Comment to the CFTC on Compute Futures and GPU Price Indexes
By R Tamara de Silva | De Silva Law Offices, LLCÂ
On October 19, 2026, De Silva Law Offices filed a comment letter with the CFTC on compute futures. These contracts could let companies lock in the cost of renting the GPU capacity used to build and run AI models. Our letter supports listing them. Its main concern is the indexes the contracts would settle on and the firms that supply their prices, many of which may also trade the futures.
The letter appears below as filed, with its footnotes as numbered notes at the end. The official copy is posted on Regulations.gov. Our article GPU Compute Futures and the CFTC covers the NYMEX and ICE proposals.
The Comment Letter
October 19, 2026
Submitted electronically via Regulations.gov
Christopher Kirkpatrick
Secretary of the Commission
Commodity Futures Trading Commission
Three Lafayette Centre
1155 21st Street NW
Washington, DC 20581
Re: Request for Comment on the Listing of Compute Derivatives Contracts; RIN 3038-AF77; 91 Fed. Reg. 54259 (Aug. 21, 2026)
Dear Mr. Kirkpatrick:
Many companies that build artificial intelligence rent the computing power they use. They pay by the hour for time on specialized chips in someone else’s data center. The Commission’s Request for Comment on the Listing of Compute Derivatives Contracts (the “Request”) calls the price of that compute “one of the most material costs of AI.”[1] No U.S. futures market for compute exists yet. NYMEX and ICE have proposed futures that would settle on indexes of GPU rental prices, and the Request asks whether compute price indexes can support a futures market.
De Silva Law Offices, LLC supports listing compute futures. Compute cannot be stored in quantity, but neither can electricity or ocean freight, and futures markets price both. In the firm’s view, the main risk lies in who supplies the prices. Compute indexes draw largely on rates and transactions supplied by the firms that sell compute or run the venues where it trades. Many of those firms may also trade the futures. Natural gas had a similar structure in 2002, and false reports to its indexes led to a series of enforcement cases. The Commission can set expectations for compute indexes, and for the firms that feed them, before compute futures trade at scale on U.S. exchanges.
I. Statement of Interest
De Silva Law Offices is a Chicago financial regulatory and litigation firm founded in 2002. Its practice centers on CFTC and NFA compliance, derivatives and futures regulation, event contracts and prediction markets, fund formation, and enforcement defense. The firm’s clients include designated contract markets, futures commission merchants, introducing brokers, commodity pool operators, commodity trading advisors, proprietary trading firms, and event contract and fintech platforms. Its Managing Partner spent more than two decades in the industry, beginning on the floor of the Chicago Board of Trade as a floor trader and floor broker.
This year the firm submitted comments to the Commission on prediction markets, fintech innovation and event contract data reporting. In 2025 it commented on perpetual futures.[2] The comments below respond principally to Questions 1(a), 1(c), 1(f), 2(a), 2(b), 2(d), 2(g), 3(a) and 4.
II. Summary of Comments
1. The Commission should set out its expectations for compute derivatives in supplemental guidance to Appendix C to Part 38, proposed for comment. Exchanges could then self-certify later contracts against known standards.
2. A compute contract should settle only on an index whose administrator can give the listing exchange, and through it the Commission, the data behind each settlement value. The administrator should also publish its methodology and contributor concentration, disclose its affiliations, and have its methodology and data controls reviewed independently.
3. Knowingly or recklessly false submissions to a compute index would fall within Section 6(c)(1)(A) of the Act and, when knowing, Section 9(a)(2). The Commission should say so, make clear that a posted rate the provider will not honor can be a false report, and encourage contributor codes of conduct modeled on FERC’s 2003 standards.
4. Barring index contributors from compute futures would exclude many natural hedgers. Exchanges should instead identify their accounts and watch their trading, and their rate changes, during the settlement window.
5. Final settlement should average daily index values over the contract month. Contract terms should state what the index measures and disclose basis risk. Spot-month limits should rest on observable index-eligible volume, with accountability levels outside the spot month. Compute futures should enter the Commitments of Traders report once they qualify.
6. Perpetual compute futures should be listed only if their reference index meets the standards in item 2. The Commission should also codify by rule the position it took on perpetual contracts in May 2026.
7. The Commission should confirm when a capacity reservation is a forward contract, what an exchange for physical in compute requires, and when a compute marketplace must register. AML expectations should focus on physical delivery, including exchanges for physical.
III. Background: What GPU Compute Is and How It Is Sold
A graphics processing unit, or GPU, is a chip first designed to render graphics. Because it performs thousands of calculations at once, it has become the leading hardware for training AI models and for running them. Nvidia’s H100 and its newer B200 are the chips named in the proposed futures. The Request describes compute as “the processing power primarily used by the large language models” at the center of the AI economy. It notes that compute “has become a multi-hundred-billion-dollar enterprise.”[3]
A buyer of GPU capacity can pay by the hour at a provider’s posted rate, with no commitment. It can also reserve or commit to capacity for months or years at a negotiated price. Sellers include the large cloud companies, often called hyperscalers; newer specialized providers, often called neoclouds; and marketplaces that resell spare capacity. In the Commission’s preliminary view, compute markets are fragmented, and prices are set mainly in “opaque bilateral transactions.”[4]
Prices differ widely by seller. In April and May 2026, Silicon Data’s index of hyperscaler H100 rental prices stayed between about $7.40 and $7.55 per GPU-hour. Its neocloud index ranged from about $2.47 to $2.76 over the same period and was roughly five times as volatile.[5]
Two firms publish the indexes behind the proposed futures. Silicon Data, which is backed by the trading firm DRW, gathers rental and transaction prices from many sources and adjusts them for differences in rental type, location and hardware. Ornn says its index is built only from executed transactions.[6] On August 11, NYMEX asked the Commission to approve monthly, cash-settled futures on Silicon Data’s H100 and B200 indexes. On September 21, the Division of Market Oversight extended its review to November 9, citing novel or complex issues and this Request.[7] ICE has said it will not list its contracts on Ornn’s indexes until the comment period closes.[8]
Exposure to compute prices already trades in other forms. At least two designated contract markets have listed or filed terms for event contracts on compute prices, and an offshore exchange lists perpetual-style compute futures.[9]
IV. Lessons From Comparable Markets (Questions 1(a) and 1(e))
Question 1(a) asks whether particular commodity markets are “especially relevant or similar to the compute cash market.” The Request also notes that compute’s issues are “informed by historical antecedents in the development of derivative markets in similar commodities.”[10] In the firm’s view, three markets are especially relevant: electricity, ocean freight and natural gas.
The Request’s footnote on historical antecedents cites electricity, the usual comparison. Like compute, electricity cannot be stored in quantity. But many electricity futures settle on prices published by regulated grid operators. NYMEX’s PJM Western Hub futures, for example, settle on the average of real-time prices at a hub on the PJM grid.[11] Compute has no public price of that kind.
Ocean freight is a closer analogue. A day of a ship’s time, like an hour of GPU time, is lost if it goes unsold. Ships, like GPU clusters, differ in size, age and efficiency, and freight rates are set in privately negotiated charters. Freight has nonetheless supported cash-settled derivatives for many years. Freight futures clear through CME and other clearinghouses and settle on the average of a Baltic Exchange index over the month.[12] NYMEX itself lists freight futures that settle on the average of a month’s daily Baltic Exchange rates, the same design the compute proposals use.[13]
The Baltic Exchange builds those rates from assessments by panels of shipbrokers. Shipowners and charterers, the parties with the most direct stake in the price, cannot serve on the panels. The Baltic Exchange’s benchmark administrator is authorized by the U.K. Financial Conduct Authority.[14] Compute indexes are built differently. Some draw on rates that capacity providers post themselves, and others on transactions from venues that a few participants operate or dominate.[15] For compute, the protections that freight builds into its panels will therefore have to come from index governance and surveillance.
Natural gas offers a warning. In 2002, the leading natural gas price indexes were built from trade data that energy companies reported to publishers such as Platts. Those indexes priced physical gas deals, and futures traders used them for price discovery. Traders at a number of companies reported false or inaccurate trades. Beginning in late 2002, the Commission charged false reporting and attempted manipulation, and by November 2003 it had imposed more than $130 million in civil penalties in energy trading cases. The Commission later used the same false-reporting theory in the LIBOR cases.[16] FERC responded in 2003 with standards for the companies that report price data and the publishers that compile it.[17] In one respect, compute today resembles natural gas before those reforms. Its indexes depend on data from interested firms, and no common standard yet governs how that data is reported.
V. The Commission Should Set Out Its Expectations in Guidance (Question 2(i))
As Section III notes, NYMEX sought approval of its compute futures under Regulation 40.3 rather than self-certifying them, and the review has been extended. That caution is understandable for a new underlier. Contract-by-contract review should not, however, become the standing way to decide what a compliant compute contract looks like.
In earlier letters, the firm urged the Commission to answer recurring structural questions once and in public, in a form that applies to everyone.[18] The same approach fits here. Because Appendix C is itself guidance, the Commission can supplement it for compute and propose the supplement for comment after it reviews the responses to this Request. Exchanges could then self-certify later contracts under Regulation 40.2 against known standards. Commission staff could spend more of its time on surveillance of live markets and less on serial product reviews.
VI. Standards for Compute Settlement Indexes (Questions 1(c), 2(a), 2(b), 2(d), and 2(g))
The Request asks whether a contract should trade if it settles to a price computed from data the Commission “may not be able to observe, verify, or surveil, in whole or in part.”[19] The firm believes it should not, unless the exchange can obtain the underlying data on request. Appendix C already contemplates that a third party may calculate a settlement price. It asks the exchange to consider a licensing agreement, to verify the provider’s safeguards against manipulation, and, where practicable, to enter into an information-sharing agreement.[20] For compute, the firm recommends that those considerations become four expectations of the index administrator.
1. Data access. For each settlement value, the administrator should be able to give the exchange, on request, the transactions and posted rates behind it, identified by source and time. Where contributors submit data, the administrator should obtain by contract the right to pass that data to the exchange. Where rates come from public listings, it should keep records of what it collected. The exchange can then make the data available to the Commission under Core Principle 4. This approach avoids requiring an exchange to sign agreements with every provider and venue in a fragmented market (Question 2(d)).
2. Methodology and concentration. The administrator should publish its methodology and its procedures for changing it. For each settlement period, it should publish the number of contributors and the share of inputs from the largest contributor and from the five largest. It should also show how much of the index comes from posted rates rather than completed transactions.[21] Ranges are enough where exact figures would reveal confidential data.
3. Affiliations. The administrator should disclose any ownership by, or affiliation with, a firm that trades compute derivatives or a venue whose transactions enter the index. It should also explain how it keeps index calculation separate from those affiliates. One provider whose index underlies a pending futures submission is backed by a trading firm, and another publishes an index while operating its own venue for compute risk transfer.[22] The firm does not suggest that either has acted improperly. It raised a related concern about combined exchange, clearing and brokerage functions in its July 8 comment.
4. Independent review. The administrator should have its methodology and data controls reviewed independently at regular intervals. The IOSCO Principles for Financial Benchmarks offer one reference.[23] Another is FERC’s 2003 policy statement, which asked the publishers of energy price indexes to verify reported data and undergo independent audits.[24]
Appendix C prefers publicly available price series. When a settlement price rests on non-public sources, it asks the exchange to publish the final settlement price and supporting information as soon as possible.[25] The conditions above would add the Commission’s access to the underlying data (Question 2(g)). The firm takes no position on whether any current compute index meets those conditions.
VII. Index Contributors Should Be Put on Notice and Surveilled Rather Than Excluded (Questions 1(f), 2(b), 2(f), and 3(e))
Question 1(f) asks, among other things, whether the parties best positioned to influence a compute reference price are the same parties that supply its data. The Act already reaches much of the conduct that concerns the Commission. Section 6(c)(1)(A), which Congress added in 2010, treats as manipulation a false, misleading or inaccurate report of market information that affects or tends to affect the price of a commodity. It reaches reckless as well as knowing conduct. Section 6(c)(1)(C) protects good-faith mistakes in reporting to a price reporting service.[26] Section 9(a)(2), the provision behind the natural gas cases, still makes knowingly false market reports a felony.[27] Once compute futures list, these provisions would reach false transaction reports that feed their settlement indexes. Because many of the firms supplying that data have never dealt with the Commission, the guidance described in Section V should tell them clearly that these provisions reach such reports.
Posted rates present a harder question. A gas trader who reported a fictitious trade made a false report. A provider that posts a rate it will not honor, in order to move an index, comes closer to spoofing. Like a spoofer, it shows a price to move the market rather than to trade at it. GPU Economy publishes an index built on posted rates. It estimates that a single provider changing its rates in every region it lists could move the index’s weekly value by as much as 10.5 percent. Its smaller monthly panel could move by as much as 12.8 percent.[28] The anti-spoofing provision would not reach a false posted rate, because it applies only to conduct on or subject to the rules of a registered entity.[29] Rule 180.1 may reach it, because it covers reckless false statements and false reports. A price-manipulation charge, by contrast, would require proof of specific intent to affect the price.[30] The firm is not aware, however, of any case applying these provisions to a posted rate.
The Commission can reduce that uncertainty in two ways. First, its guidance can state that a posted rate is a report concerning market information when the provider knows the rate is used in a settlement index. Providers would then be on notice that a knowingly or recklessly false rate can violate Section 6(c)(1)(A). Second, exchanges can expect index administrators to obtain attestations, from contributors that supply rates under contract, that those rates could be booked when read.[31] Attestations would improve the index and would also make each attested rate a representation that the false-reporting provisions more clearly reach.
The firm also recommends that the Commission encourage contributor codes of conduct modeled on FERC’s 2003 standards. FERC asked companies that report trade data to adopt a code of conduct, separate reporting from trading, keep records and correct errors. It gave companies that met its standards a rebuttable presumption of good faith.[32] The Commission could make clear that following a published code of that kind is strong evidence of the good faith Section 6(c)(1)(C) protects.
The Commission should not, however, exclude contributors from the market. One exchange’s terms for compute event contracts bar from trading the ultimate beneficial owners of any entity able to influence the settlement data.[33] That approach may suit an event contract. In a futures market, the firms that post rates and report transactions are the neoclouds, data center operators and marketplaces whose price risk the contract exists to transfer. A ban would exclude many natural hedgers. Exchanges should instead identify the accounts of index contributors and their affiliates, require position reporting at levels the exchange sets, and watch their trading closely during the settlement averaging window. The time-stamped records described in Section VI would let an exchange match each contributor’s rate changes against its positions.
VIII. Contract Design, Position Limits and Public Data (Questions 1(g), 2(h), and 3(c))
The NYMEX and ICE contracts settle on the average of daily index values over the contract month.[34] The firm supports that design. A monthly average is harder to move than a single day’s value, and freight and many power contracts settle the same way. The Commission’s guidance should treat averaging as the expected design. An exchange that proposes single-day settlement should have to show why the underlying data can support it.
Contract terms should also state plainly what the index measures: the price type (on-demand, spot or reserved), the hardware and the geography. Both pending NYMEX contracts settle on on-demand prices for U.S. capacity.[35] The Request notes, however, that non-public bilateral agreements appear to “carry the majority of economic value,” and Section III shows how far prices differ by seller.[36] In response to Question 3(c), contract terms and customer disclosures should describe this basis risk in specific terms. A commercial user should be able to judge whether the index tracks its own costs.
On Questions 1(g) and 2(h), compute has no deliverable supply in the usual sense. A cash-settled contract involves no delivery, and an unused GPU-hour cannot be stored. For cash-settled contracts, Appendix C asks the exchange to analyze the size and liquidity of the underlying cash market in the way it would estimate deliverable supply.[37] For compute, the closest measure is the volume of index-eligible transactions and rates observed during the averaging period. Regulation 150.5(b) requires a spot-month limit on a contract in a physical commodity, a term that includes exempt commodities. It allows an alternative to the 25-percent-of-deliverable-supply standard if the limit is necessary and appropriate to reduce the potential threat of manipulation or price distortion.[38] Exchanges should use that alternative to set conservative spot-month limits based on index-eligible volume. Outside the spot month, accountability levels backed by large-trader reporting fit better than hard limits until the market has more history.
Finally, the firm encourages the Commission to include compute futures in the Commitments of Traders report once they meet its criteria for publication. In its July 31 comment, the firm asked the Commission to consider publishing event contract position data on the same model.[39]
IX. Perpetual Compute Futures Should Meet the Same Index Standards (Questions 4(a) and 4(b))
The firm supported the listing of perpetual futures in its 2025 comment and still does. Perpetual contracts avoid the cost of rolling expiring positions and can bring offshore trading into regulated markets. Their risks can be managed with margin, position limits and surveillance.[40] As noted in Section III, perpetual-style compute futures already trade offshore.
A compute perpetual would, however, rely on its index more heavily than a dated contract does. A monthly contract depends on the index during one month, and averaging dilutes any single day. A perpetual’s funding payments are typically calculated at regular intervals from the gap between the contract price and the index. The index therefore matters every day a position is open. Perpetual compute futures should be listed only if their reference index meets the standards in Section VI. Their contract terms should state the funding-rate method and any caps.
The firm also renews its 2025 recommendation that the Commission address perpetual contracts by rule.[41] The Commission’s May 29, 2026 order approving a perpetual contract for listing as a futures contract, and the policy statement that accompanied it, are now being challenged in court.[42] The Commission should codify its position by rule, after notice and comment. A commercial user should not have to wonder whether a hedge it entered as a futures contract will later be treated as a swap.[43]
X. Customer Protection and Anti-Money Laundering (Questions 3(a), 3(b), and 3(d))
Question 3(a) asks whether compute markets raise heightened anti-money laundering and know-your-customer concerns. The firm’s Managing Partner serves as AML compliance officer for CFTC-registered firms. In the firm’s view, the answer depends on where the futures meet the physical market.
The firm has not identified an AML risk, or a challenge for intermediaries, specific to cash-settled compute futures that existing rules fail to address. Futures commission merchants and introducing brokers must already maintain AML programs, verify customer identity, identify the beneficial owners of legal entity customers, and file suspicious activity reports.[44]
The concern arises at delivery, which can itself raise sanctions and export-control questions. In Question 3(b), the Request describes compute as geopolitically sensitive.[45] Physical delivery, or an exchange for physical that pairs a futures position with a transfer of capacity, could give access to advanced computing to a party that no provider has screened. An exchange that permits either should require delivery only through providers that screen the receiving party, at the point of transfer, for sanctions and export-control purposes. The parties should keep screening records and produce them on request.
On Questions 3(b) and 3(d), disclosure should be specific to compute. A retail customer should be told what the index measures and how it is built. The customer should also learn that the index may not track any particular provider’s prices, and that government action on chips or data centers can move the market abruptly. For perpetual contracts, the disclosure should explain funding payments in plain terms.
XI. Forward Contracts, Exchanges for Physical, and Compute Marketplaces (Request Footnote 6)
The Request also invites comment on physically settled compute swaps and on compute contracts listed on swap execution facilities.[46] The firm urges the Commission to address the physical market directly as well, because many of the companies affected by this Request do business only in that market.
First, the Commission should confirm when a reservation of compute capacity is a forward contract. The 2012 product definitions release treats an intangible commodity as a nonfinancial commodity, eligible for the forward contract exclusion, if ownership of it can be conveyed and it can be consumed. The commodity must also be capable of physical delivery and must not be an excluded commodity. The same release discusses capacity and tolling arrangements in which the buyer pays for the use of a facility.[47] A reservation that gives a buyer the use of specified GPU capacity for a stated term, with intent to deliver, appears to fit the release’s framework. Guidance should confirm that it does and identify the features that might change the answer. Those could include a right to settle the difference in cash, a severable price option, or capacity that is not dedicated to the buyer.
Second, the Commission should state what an exchange for physical in compute requires. One firm preparing to launch a designated contract market for compute futures has announced an exchange-for-physical network that would let its customers convert futures positions into GPU capacity.[48] The physical leg of such a transaction should be a bona fide, documented transfer of the right to use specified capacity, not a paper transfer made only to move a futures position.
Third, the Commission should describe when a compute marketplace must register, so that marketplaces can design their products accordingly. A platform where customers buy capacity they intend to use is a commercial marketplace. A platform that lists standardized forward contracts, lets multiple participants trade reserved capacity among themselves, or offers offsetting positions or cash settlement may be offering futures or swaps. If so, it must be designated as a contract market or, if it offers only swaps, registered as a swap execution facility.[49]
XII. Conclusion
Compute futures could give the companies that build and use AI what energy and freight futures give their users: a public price and a way to lock in costs ahead of time. Freight demonstrates that a futures market can rest on an index of a perishable service priced in private. The natural gas cases show what can happen when the firms that report prices also trade on them and the reports are not verified. The firm respectfully urges the Commission to set its expectations for compute now, while compute futures have yet to trade at scale on U.S. exchanges.
The firm appreciates the opportunity to comment and would be pleased to provide further information or to answer any questions the Commission or its staff may have.
Respectfully submitted,
R Tamara de Silva
Managing Partner
De Silva Law Offices, LLC
cc: The Honorable Michael S. Selig, Chairman
Notes
[1] Request for Comment on the Listing of Compute Derivatives Contracts, 91 Fed. Reg. 54259, 54262 (Aug. 21, 2026) (“Request”), https://www.federalregister.gov/documents/2026/08/21/2026-17163/request-for-comment-on-the-listing-of-compute-derivatives-contracts.
[2] Comment of De Silva Law Offices, LLC on the Prediction Markets Advance Notice of Proposed Rulemaking, RIN 3038-AF65 (Apr. 30, 2026) (Comment No. 115376), https://comments.cftc.gov/PublicComments/ViewComment.aspx?id=115376; Comment of De Silva Law Offices, LLC on the Request for Information: Identifying Regulations To Facilitate Innovation and Competition to Financial Products and Services for Fintech Firms, RIN 3038-ZA24 (July 8, 2026), https://www.regulations.gov/comment/CFTC-2026-1321-0006; Comment of De Silva Law Offices, LLC on Data Reporting Requirements for Certain Event Contracts, RIN 3038-AF73 (July 31, 2026), https://www.regulations.gov/comment/CFTC-2026-1453-0016; Letter from R. Tamara de Silva, De Silva Law Offices, LLC, to Christopher J. Kirkpatrick, Secretary, Commodity Futures Trading Commission, Re: Comments on Potential Listing and Regulation of Perpetual Futures (“Perpetual Swaps”) (May 20, 2025) (Comment No. 74864), https://comments.cftc.gov/PublicComments/ViewComment.aspx?id=74864.
[3] Request, 91 Fed. Reg. at 54260.
[4] Id. at 54262 (“opaque bilateral transactions”); id. at 54263 (Question 1(a)) (asking commenters to “distinguish between on-demand, spot, reserved, committed purchase modes”).
[5] Carmen Li, H100 Index, May 2026: Same Chip, Two Tiers, Two Regimes, Silicon Data (June 16, 2026) (reporting annualized volatility of 3.8 percent for the hyperscaler index and 18.4 percent for the neocloud index), https://www.silicondata.com/blog/h100-index-may-2026-two-tiers-two-regimes.
[6] Silicon Data, GPU Rental Price Indices (Methodology) (each record is standardized “to a benchmark-equivalent contract, accounting for rental type, geography, CPU platform, and GPU variant”), https://www.silicondata.com/products/silicon-index?gpu=h100; CME Group, Press Release, CME Group and Silicon Data to Launch Compute Futures on October 5 to Unlock New Way to Hedge AI Risks (Aug. 11, 2026) (describing Silicon Data as backed by the trading firm DRW), https://www.cmegroup.com/media-room/press-releases/2026/8/11/cme_group_and_silicondatatolaunchcomputefuturesonoctober5tounloc.html; Intercontinental Exchange, Press Release, ICE and Ornn to Launch GPU Compute Futures Contracts (May 19, 2026) (OCPI “is the first compute index to be built only from printed transactions”), https://ir.theice.com/press/news-details/2026/ICE-and-Ornn-to-Launch-GPU-Compute-Futures-Contracts/default.aspx.
[7] NYMEX Submission No. 26-370, Voluntary Submission for Product Approval (Aug. 11, 2026), https://www.cftc.gov/filings/ptc/ptc08112615405.pdf; Letter from DJ Hennes, Acting Director, Division of Market Oversight, to Maureen Guilfoile, CME Group (Sept. 21, 2026) (extending review under Regulation 40.3(c)(2) “until the end of November 9, 2026”), https://www.cftc.gov/filings/documents/2026/orgdcmnymexcompcontr260921.pdf.
[8] ICE Futures U.S., Exchange Notice, New Compute Index Reference Data to Be Made Available on Monday, October 5, 2026 (Sept. 29, 2026), https://www.ice.com/publicdocs/futures_us/exchange_notices/ICE_Futures_US_Exchange_Notice_ORNN_Compute_4Q2026_20260929.pdf.
[9] See QCX LLC d/b/a Polymarket US, Compute Index Value Contracts, Terms and Conditions (Aug. 20, 2026), https://www.cftc.gov/filings/ptc/ptc08202618957.pdf; Sebastian Moss, Kalshi Users Able to Wager on Nvidia GPU Compute Prices, Based on Ornn’s Compute Derivatives Platform, DatacenterDynamics (Mar. 4, 2026), https://www.datacenterdynamics.com/en/news/kalshi-users-able-to-gamble-on-nvidia-gpu-compute-prices-based-on-ornns-compute-derivatives-platform/; Naga Avan-Nomayo, Crypto-Style Derivatives Reach AI Compute Ahead of Planned CME and ICE Futures: Bernstein, The Block (July 17, 2026) (reporting that Architect’s AX exchange, registered in Bermuda, “now lists perpetual-style compute futures”), https://www.theblock.co/post/408758/crypto-style-derivatives-reach-ai-compute-ahead-of-planned-cme-and-ice-futures-bernstein.
[10] Request, 91 Fed. Reg. at 54262 & n.45 (citing electricity); id. at 54263 (Question 1(a)).
[11] CME Group, PJM Western Hub Peak Calendar-Month Real-Time LMP Futures, Contract Specifications (Floating Price is “the arithmetic average of the PJM Western Hub Real-Time LMP for peak hours”), https://cmegroup.com/markets/energy/electricity/pjm-western-hub-peak-calendar-month-real-time-lmp.contractSpecs.html.
[12] Baltic Exchange, FFAs (“Freight futures are Asian style, meaning that they settle on the average of the underlying spot Baltic index”; naming CME, EEX, ICE and SGX as clearing providers), https://www.balticexchange.com/en/FFAS.html.
[13] NYMEX, Special Executive Report No. 8031 (Nov. 21, 2017) (for the Freight Route TD3C (Baltic) Futures, the Floating Price “is equal to the arithmetic average of the rates for each business day” that the Baltic Exchange publishes for the route during the contract month), https://www.cftc.gov/filings/ptc/ptc011018nymexdcm003.pdf; see also NYMEX Rulebook ch. 684 (Freight Route Middle East to China (TD3C) (Baltic) Futures), https://www.cmegroup.com/rulebook/NYMEX/6/684.pdf.
[14] Baltic Exchange, Frequently Asked Questions (“Baltic Data is based on assessments made by Panellist Brokers”; “Shipowners or charterers cannot be panellists”; Baltic Exchange Information Services Ltd “is authorised by the UK’s Financial Conduct Authority”), https://www.balticexchange.com/en/who-we-are/faqs.html.
[15] Request, 91 Fed. Reg. at 54263 (Question 2(b)); see also supra note 6.
[16] See, e.g., CFTC Release No. 4728-02 (Dec. 19, 2002) (Dynegy Marketing & Trade and West Coast Power LLC; attempted manipulation and false reporting), https://www.cftc.gov/opa/enf02/opa4728-02.htm; CFTC Release No. 4846-03 (Oct. 1, 2003) (complaint alleging false reporting and attempted manipulation by American Electric Power and AEP Energy Services), https://www.cftc.gov/opa/enf03/opa4846-03.htm; CFTC Release No. 4869-03 (Nov. 25, 2003) (orders against Reliant Energy Services and CMS subsidiaries; quoting the Director of Enforcement: “Over the past 11 months the Commission has imposed more than $130 million in civil penalties”), https://www.cftc.gov/opa/enf03/opa4869-03.htm; see also In re Barclays PLC, CFTC No. 12-25 (June 27, 2012) (LIBOR and Euribor), https://www.cftc.gov/sites/default/files/idc/groups/public/@lrenforcementactions/documents/legalpleading/enfbarclaysorder062712.pdf.
[17] Policy Statement on Natural Gas and Electric Price Indices, 104 FERC ¶ 61,121 (July 24, 2003) (Docket No. PL03-3-000).
[18] See Comment on RIN 3038-ZA24, supra note 2 (urging a rule rather than serial no-action letters); Comment on RIN 3038-AF73, supra note 2 (supporting codification of the event contract reporting framework); Letter of May 20, 2025, supra note 2 (urging formal rulemaking on perpetual futures).
[19] Request, 91 Fed. Reg. at 54263 (Question 1(c)).
[20] 17 C.F.R. pt. 38, app. C, para. (c)(3), (c)(3)(i).
[21] Other commenters make related points. See Comment of GPU Economy at 2, 9 (posted Oct. 5, 2026) (a settlement reference “should not blend” posted rates with marketplace prices), https://www.regulations.gov/comment/CFTC-2026-1850-0025; Comment of Compute Capital (USA), Inc. at 4 (Sept. 25, 2026) (a contract that settles on posted supplier rates “is readily susceptible to manipulation within the meaning of Appendix C” unless the exchange can see, and the Commission can reconstruct, the bilateral market), https://www.regulations.gov/comment/CFTC-2026-1850-0018.
[22] CME Group Press Release, supra note 6; ICE and Ornn Press Release, supra note 6 (“Ornn Exchange acts as a venue for compute risk transfer.”); see also Comment on RIN 3038-ZA24, supra note 2.
[23] Int’l Org. of Sec. Comm’ns, Principles for Financial Benchmarks: Final Report, FR07/13 (July 2013), https://www.iosco.org/library/pubdocs/pdf/IOSCOPD415.pdf.
[24] 104 FERC ¶ 61,121, supra note 17, at P 33.
[25] 17 C.F.R. pt. 38, app. C, para. (c)(3)(v).
[26] Dodd-Frank Wall Street Reform and Consumer Protection Act, Pub. L. No. 111-203, § 753(a), 124 Stat. 1376, 1750 (2010); 7 U.S.C. § 9(1)(A), (C); see also 17 C.F.R. § 180.1(a)(4) (implementing both provisions).
[27] 7 U.S.C. § 13(a)(2).
[28] Comment of GPU Economy, supra note 21, at 2, 7 (“One provider moving its rates in every region it lists can shift [GPU Economy’s weekly median] by up to 10.5%, and up to 12.8% in our smaller fixed monthly panel.”); id. at 10 (“A rate offered with nothing behind it is the one route to influence that public data cannot close.”).
[29] 7 U.S.C. § 6c(a)(5) (reaching “trading, practice, or conduct on or subject to the rules of a registered entity”).
[30] See 17 C.F.R. § 180.1(a)(2), (4) (intentional or reckless false statements and reports); 7 U.S.C. § 9(3); 17 C.F.R. § 180.2 (price manipulation); Prohibition on the Employment, or Attempted Employment, of Manipulative and Deceptive Devices and Prohibition on Price Manipulation, 76 Fed. Reg. 41398, 41407 (July 14, 2011) (Rule 180.2 requires “the requisite specific intent”; “recklessness will not suffice”).
[31] See Comment of GPU Economy, supra note 21, at 10 (suggesting consideration of whether panel members should “attest that a posted rate could be booked at the time it was read”).
[32] 104 FERC ¶ 61,121, supra note 17, at P 34; see id. at PP 5, 38.
[33] Polymarket US Terms and Conditions, supra note 9 (Eligibility).
[34] NYMEX Rulebook chs. 1045 and 1047 (Floating Price is the arithmetic average of the index’s on-demand settlement prices for each Business Day of the contract month), in NYMEX Submission No. 26-370, supra note 7; ICE Futures U.S. Exchange Notice, supra note 8 (final settlement is the average of daily index prices for the contract month).
[35] NYMEX Rulebook chs. 1045 and 1047 (on-demand settlement prices; “Geography: United States”), supra note 34.
[36] Request, 91 Fed. Reg. at 54263 (Question 1(b)).
[37] 17 C.F.R. pt. 38, app. C, para. (c)(2).
[38] 17 C.F.R. § 150.5(b)(1)(i)–(ii), (b)(2); see 17 C.F.R. § 150.1 (defining “physical commodity” to include any exempt commodity).
[39] Comment on RIN 3038-AF73, supra note 2, at 2, 7 (inviting the Commission to consider whether aggregated position data for covered event contracts “warrants periodic publication on the model of the Commitments of Traders reports”).
[40] Letter of May 20, 2025, supra note 2.
[41] Id. (the CFTC “should address perpetual futures contracts through formal rulemaking rather than through enforcement actions”).
[42] Chi. Mercantile Exch. Inc. v. Selig, No. 1:26-cv-02157 (D.D.C. filed June 18, 2026) (challenging the Commission’s May 29, 2026 order approving KalshiEX LLC’s bitcoin perpetual contract for listing as a futures contract and the accompanying Policy Statement Concerning the Listing of Perpetual Contracts).
[43] At least one commenter predicts that perpetual compute futures “will function more like swaps products.” Comment of Capital Markets & Technology Partners at 18 (Sept. 15, 2026), https://www.regulations.gov/comment/CFTC-2026-1850-0010.
[44] 31 C.F.R. §§ 1026.210 (AML program), 1026.220 (customer identification), 1010.230 (beneficial ownership), 1026.320 (suspicious activity reports).
[45] Request, 91 Fed. Reg. at 54264 (Question 3(b)).
[46] Request, 91 Fed. Reg. at 54260 n.6.
[47] Further Definition of “Swap,” “Security-Based Swap,” and “Security-Based Swap Agreement”; Mixed Swaps; Security-Based Swap Agreement Recordkeeping, 77 Fed. Reg. 48208, 48233 (Aug. 13, 2012) (intangible commodities); see id. at 48238–42 (volumetric optionality, including capacity contracts and tolling agreements, and agreements for the “usage of a specified facility”).
[48] Architect Financial Technologies, Press Release, Architect Financial Technologies Partners with Compute Desk to Launch Exchange-for-Physical Market for GPU Compute (July 8, 2026) (describing a network that would let customers of Architect’s forthcoming designated contract market “convert compute futures positions into real GPU capacity”), https://www.prnewswire.com/news-releases/architect-financial-technologies-partners-with-compute-desk-to-launch-exchange-for-physical-market-for-gpu-compute-302820762.html.
[49] See 7 U.S.C. §§ 6(a) (futures), 7b-3(a)(1) (swaps).
De Silva Law Offices, LLC advises exchanges, trading firms, intermediaries and companies entering CFTC-regulated markets for the first time, including compute providers, compute marketplaces and benchmark administrators. Our futures and derivatives practice covers product classification, exchange and intermediary registration, compliance and enforcement defense. Questions about this letter may be directed to info@desilvalawoffices.com or 312-500-8424.
NB: This comment letter is a matter of public record and is not legal advice. Reading it does not create an attorney-client relationship. R Tamara de Silva is the Managing Partner of De Silva Law Offices, LLC, Chicago, and writes on financial regulation, derivatives and prediction markets.