Christopher Dilts | Bloomberg | Getty Images
Contracts on the new “compute futures market” from CME Group will be based on graphics processing units (GPU) price indexes from Silicon Data, the companies said in a statement released Tuesday announcing the joint venture, which still is pending regulatory review.
The new market will let investors lock in a price for computing capacity based on a GPU benchmark, which can be used to hedge against rising GPU rental rates and other operational costs in the enormous and multifaceted AI buildout.
“GPU markets … have historically lacked standardized reference pricing,” Carmen Li, chief executive of Silicon Data, said in the release. “The launch of compute futures is an important step toward giving AI builders, cloud providers and investors more reliable tools for valuation, hedging and long-term planning.”
Futures markets are traditionally associated with basic commodities like foodstuffs, metals, and petroleum products, but they’ve also popped up for assembled components in rapidly developing segments of advanced industrial sectors.
During the broadband explosion in the late 1990s, the broadband services division of Enron aimed to sell unused capacity on its network of fiber optic cables prior to the company’s spectacular failure.
Silicon Data sells access to specialized price indexes to clients, similar to the consumer price index or personal consumption expenditures price index, except for semiconductors. Its products include a standardized GPU price index, a RAM index and projections for GPU rental prices.
Wall Street doesn’t see demand for GPUs, or more traditional central processing units (CPUs), slowing down any time soon.
“Agentic AI requires entirely new racks of CPU servers that sit alongside GPU infrastructure and run to power the work of all these agents,” analyst Shawn Kim at Morgan Stanley wrote in a report Monday.
“The AI system in the future will look like a distributed system consisting of GPU racks for dense model compute … [and] agentic CPU racks for orchestration, processing data and tool execution,” Kim said.
Memory chip prices soared in the first quarter as AI drove increased demand for CPUs. Hyperscalers increased capital spending across the board while executives expressed concerns about a bottleneck in memory that’s driving input costs higher.
Memory chip makers are projecting huge profit margins through this year and next as valuations have skyrocketed.
About the Author
