The Commodity Futures Trading Commission sent a draft request for public comment on AI compute derivatives to the White House budget office for review in mid-August, moving to set rules for a futures market that does not exist yet. Once the Office of Management and Budget clears it, the request opens a 30-to-60-day public comment window covering compute cash markets, manipulation risk, customer protection and what the agency calls perpetual compute futures. “America cannot win the AI race without a robust derivatives market for compute,” CFTC Chairman Michael Selig said, calling the request “the first step toward establishing clear rules.”
The clock is running because exchanges are not waiting. CME Group plans to launch two contracts on October 5: Silicon Data H100 Rental Index Futures and Silicon Data B200 Rental Index Futures, each representing a month of computing power priced against Nvidia’s two most widely deployed data center chips. CME built the index with Silicon Data, a GPU market-intelligence firm. “Compute has become the currency of the AI age, and this innovative market will bring transparency to current and future costs,” said Pete Keavey, who leads the effort at CME.
Intercontinental Exchange is building its own product on a different index, the Ornn Compute Price Index, which tracks live-traded spot prices across several hardware types rather than a single benchmark chip. A third entrant, Architect Financial Technologies, acquired the health-data company IMX Health and is seeking CFTC approval to launch an exchange built around compute futures and options across multiple GPU vendors, plus a mechanism it calls ComputeConnect that lets a futures position convert directly into GPU capacity. “We can build products addressing actual compute buyers and sellers, not just financial speculators,” said Brett Harrison, Architect’s chief executive, who has projected the market could reach $10 trillion in annual notional value by 2030.
What the CFTC actually has to referee is whether the underlying price is real. Unlike oil or wheat, benchmarked for decades against deep, liquid physical markets, GPU rental rates come from a small number of cloud providers and brokers with no long pricing history and no standard way of quoting them. A futures contract is only as trustworthy as the index it tracks, and Silicon Data, Ornn and Architect are all, in effect, asking the market to trust a benchmark that did not exist a year ago.
The timing tracks the buildout it is trying to price. Compute capital spending is on pace for $600 billion to $800 billion this year, with projections reaching $2 trillion by 2030, and lenders are already treating chips as collateral: Nvidia signed six-institution memorandums this month aiming to mobilize more than $500 billion in debt secured against the hardware itself. Futures contracts would add a further layer on top of a compute market that is still being priced ad hoc: OpenAI, for one, started charging a premium for faster inference on the same hardware this month, the kind of price that has never traded on an exchange.
Neither CME nor the CFTC has said what happens if the comment period is still open on October 5. The exchanges are building their contracts on the current timeline regardless.
Sources: Crypto Briefing, The Star, Markets Media
–
By the Control Plane Editorial Team