Hyperliquid Policy Center Urges CFTC to Permit AI Compute Derivatives

The Hyperliquid Policy Center has asked the Commodity Futures Trading Commission to allow U.S. futures and perpetual contracts tied to GPU rental prices and the cost of running AI models in an Oct. 20 letter. According to CryptoNews, the independent research and advocacy organization argues that data-center operators, lenders, and artificial intelligence developers need standardized ways to manage ongoing computing-price risk.

Signed by CEO Jake Chervinsky and senior counsel Brad Bourque, the regulatory submission responds to the commission's request for comments on compute derivatives. The letter argues that rented computing capacity qualifies as a commodity under the Commodity Exchange Act, supporting both dated futures for scheduled expenses and perpetual contracts without expiration dates for ongoing costs.

Proposed Benchmarks and Market Safeguards

The policy center proposes contracts tied to GPU rental prices and the cost of running AI models. GPU-hour contracts would track rental prices of specified hardware, while inference contracts would reference access to a stated volume of AI-model tokens. Citing Silicon Data marketplace rates, the submission notes Nvidia H100 rental rates at approximately $2.50 to $3.19 per GPU-hour in the second half of 2024, compared with $1.92 to $2.00 in the second half of 2025.

To address market risks such as concentrated supply and hardware turnover, the group recommends position limits, independent price contributors, and filtering methods. For perpetual contracts, the submission suggests funding-rate caps and calculations spread over longer periods. Customer-protection proposals include leverage limits, open-interest caps, and funding-cost disclosures in dollar terms.

Key facts

• The Hyperliquid Policy Center submitted an Oct. 20 letter to the CFTC regarding compute derivatives.
• The proposal requests U.S. futures and perpetuals tied to GPU rentals and AI inference costs.
• CEO Jake Chervinsky and senior counsel Brad Bourque signed the response.
• Nvidia H100 marketplace rental rates reached $2.50 to $3.19 per GPU-hour in H2 2024.
• Epoch AI research cited in the letter estimates private AI data center and hardware spending reached roughly 0.8% of U.S. GDP in Q1 2026.