Kalshi seeks approval to relist GPU compute markets in U.S.
As demand for AI infrastructure grows, Kalshi is seeking federal approval to resume listing derivatives tied to the future cost of renting GPUs. The proposed contracts are positioned as a way to keep price formation for a key AI input within the U.S. rather than in offshore markets.
Highlights
- Kalshi requested CFTC approval to relist GPU compute derivatives, aiming to enable trading on future costs of renting AI-operating GPUs in the U.S.
- Earlier in 2024, compute markets were reportedly removed from Kalshi over national security concerns, despite the Commerce Department denying direct involvement.
- Approval would grant AI labs and data center operators a domestic hedging tool for GPU price volatility, following the CFTC's August public comment request that closed Tuesday.
Regulatory push for compute derivatives
As reported by Axios, Kalshi is asking the Commodity Futures Trading Commission to allow it to relist compute derivatives, which would let users trade on the future cost of renting GPUs used to run AI.The prediction market had reportedly been told earlier this year to remove compute markets over national security concerns, although the Commerce Department denied issuing such an order. Sudhir Jain, Kalshi's chief compliance officer, said Wednesday on LinkedIn that the company had submitted comments to the CFTC in support of the products.
Jain argued that the contracts would help ensure GPU price discovery happens in the U.S., not offshore. He wrote that compute already fits the definition of a commodity and said both buyers and providers have reasons to hedge pricing, including AI labs seeking to lock in costs and data center operators looking to protect revenue tied to GPU investments.
AI infrastructure pricing and market implications
The CFTC declined to comment on Kalshi's request. The agency issued a request for public comment on compute contracts in August, and that comment period ended Tuesday.At the time, CFTC Chair Michael Selig said American markets could play the same role for the "intelligence economy" that they had for commodities central to the industrial economy. If approved, the contracts could give AI companies and infrastructure providers a new tool to manage volatility in one of the sector's most important operating costs.
In our earlier article on the U.S. AI data center buildout, we examined how soaring financing costs, construction delays, and power constraints are putting the $10 trillion infrastructure boom under pressure. We also highlighted growing investor skepticism as projects get postponed or scaled back, raising doubts about how fast computing capacity can realistically expand.