Carmen Li's Plan to Build a Compute Futures Market Could Surpass Oil's $2.5T Market

For commodities traders and macro investors, the emergence of a structured compute futures market represents the next frontier in alternative assets. With global GPU spend projected to exceed $1 trillion annually, Carmen Li's plan to index and financialize compute power could unlock hedging tools and speculative opportunities on a scale comparable to crude oil. The partnership with DRW's Don Wilson signals serious market-making firepower behind the effort, potentially transforming how AI infrastructure is priced, traded, and risk-managed.

Carmen Li is orchestrating one of the most ambitious financial infrastructure projects of the AI era: creating a fully fledged futures market for compute power. As CEO of two companies—Silicon Data and Compute Exchange—Li is methodically assembling the building blocks needed to turn GPUs from specialized physical assets into a tradable commodity on par with oil, wheat, or Treasury bonds. The effort, detailed in a live Odd Lots podcast recording in New York, represents a leap forward in financializing artificial intelligence’s most critical resource.

“

With global GPU spend projected to exceed $1 trillion annually, Carmen Li's plan to index and financialize compute power could unlock hedging tools and speculative opportunities on a scale comparable to crude oil.

The timing is not accidental. AI training and inference demand has triggered unprecedented volatility in GPU pricing and availability. Spot prices for high-end chips like Nvidia’s H100 can swing dramatically based on supply bottlenecks, new model launches, or geopolitical export controls. This uncertainty forces AI labs and cloud providers to hoard capacity or accept arbitrary pricing—a classic market failure that commodity futures were invented to solve. Li’s vision is to give market participants tools to hedge compute cost risk, speculate on future GPU values, and ultimately discover a fair global price for a teraflop-hour of processing.

The two-company structure is a masterclass in market design. Silicon Data is building the pricing index—the foundational layer that any derivatives market requires. Constructing a compute index is fiendishly difficult: it must account for heterogeneous hardware (different GPU generations, memory configs, interconnect speeds), location (power costs and latency vary), and even condition (new vs. used chips, similar to the used-car market). Li acknowledges the challenge of ‘standardizing compute,’ a prerequisite for a credible benchmark that can underpin financial contracts. Compute Exchange, meanwhile, is the spot marketplace where buyers and sellers transact today, generating the price discovery data that feeds the index. This creates a virtuous loop: spot liquidity builds index credibility, and a robust index enables the eventual launch of futures and options.

Partnering with Don Wilson, founder of the giant trading firm DRW, brings institutional heft and market microstructure expertise. Last year, Wilson publicly mused that the GPU market might one day surpass the size of oil. The oil futures market—including Brent and WTI—represents a $2.5 trillion-plus notional value, so the ambition is staggering. If even a fraction of global compute spend, projected to exceed $1 trillion annually later this decade, migrates into standardized derivative contracts, the opportunity is enormous. For financial institutions, a compute futures market would open a new asset class uncorrelated with traditional equities or bonds, attracting commodity trading advisors, macro hedge funds, and corporate hedgers.