite novel: when compute is traded like oil or electricity, companies that own racks, chips, and megawatt-scale power will be seen as owners of the underlying asset. A futures price directly observable in the market will, for the first time, provide a valuation for the marginal cost of AI compute capacity.
CME treats AI compute as Wall Street's "new oil"
CME first proposed this concept in May of this year, when it announced a partnership agreement with Silicon Data, a GPU benchmarking company backed by trading firm DRW. In August, the listing date was finalized: two contracts to be launched on October 5, subject to regulatory approval. Both contracts are based on Silicon Data's Nvidia chip index. The hourly compute rental price for B200 is about $5.86, while the older H100 is about $2.77. CME's contracts will track the forward value of such prices over the next 36 months.
CME has never been shy about its ambitions. Pete Keavey, CME's global head of energy and environmental products, said when announcing the product launch: "Compute has become the currency of the AI era." CME CEO Terry Duffy made a similar statement in May, calling compute "the new oil of the 21st century." DRW founder Don Wilson predicted that before compute hedging mechanisms are introduced, compute will become the world's largest commodity.
Why compute needs to be priced now
The key to this strategy lies in scale and volatility. Boston Consulting Group (BCG) predicts that the AI compute market will expand from about $360 billion in 2025 to nearly $2.3 trillion by 2030. Against this backdrop of expansion, price volatility is extremely severe: during the semiconductor shortage in early 2024, the rental price of an H100 GPU was about $8 per hour, but by the end of last year it had fallen below $2. This volatility is exactly what commercial buyers want to hedge against.
BlackRock CEO Larry Fink laid out the logic plainly at the Milken Institute conference in May: "There will be a whole new asset class—buying compute futures." He listed compute alongside electricity and semiconductors as assets in obvious short supply in the United States. Brett Harrison, CEO of derivatives platform Architect, estimates that by the end of this decade, the notional contract size of compute futures will reach $10 trillion per year.
Miners already hold the underlying asset
This is where the Bitcoin mining industry comes in. CoinShares' Q1 2026 mining report shows that over the past year, listed mining companies signed AI and high-performance computing (HPC) contracts worth more than $70 billion. CoinShares expects that by the end of this year, AI business will contribute 30% to 70% of these miners' revenue, compared with about 30% currently. In CoinShares' words, TeraWulf, Core Scientific, Cipher Mining, and Hut 8 have become "data center operators that mine Bitcoin on the side."
Behind the intense activity is straightforward economics. Bitcoin mining hardware costs an estimated $700,000 to $1 million per megawatt, while AI systems cost $8 million to $15 million per megawatt. IREN and Bitfarms are shifting toward HPC suppliers. A liquid compute market can give this group of companies what oil drillers and power producers have always had: a tool to hedge their own output.
The benchmark still needs to take shape
Nothing is settled yet. In addition to CME, Intercontinental Exchange (ICE) also announced in May that it plans to launch its own GPU compute futures based on another index from Ornn. Architect acquired a CFTC-regulated futures exchange to develop the American Innovation Exchange. China is also considering launching its own version of AI compute futures.
The real challenge is that two-thirds to three-quarters of futures contracts ultimately fail to generate sufficient profits. Compute itself is heterogeneous—an hour on an H100 is not the same as an hour on a B200; the two key indices sometimes disagree on valuation; and the supply side of this commodity is Nvidia and a handful of hyperscalers. Whether the market ultimately validates or disproves this AI boom, the value of compute futures as a signal may be greater than its value as a market.
