A CME Group sign above the trading pit of the former Chicago Board of Trade (CBOT) on Thursday, November 13, 2025 in Chicago, Illinois, USA.
Christopher Dilts | Bloomberg | Getty Images
A new futures market for semiconductors will allow traders to hedge their investments in artificial intelligence by betting on the increasingly expensive price of computing power.
CME Group’s new “computing futures market” contracts will be based on Silicon Data’s graphics processing unit (GPU) price index, the companies said in a statement announcing the joint venture on Tuesday, which is still pending regulatory review.
The new market allows investors to lock in the price of computing power based on GPU benchmarks, which can be used to protect against rising GPU rental fees and other operating costs in large-scale, multifaceted AI builds.
“The GPU market…historically has not had a standardized reference price,” Silicon Data CEO Carmen Li said in a release. “The launch of Computing Futures is an important step toward providing AI builders, cloud providers, and investors with more reliable tools for evaluation, hedging, and long-term planning.”
While futures markets are traditionally associated with basic commodities such as foodstuffs, metals, and petroleum products, markets are also emerging for assembled components in rapidly developing sectors of the industrial sector.
During the broadband explosion of the late 1990s, Enron’s broadband services division sought to sell unused capacity on its fiber-optic cable network before the company became a fiasco.
Silicon Data sells clients access to specialized price indexes, similar to the Consumer Price Index and the Personal Consumption Expenditure Price Index, excluding semiconductors. Its products include a standardized GPU price index, RAM index, and GPU rental price predictions.
Wall Street doesn’t see demand for GPUs, or more traditional central processing units (CPUs), slowing down anytime soon.
“Agent AI requires a whole new rack of CPU servers that sit alongside the GPU infrastructure and run to power the work of all these agents,” Morgan Stanley analyst Sean Kim said in a report Monday.
“Future AI systems will look like distributed systems consisting of GPU racks for dense model computing and agent CPU racks for orchestration, data processing, and tool execution,” Kim said.
Memory chip prices soared in the first quarter as demand for CPUs increased due to AI. Hyperscalers increased capital spending across the board, but executives expressed concerns about memory bottlenecks that were driving up input costs.
With valuations soaring, memory chip makers are anticipating huge profit margins this year and into next.
