Hyperliquid Policy Center has asked the CFTC 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.
Summary
- Hyperliquid Policy Center wants U.S. futures and perpetuals tied to GPU rentals and AI inference costs.
- The group says perpetuals could help data centers and AI developers hedge ongoing price risks.
- Proposed safeguards include leverage limits, funding-rate caps and transparent pricing benchmarks.
- The letter supports dated futures for scheduled expenses, such as planned AI training runs.
Hyperliquid Policy Center said data-center operators, lenders and AI developers need ways to manage computing-price risk, urging the agency to review contracts under its existing rules. CEO Jake Chervinsky and senior counsel Brad Bourque signed the response to the regulator’s request for comments on compute derivatives.
— Hyperliquid Policy Center (@HyperliquidPC) October 20, 2026Today, HPC filed a comment letter asking the @CFTC to bring compute perpetual contracts to U.S. markets.
Data centers, their lenders, and the AI developers that rent compute all carry price risk, and the exposure runs around the clock with no fixed end date. Our comment explains… https://t.co/SpHZvLo5KR
The policy center describes itself as an independent research and advocacy organization seeking regulated American access to onchain markets, including Hyperliquid. Its letter argues that rented computing capacity qualifies as a commodity under the Commodity Exchange Act, while supporting contracts linked to the cost of running AI models.
The submission asks regulators to allow competing price benchmarks, support both dated futures and perpetuals, and address market risks through trading safeguards and oversight. It is a policy recommendation, not an announcement of a Hyperliquid compute-contract launch or a regulatory approval.
AI compute perpetuals would target ongoing costs
In the letter’s account, computing capacity creates price exposure that continues around the clock. Data centers earn rental revenue by the hour, while inference platforms pay for computing power each time customers use an AI model.
“A GPU-hour exists only in the hour it is used,” the policy center wrote.
The organization argues that perpetual contracts could match those ongoing costs and revenues because they have no expiration date. Its explanation describes recurring funding payments between buyers and sellers that help keep a perpetual’s price aligned with its reference benchmark.
According to the submission, a data-center operator could use such a contract to protect against falling rental revenue, while an AI platform could hedge rising computing bills. Lenders financing hardware face related exposure because lower rental income can weaken borrowers’ cash flows and the value of their collateral.
The letter favors dated futures for costs with a known endpoint, including a training run planned for a specific quarter or a scheduled capacity purchase. It says successive monthly contracts can produce a forward price curve that lenders and investors need when valuing financing agreements.
GPU rentals and inference need different benchmarks
The policy center proposes contracts tied to GPU rental prices and the cost of running AI models.
GPU-hour contracts would track the rental price of specified hardware. Inference contracts would reference access to a stated volume of AI-model tokens, the units used to measure text that a model processes and produces.
According to the letter, the second category would serve application developers and inference platforms whose bills depend on token prices. A hardware-rental contract would provide an imperfect hedge because inference costs depend on both computing prices and model efficiency.
Citing Silicon Data’s historical price series, the submission puts Nvidia H100 marketplace rental rates at approximately $2.50–$3.19 per GPU-hour in the second half of 2024, compared with $1.92–$2.00 in the second half of 2025.
The letter cites Epoch AI research estimating that private investment in AI data centers, computing hardware and networking equipment reached roughly 0.8% of U.S. GDP in the first quarter of 2026.
Despite that spending, the policy center describes a market where most economic value changes hands through privately negotiated agreements. Its account distinguishes publicly visible hourly rentals from reserved capacity and multiyear purchase commitments whose prices generally remain private.
Benchmark selection would therefore determine which exposures a contract can cover, the organization argues. Its proposed specifications include chip type, location, rental period, cluster size, service terms and whether a provider can interrupt access.
Pricing safeguards would address thin markets
The letter identifies concentrated supply, limited transaction visibility and hardware turnover as unresolved issues for compute derivatives.
According to the policy center, a benchmark built from relatively few transactions could be vulnerable to manipulation. It proposes independent price contributors, methods that filter out unusual observations, position limits, and information-sharing arrangements.
For perpetual contracts, the submission recommends funding-rate caps or calculations spread over longer periods to limit sudden funding costs around model releases and hardware shortages.
New chip generations create a separate benchmark problem. The organization proposes advance notice and gradual transitions when an index replaces older hardware, citing the disclosed schedules used by Hyperliquid oil perpetual markets when their reference futures move between contract months.
Its customer-protection proposals include leverage limits, open-interest caps and disclosures showing funding costs in dollar terms over different holding periods. The letter calls for venues to explain liquidation thresholds and any discretion retained by exchange operators or index administrators.
CFTC review follows earlier perpetual-market proposals
On Aug. 19, the CFTC announced its compute-derivatives consultation, seeking feedback on cash-market liquidity, manipulation, customer protection and perpetual compute futures.
Chairman Michael Selig described derivatives markets as part of America’s AI ambitions:
“America cannot win the AI race without a robust derivatives market for compute.”
The policy center’s letter says NYMEX and ICE Futures U.S. have proposed cash-settled compute futures, although neither exchange’s contracts had begun trading when the submission was written. It describes existing compute-price event contracts and perpetual offerings through a Bermuda-licensed venue.
Earlier, crypto.news covered the group’s Aug. 26 request for regulated energy perpetuals alongside trade[XYZ]. Their filing sought a U.S. path for contracts tied to West Texas Intermediate crude, Brent crude and Henry Hub natural gas.
In an Aug. 24 submission, the group asked the SEC and CFTC to regulate stock-linked perpetual contracts as security futures under their joint oversight. The group argued that regulators should examine a contract’s structure before assigning jurisdiction based on its underlying asset.
On Oct. 19, the organization submitted five SEC rule recommendations concerning Regulation Crypto Assets, covering airdrops, secondary trading, developer protections, automated token buybacks, and state securities requirements.