CME Plans Futures Contracts for GPU Computing Power

CME and Silicon Data are trying to turn computing power into a futures market. Caroline Hyde frames the move through CME chief Terry Duffy’s formulation: compute is “the new oil.” The consequence, as Katherine Doherty describes it, is that compute would move from something market participants are trying to observe and value into something they can hedge against future price moves.

The first step was not a futures contract. It was an index. Silicon Data, according to Doherty, created the index to give investors and market participants more transparency around “putting a number” on the value of compute technology. That index was the mechanism for tracking the price of compute. The CME plan adds the next layer: futures contracts that would let market participants hedge against price moves and some of the risks around those prices.

Doherty compares the logic to oil and other commodities: once there is a way to track the market, firms and investors can begin treating compute as an exposure with both current and future pricing risk. Bloomberg’s on-screen graphic stated the development directly: “CME to create futures market for computing power.” Doherty describes the aim as bringing compute into the institutional space, where tech firms and investors could help build the market and give it the liquidity needed to function.

A futures contract, as Ed Ludlow puts it, is an agreement to buy or sell something later at a fixed price. In practical terms, that lets a participant lock in a cost or bet on a price move in either direction. The question for compute is what exactly gets referenced: Ludlow points to GPUs, GPU hours, or a similar measure of processing capacity, and asks whether CME and Silicon Data have worked out how that would function in the real world.

Doherty points back to the index as the place where those details begin. The index is where the market gets into GPUs and processing power: what is being measured today, how that processing power is priced, and how a current reference point can be established. The futures contract would then allow market participants to express a view on where that price moves in the future.

The intended function is clear, but the implementation is still developing. Participants would be able to look at where GPU processing power trades today and where they expect it to trade in the coming weeks or months. Doherty also emphasizes that the market “has to develop,” and that it needs the blessing of regulators.

Compute futures are therefore being described as a planned market built on a tracking mechanism, not as a fully mature contract with every operational question already settled. The price of compute is not only being discussed as an operating cost for companies that need processing power; CME and Silicon Data are trying to make it something that can be priced forward and hedged.

Silicon Data’s index was presented as more than a transparency tool for compute buyers. Caroline Hyde asks whether founder Carmen Lee, described as a former trader, began with a narrower goal — helping buyers understand one compute offering against another — or whether the ambition was to make compute a fully fledged asset class with a broader ecosystem around it.

Doherty’s answer is that the ecosystem was the point. In her account, Silicon Data’s work was not merely a procurement aid for companies buying compute. The index was “one pillar” of something larger: a market structure that could support institutional participation.

Before CME’s involvement, Doherty says, there was not “a big player like CME” bringing compute as an asset class to market in this way. CME’s role matters because it already has derivatives traders and an institutional customer base. That means the prospective market would not rely only on natural users of compute — technology firms, AI providers, and hyperscalers hedging their own costs. It would also bring in an institutional base accustomed to looking at oil, metals, and other commodities as tradable asset classes.

The market envisioned here is mixed by design. One side includes companies exposed to compute prices because they need processing power. Another includes investors and traders who already participate in derivatives markets and may approach compute in the same way they approach other commodities. Doherty describes those worlds as combining to build “this ecosystem, this new market.”

CME’s involvement changes the scale of the compute-derivatives idea. Ed Ludlow asks whether the next step could be similar proposals from other players, including in other jurisdictions. Doherty says that some exchanges and upstarts have already appeared in the space, but not at CME’s scale. She identifies CME as the largest U.S. derivatives exchange and treats its entry as a marker that compute derivatives are moving beyond smaller experiments.

Her expectation is that other large exchange operators are likely to follow with their own marketplaces. The rationale is not exclusivity but liquidity: more trading, more interest, and more participants can strengthen the market overall. Doherty explicitly says it does not need to be only one exchange.

The prospective compute futures market is not presented as a CME monopoly over the financialization of processing power. It is an early institutional move in a market that may require multiple venues, regulatory blessing, and enough participation from both commercial hedgers and financial traders to make the contracts useful.