Wall Street puts a price on AI compute as miners pivot

CME Group has announced that the first compute futures contract will list on October 5, 2026, pending regulatory approval. This would provide the market with a public reference price for the input resource used by any artificial intelligence application worldwide.

For Bitcoin miners that have been spending the last year transforming themselves into providers of compute resources for AI data centers, this means that they can be more than sellers of hashpower; they can also become producers of a marketable good.

The dilemma that the author of a mining company’s business plan should solve is novel. With compute trading like oil or electricity, the companies that possess the racks, chips, and megawatts would be considered owners of the underlying. A futures price that is available to be observed in the market would, for the first time, give an estimated value to the marginal cost of AI compute capacity.

CME targets AI compute as the ‘new oil’ of Wall Street

The CME first brought up the notion in May when it revealed a collaboration deal with Silicon Data, which is a GPU benchmarking company funded by the trading company DRW. August saw the establishment of the date: the launch of two contracts on October 5, contingent on the approval of the regulators. Both contracts are based on Silicon Data’s index of Nvidia silicon. As reported by the Financial Times, the hourly rental rate of capacity on B200 is approximately $5.86 compared to $2.77 on the older model H100, and CME’s contracts will track the forward value of such pricing for the next 36 months.

The representatives of the CME have never been reluctant to talk about it. “Compute has become the currency of the AI age,” said Pete Keavey, CME’s global head of energy and environmental products, announcing the launch of the product. Terry Duffy, CEO at CME, made the same statement in May, calling compute “the new oil of the 21st century.” Don Wilson, the founder of DRW, has predicted that compute will be the largest commodity in the world until the introduction of compute hedging mechanisms.

Why a price on compute matters now

The strategy hinges on scale and volatility. The Financial Times reported that Boston Consulting Group predicted that the AI compute market would expand from around $360 billion in 2025 to nearly $2.3 trillion in 2030. The volatility in prices is extreme within the context of that expansion: while an H100 GPU was rented at around $8 per hour amid the 2024 semiconductor shortage earlier in the year, it dropped to under $2 late last year. Such volatility is precisely what commercial buyers seek to hedge.

BlackRock CEO Larry Fink laid out the case very succinctly at the Milken Institute conference in May. “There will be a whole new asset class of buying compute futures,” Fink said, citing compute along with power and semiconductors as assets which the US is simply short on. Brett Harrison, CEO of the derivatives platform Architect, estimated compute futures would amount to a notional contract of $10 trillion per year by the end of the decade.

Miners already own the underlying

Here comes into the picture the industry of Bitcoin mining. Within the past year, listed miners have inked over $70 billion worth of AI and high-performance computing contracts, says a report from CoinShares’ Q1 2026 mining report. CoinShares anticipates AI to generate between 30% and 70% of the revenues of those miners by year-end, compared with the current 30%. TeraWulf, Core Scientific, Cipher Mining, and Hut 8 have become, in CoinShares’ words, data center operators who also happen to be mining Bitcoin.

Economics make sense of the flurry of activity. The cost of Bitcoin mining hardware is estimated to be around $700,000 to $1 million per megawatt, compared with $8 million to $15 million per megawatt for AI systems. IREN and Bitfarms are switching to HPC suppliers. A liquid compute market can bring to this group what oil drillers and power producers have had all along: a way to hedge their output.

A benchmark still has to form

Nothing is set in stone. Not only CME; in May, Intercontinental Exchange announced its intentions to list its own GPU compute futures based on another index from the company Ornn. And Architect has purchased a futures exchange regulated by the CFTC to develop the American Innovation Exchange. Cryptopolitan reports that China is considering its version of AI compute futures.

The Financial Times draws attention to the real challenge: two-thirds to three-quarters of futures contracts end up not generating sufficient profit. Compute is heterogeneous (one hour on an H100 is different from one hour on a B200), the two key indices sometimes differ in their assessments, and the suppliers of the commodity are Nvidia and a few hyperscalers. Regardless of whether the market will confirm the AI craze or discredit it, the Financial Times argues that compute futures can be important as a signal rather than a market.