Wall Street is moving closer to treating artificial intelligence computing power like a tradable commodity.
The shift could reshape how data-center operators value and manage expensive GPU capacity. Bitcoin miners moving into AI infrastructure may gain a new tool for protecting future revenue. CME Group’s planned compute futures could provide the market with a public reference price for that capacity.
CME plans to launch its first compute futures contracts on October 5, 2026, pending regulatory approval. The contracts will track Nvidia GPU compute pricing for periods extending as far as 36 months.
That structure could give operators an observable forward price for AI compute capacity. It may also help companies hedge swings in the value of racks, chips, power, and data-center infrastructure.
For Bitcoin miners, the development carries added relevance. Several listed miners have spent the past year expanding beyond hashpower and into AI data-center services.
CME brings AI compute pricing to Wall Street
CME first disclosed its compute futures plans in May through a partnership with Silicon Data. The GPU benchmarking company has backing from trading firm DRW.
By August, CME had set October 5 as the planned launch date for two contracts. Both will rely on Silicon Data’s Nvidia-focused pricing index.
The Financial Times reported that Nvidia B200 capacity rents for about $5.86 per hour. By comparison, older H100 capacity costs around $2.77 per hour.
CME’s contracts will track the forward value of such pricing for up to 36 months. That could help establish a benchmark in a market with wide price swings.
H100 rental costs reached roughly $8 per hour during the 2024 semiconductor shortage. Rates later fell below $2, according to the Financial Times.
Pete Keavey, CME’s global head of energy and environmental products, described compute as the currency of the AI era. CME CEO Terry Duffy also compared compute with oil when discussing the market in May.
DRW founder Don Wilson has argued that compute could become one of the world’s largest commodities.
The Financial Times cited Boston Consulting Group estimates showing the market expanding from $360 billion in 2025. BCG expects that figure to approach $2.3 trillion by 2030.
BlackRock CEO Larry Fink also backed the broader idea at the Milken Institute conference in May. He said compute futures could form a new asset class alongside power and semiconductors.
Architect CEO Brett Harrison estimated annual compute futures notional value could reach $10 trillion by decade-end.
Bitcoin miners could hedge expanding AI revenue
Bitcoin miners already control infrastructure that could benefit from a liquid compute futures market.
CoinShares said listed miners signed more than $70 billion in AI and high-performance computing contracts during the past year. Its Q1 2026 mining report expects AI to generate between 30% and 70% of miner revenue by year-end.
TeraWulf, Core Scientific, Cipher Mining, and Hut 8 have increasingly moved toward data-center operations. IREN and Bitfarms have also expanded into HPC services.
The cost gap helps explain the pivot. Bitcoin mining hardware costs about $700,000 to $1 million per megawatt.
AI systems can cost between $8 million and $15 million per megawatt. Compute futures could help operators hedge revenue tied to those larger investments.
CME is not alone in pursuing the market. Intercontinental Exchange announced plans in May for GPU compute futures using Ornn’s index.
Architect also acquired a CFTC-regulated exchange to develop the American Innovation Exchange.
However, a reliable benchmark still needs to emerge. The Financial Times noted that many futures contracts fail to attract enough profitable trading activity.
Compute also varies across hardware generations. One hour on an H100 carries a different value from one hour on a B200.
The leading indices can also produce different assessments, while Nvidia and major hyperscalers control much of the supply.
Even so, compute futures could still become an important pricing signal. For Bitcoin miners, that signal could help turn AI compute revenue into something they can hedge.
