Research · market structure · updated 2026-10-21 · docket CFTC-2026-1850
Compute Derivatives: A Summary of Public Comments
The CFTC docket on compute derivatives received 103 submissions. We count 92 letters after removing duplicates, a withdrawn upload and short comments that took no position on the questions. This summary groups the letters by their positions on the key questions the CFTC asked. Each letter, with a summary and a link to the author's full letter, is listed on our CFTC monitor.
- 01 · Whether Contracts Should List · Question 1 of 17
Should compute contracts list?
- List once named conditions are met69%
- List the contracts now22%
- The cash market cannot support a sound contract today9%
“The comment explains how, with a worked example, and asks the Commission to require five safeguards before a designated contract market (DCM) lists compute derivatives”
“Dragonfly urges the Commission to apply these existing, flexible principles without imposing new one-size-fits-all audit or contributor requirements that could discourage market entry.”
“The present underlying market is not ready to support listed derivatives.”
- 01 · Whether Contracts Should List · Question 2 of 17
Does the CFTC need new rules?
- Compute-specific guidance under existing rules59%
- No new rules or guidance needed34%
- New rules or a new regime7%
“It should articulate what a robust compute settlement index looks like and let DCMs list contracts that settle to indices meeting that standard.”
“The Commission does not require novel statutory authorities to oversee compute derivatives.”
“the Commission should mandate regulator-accessible reporting of bilateral prices and capacity, vest governance in an independent body”
- 02 · What Can Set the Settlement Price · Question 3 of 17
Can posted rates feed the settlement price?
- Only if verified, executable, capped or disclosed73%
- Executed transactions only22%
- Posted rates can feed the price5%
“Consider whether panel members should attest that a posted rate could be booked at the time it was read.”
“build settlement indexes from verified, actual transactions, not providers' self-reported or posted rates”
“In compute, the conventional distinction between quotes and trades is less meaningful than in other commodities because posted rates are directly executable through API calls, while reported transaction data remains comparatively sparse.”
- 02 · What Can Set the Settlement Price · Question 4 of 17
Should contributor floors and caps be fixed?
- Each exchange or administrator sets and discloses its own65%
- Publish counts and concentration, no fixed floor23%
- Set a fixed minimum or cap12%
“As illustrative starting points, not empirical findings: at least five unaffiliated contributors and no group above one quarter of weighted inputs, calibrated and disclosed by the DCM.”
“disclosure, with each published value, of the number of contributors and the concentration of contribution.”
“we suggest that a settlement index be calculated each business day from no fewer than 5 independent providers, with no single provider accounting for more than 20% of index weight.”
- 03 · What One Contract Measures · Question 5 of 17
What should one contract measure?
- One tightly defined GPU-hour34%
- A reference grade plus priced differentials31%
- Other: leave it to exchanges and the market19%
- A unit of performance or delivered work11%
- Other: tokens or inference output5%
These answers are different designs, not points on one scale, so they carry their own colors.
“Define the underlier as a bounded economic identity: contract terms should specify the dimensions that materially determine equivalence rather than treating "compute" as a unitary commodity.”
“Agricultural futures solved a version of this problem with published grade differentials; compute needs the same discipline.”
“the market should be permitted to determine which underliers, contract grades, settlement structures, and reference prices prove most useful.”
- 04 · Who Runs and Checks the Index · Question 6 of 17
Can the index administrator trade the contract or hold compute?
- Conflicts allowed, disclosed and managed55%
- Administrator may not trade the contract or hold compute45%
“the administrator should keep its index business separate from any trading interest and disclose its ownership”
“an administrator should not trade the derivative or hold material physical exposure to the compute capacity its index measures.”
- 04 · Who Runs and Checks the Index · Question 7 of 17
Should an auditor check the index data?
- Required as a condition67%
- Encouraged28%
- Not a listing condition4%
Question 1(d) asks whether an independent party should audit the index and the prices that feed it.
“price reports should be audited against the underlying transaction and delivery records”
“An index whose values can be rebuilt from retained inputs can be audited.”
“the existing protections described above in the context of natural gas are sufficient.”
- 04 · Who Runs and Checks the Index · Question 8 of 17
Who sees the inputs behind each price?
- Detail to the exchange and CFTC; summaries public84%
- Input-level detail published16%
“A listing package should therefore establish enforceable access to underlying reporter evidence, including counterparty and affiliation information when needed for an investigation, through arrangements that protect confidentiality.”
“Methodology and raw panel data must be public and independently recalculable.”
- 04 · Who Runs and Checks the Index · Question 9 of 17
Should the exchange get access to each source's data?
- One arrangement with the index administrator54%
- An arrangement reaching every source31%
- No mandatory arrangement15%
Question 2(d) asks whether an exchange should hold an information-sharing arrangement with each venue and provider whose data enters the settlement price. The arrangement lets exchange surveillance staff see the trades behind the price.
“Information-sharing arrangements should be enforceable, give timely access to contributor and counterparty details, and reach through any administrator to the underlying evidence.”
“Information-sharing with venues is necessary and not sufficient.”
“requiring such arrangements would give compute suppliers an effective veto over whether a contract can be listed at all”
- 05 · Perpetual Futures and the Review Path · Question 10 of 17
Should perpetual compute futures be allowed?
- Allow with added safeguards49%
- Not at first29%
- Allow perpetuals20%
- Do not allow2%
“For any compute contract — and especially any perpetual — require the DCM to assess the channel between derivative price signals and automated physical capacity management, with rate limits or confirmation gates on redemption-driven provisioning”
“Defer broad retail access and perpetual compute futures until fixed-date markets, reference prices and surveillance arrangements have demonstrated resilience”
“The Commission should therefore permit exchanges to develop both structures where the contracts satisfy applicable requirements under the Commodity Exchange Act and Commission regulations”
- 05 · Perpetual Futures and the Review Path · Question 11 of 17
How should the first contracts be reviewed?
- Self-certification under existing rules50%
- Commission approval or added review50%
“DCMs should also retain access to the Part 40 self-certification process for new compute contracts.”
“Direct that first-of-kind compute contracts proceed by Commission approval under 17 CFR 40.3 rather than self-certification under 17 CFR 40.2 for a defined initial period”
- 06 · Liquidity and Position Limits · Question 12 of 17
Where does liquidity come from?
- Build it on purpose: market makers, one instrument, fewer months46%
- A listed contract creates it36%
- The cash or OTC market has to come first18%
Question 1(b) asks how to measure the size and liquidity of compute markets, which bears on whether the market is deep enough to support a contract. Letters divide on where liquidity comes from. Compute Capital writes that listed futures “inherit liquidity from the OTC market they sit next to.” Douro Labs writes that “a liquid listed contract is what standardizes and deepens an underlying cash market.”
“Listings should concentrate liquidity on high-granularity tenors within a 12-month window.”
“A well-regulated compute futures market will strengthen the underlying cash market by promoting common commercial specifications, supporting transparent price discovery, and expanding hedging opportunities.”
“the “danger is excessive speculation without a strong underlying cash market, which can create noisy price discovery and poor hedging performance.””
- 06 · Liquidity and Position Limits · Question 13 of 17
Position limits or accountability levels?
- Hard limits sized to the cash market67%
- Accountability levels at the exchange’s discretion33%
Question 1(g) asks how to estimate deliverable supply for position limits or accountability levels.
“The Commission should size position limits against transferable capacity, not against the market's apparent scale, precisely because the apparent scale is so much larger than the corner-relevant quantity.”
“On position limits and accountability, PMEX believes that, given current data limitations, position accountability may be more readily calibrated than a hard numerical limit.”
- 07 · Customer Protection · Question 14 of 17
Should retail traders have access?
- Open to retail with standard protections59%
- Retail with added limits25%
- Institutions only at first16%
Question 3(d) asks what protections retail users of compute derivatives need.
“Do not create a bespoke retail regime keyed to the underlier.”
“We support proportionate retail protections based on leverage, complexity, liquidity and delivery obligations. Interfaces should distinguish purchasing compute from trading a derivative, with clear loss scenarios and appropriate access checks and limits.”
“I would be hesitant to make compute derivatives available to retail before any reference-pricing structure discussed under Question 2(g) exists.”
- 07 · Customer Protection · Question 15 of 17
Export controls
27 of 92 letters ask the CFTC to weigh export controls on advanced chips, or to have exchanges disclose that risk.
Question 3(b) asks about disclosure for a contract that settles against a “geopolitically sensitive commodity.”
“Consult BIS and OFAC on the export-control and sanctions implications of synthetic exposure to restricted compute capacity before the first compute contracts are listed.”
“Consult the Department of Commerce so that no compute contract, particularly a physically settled one, becomes a channel for parties restricted under U.S. export controls to obtain compute.”
“The greater risk is at any physical-delivery or EFRP leg, where delivered compute could be used to evade export controls or sanctions, so those legs should require identification of the ultimate capacity user and jurisdictional screening (Q3(a)).”
- 08 · Also Raised by Commenters · Question 16 of 17
Physical delivery alongside cash settlement
21 of 92 letters argue for physical delivery of compute alongside cash settlement.
The CFTC's questions assume cash-settled contracts. Another 6 letters argue for cash settlement alone.
“The Commission should not presume that cash settlement is the only viable design.”
“Hyperbolic agrees that cash settlement will come first but urges the CFTC to recognize that physical delivery is practical.”
“We encourage the Commission to accommodate physical delivery alongside cash settlement, with enforceable access rights, clear grading, and realistic deliverable supply estimates.”
- 08 · Also Raised by Commenters · Question 17 of 17
Should token or inference contracts list?
- Allow token or inference contracts now64%
- Later, after GPU-hour contracts or once measurable36%
The CFTC's questions treat compute as GPU time.
“Both forms of compute are suitable underliers for derivatives contracts, and they serve different participants.”
“For contracts based on inference service we are missing a common specification of the deliverable, followed by evidence of transactions and delivery within that specification.”
09 The Full Record
Every letter's position on each question, with an excerpt. Each name links to the letter on the docket.
- → 77 of 92 letters took a position
- → 10 state more than one position
- → Secondary positions are marked in the dropdown below
Commenters and excerpts
List once named conditions are met (53)
- 1791 Technologies: “This response does not recommend prohibiting compute derivatives. It recommends making the transformation from heterogeneous market observations to a legally operative settlement value auditable and explicitly scoped.”
- Abaxx Technologies: “That sparsity is not a reason to forgo listing. It is a reason to make data production a condition of cash-settled listing”
- Alec Snelling: “I support allowing designated contract markets to list compute derivatives, with guardrails.”
- Andrew Huff: “On Question 1(c), I would not permit a contract to settle to a price computed from data that the DCM and the Commission cannot observe or verify on request.”
- Animica Labs: “The Commission's most durable available step is to make observability a listing condition rather than a hoped-for disclosure”
- Architect Financial Technologies: “A DCM should list a compute derivative only if the DCM is able to demonstrate a precisely defined underlying commodity, a settlement index that is not readily susceptible to manipulation”
- Babbage: “Listing proponents should demonstrate suitability for the particular contract.”
- CCE: “We urge the Commission to permit the development of well-designed compute derivatives while requiring evidence that settlement and delivery arrangements can support market integrity.”
- Compute Capital: “Compute Capital does not ask the Commission to block listed compute futures. It asks the Commission to treat listed futures and institutional OTC as one market-structure project.”
- Compute Desk: “should not restrain the listing of compute derivatives, so long as the price references a reliable, transparent index.”
- Crypto Council for Innovation: “compute contracts may warrant particularly careful review of benchmark representativeness, source independence, manipulation resistance, and settlement methodology before listing.”
- Dartmouth Advisory Partners: “The Commission should list compute contracts only against indices whose inputs the exchange and the Commission can observe, reconstruct after the fact, and classify the same way across venues.”
- De Silva Law Offices: “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.”
- DNA Compute: “We support the listing of regulated compute derivatives, and we do not ask the Commission to delay or withhold approval of any contract.”
- Evelyse Carvalho Ribas: “I suggest that the productive question is not whether an existing series can be made to satisfy the criteria, but what governance a compute index would have to have before a contract could sensibly settle to it.”
- Federico M. Bandi and Yinan Su: “While we do not believe that compute’s distinctive features are an insurmountable obstacle to a viable futures market, compliance with the IOSCO principles requires care and”
- Fiftyrivers: “We think the Commission should expect DCMs to build these deliberately rather than assume the cash market will supply them.”
- Fullview: “If no series satisfies the relevant requirements, it should develop adequate support or reconsider the contract design before relying on the reference for listing.”
- Futures Industry Association: “FIA believes strongly that only contracts that can be effectively risk-managed should be listed and cleared.”
- Gemini: “provided the integrity of the index is assured through governance, methodology, and auditability.”
- Hedge Quantitative Research: “The public record supports listing a monthly, cash-settled contract on a GPU rental index for the owners and lenders who carry the exposure, provided the Commission can see the index's inputs”
- Houman Shadab: “Non-public data is therefore not disqualifying under existing guidance. Unverifiable data should be.”
- Hydra Host: “Require that any such index be built only from real, completed rental transactions for graded compute, reported once by the party closest to the servers”
- Hyperbolic Labs: “What the Commission should do if no qualifying series is available is permit the listing exchange to identify or construct a reference that meets the criteria, publish the methodology and governance, and demonstrate reliability.”
- ICX Global: “We believe, however, that benchmark transparency, data quality, contributor independence and governance need to be established before the market grows to scale.”
- InfraSight Software: “Compute derivatives can support the AI Action Plan's goal of broader access to compute, but only if the thing being priced is well defined.”
- Kottos AI: “Before listing, the DCM should demonstrate which service attributes can be independently verified and how remaining verification gaps are handled.”
- Li Li: “I support the development of compute derivatives within an appropriate regulatory framework.”
- Liquid Compute (PMEX Markets): “PMEX does not believe a listed contract should settle to a reference price whose material inputs cannot be made available, on an appropriate confidential basis, to the listing DCM and the Commission for verification and surveillance.”
- Mercatus AI: “Mercatus AI supports a path for cash-settled compute products under clear standards.”
- NativX: “We ask the Commission to recognize energy-normalized compute benchmarks as a viable basis for derivatives, permit appropriately controlled reports of completed bilateral sales as inputs, and assess listing readiness through demonstrated liquidity, hedging utility and surveillance access.”
- Ola Ologunde: “Compute can support standardised derivatives, but only if the market agrees on what is being priced, settlement sources disclose how they adjust for it, and participants understand what a rent contract leaves uncovered.”
- Ömer Demirel: “The comment explains how, with a worked example, and asks the Commission to require five safeguards before a designated contract market (DCM) lists compute derivatives”
- OneChronos: “Compute derivatives can be listed today, provided contract design begins where the asset is well defined.”
- Persistence Analytics Group: “Before compute can function as a reliable financial underlier, the market should establish what specific economic object is being represented, what evidence supports that representation”
- PSYCHE: “A cash-settled contract can be written and listed, on the showing Core Principle 3 and Appendix C ¶(c) already require of its settlement reference, against a published compute price series whose dimensions and change control the contract rules describe”
- PTG: “The Commission should confirm that compute derivatives can list on regulated U.S. DCMs when the contract is well specified and the settlement reference is robust.”
- Samsung Next Fund: “We also believe that appropriate safeguards would be important to the development of any such market.”
- Setara Financial: “Setara strongly supports allowing regulated U.S. venues to develop compute derivatives subject to appropriate standards for market integrity, customer protection, benchmark quality, transparency, settlement and risk management.”
- Silicon Data: “Instead, each contract proposal should be evaluated and approved independently under the Core Principles according to the maturity and readiness of its specific underlying infrastructure.”
- Simeon Bochev: “Approval against stated conditions, with mandatory composition disclosure, seems to me to dominate both prohibition and approval as filed.”
- Solutions by Jewel: “SBJ recommends that each DCM submission under 17 CFR 40.2 or 40.3 for a compute derivative include an Input Evidence Profile for the settlement reference price.”
- Squaretower: “If done correctly, these financial instruments could enable American capital markets to drive and sustain the greatest industrial buildout of our present time.”
- Steven Singleton: “A concrete fix would condition self-certification eligibility on public disclosure of the number and market share of rate-contributing providers feeding any settlement index”
- Steven Skinner: “I would not list such a contract unless the index inputs are tied to records of what was actually delivered, and the exchange can reach those records.”
- Submitter listed as "Grok": “The next step is a framework that lets principals offer short-tenor forwards on verified capacity, settled to prices no single supplier controls.”
- Taylor Holland: “Cash settled compute futures based on an index should disclose, in addition to index data sources, calculation methodology, and price history, the index’s limitations and reasons it may not fully proxy and converge to compute’s underlying economic cost.”
- TCX: “Require DCMs seeking to list compute derivatives to demonstrate that settlement indices rely on verified, arm's-length execution data rather than unverified supplier rate cards or self-reported surveys.”
- Trevor Love: “Before a broad compute derivative is listed, the Commission should require the listing DCM to: 1. Define the underlier by operational service tier, including availability, scheduling, and interruption rights.”
- Vandermolen Power Limited: “Begin with narrowly defined, fixed-date contracts and clearly specified hardware or workload profiles; broaden only when transaction data and conversion evidence support it.”
- Vast.ai: “Vast supports well-regulated compute derivatives markets that help compute buyers and providers manage price risk.”
- VeloxVFX: “VeloxVFX cannot conclude that a derivative settling to data that the Commission cannot adequately observe, verify, or surveil is appropriate absent a demonstrated evidentiary and surveillance basis sufficient for the applicable contract and Core Principle obligations.”
- Wei Wang: “I recommend the Commission make the following verifiability thresholds a precondition for listing any compute futures contract”
- Secondary position ACU Markets: “It is not a reason to abandon compute derivatives; it is a reason to set the conditions before contracts are listed rather than after.”
- Secondary position ACU Markets: “A contract can list at Stage 2 on a single specified grade, which is approximately what is before the Commission now.”
- Secondary position Douro Labs: “It should articulate what a robust compute settlement index looks like and let DCMs list contracts that settle to indices meeting that standard.”
- Secondary position Ornn AI: “The relevant question is whether a particular contract defines the exposure with sufficient precision, uses a reliable settlement mechanism, and is not readily susceptible to manipulation.”
- Secondary position Submitter listed as "Nico": “Before a DCM lists such a contract, the Commission should require the submission to state (i) which posted series, (ii) who administers it, (iii) whether transacted volume is captured, and (iv) what independent verification of that data exists”
List the contracts now (17)
- Andreessen Horowitz (a16z): “The self-certification and approval procedures in Part 40 are sufficient for these products, and new rulemaking is not needed before listing.”
- CME Group: “CME believes it is an appropriate, if not necessary, time for compute derivatives contracts to be offered to the industry for risk management and price discovery purposes.”
- Douro Labs: “The Commission should therefore not condition listing on the maturation or standardization of the compute cash market.”
- Dragonfly: “Dragonfly urges the Commission to apply these existing, flexible principles without imposing new one-size-fits-all audit or contributor requirements that could discourage market entry.”
- DRW Holdings: “Permit well-designed conventional compute futures to proceed on the ordinary product-listing path without further delay”
- Hivemind Capital Partners: “Advance DCM/SEF Listings for Cash-Settled Rolling GPU Benchmarks: Provide clear guidance affirming that transparent, volume-weighted median benchmark methodologies meet CEA Core Principle 3 standards.”
- Hyperliquid Policy Center: “The Commission therefore should not condition listing on a level of cash-market transparency that the market can develop through trading”
- Intercontinental Exchange: “Such traits are not unique to compute and should not be a barrier to analysis under the existing Core Principle 3 framework with which the Commission has vast experience.”
- Jump Trading Group: “The Commission should permit a broad range of qualified DCMs to compete in developing compute products, consistent with the Commodity Exchange Act”
- Kalshi: “A categorical ban on derivatives contracts for compute or a delay in introducing these contracts at a time when the market is already evolving is harmful to American SI policy and goals”
- Nodal Exchange: “the challenges identified in the RFC are not unique and should not be an impediment to the successful launch of compute derivatives.”
- OrchestraOS: “I recommend no particular benchmark or administrator, and do not ask the Commission to delay listing until one is adopted.”
- Ornn AI: “The heterogeneity of compute should be addressed through careful contract and benchmark design rather than treated as a categorical barrier to market development.”
- Paradigm and Andromeda: “We therefore urge the Commission to permit DCMs and others to develop a nd launch compute derivatives under appropriate oversight but without the delays or onerous conditions that would hamstring these critical markets.”
- Polymarket US: “Holding the listing of these contracts pending selection of a single benchmark would not meaningfully add to these tools and would delay the hedging instruments the rapidly evolving market needs.”
- The Digital Chamber: “The current opacity of the underlying compute market is a reason to encourage, rather than delay, the development of listed derivatives.”
- Voltry: “We do not ask the Commission to delay any contract, to prefer any index, or to mandate any provider.”
- Secondary position Architect Financial Technologies: “Architect believes that GPU-hour underliers are suitable for listing now.”
- Secondary position PTG: “Compute derivatives raise new questions, but they can be answered through contract design and clear standards rather than delay.”
- Secondary position Silicon Data: “We encourage the Commission not to delay listing until the physical market standardizes, nor to treat all compute derivative proposals as a single monolithic category.”
The cash market cannot support a sound contract today (7)
- ACU Markets: “The cash market does not today support a cash-settled contract that satisfies Core Principle 3.”
- ACU Markets: “Our first submission told the Commission that the compute cash market cannot yet support a sound cash-settled contract.”
- Balthasar Issa: “The volume and concentration data needed for this doesn’t seem to be publicly available at present, which in my opinion is an argument for not listing these contracts before that data exists and can be verified.”
- Better Markets: “The Commission must ensure that compute derivatives contracts are able to be standardized, and are not susceptible to manipulation, before allowing them to trade on a DCM.”
- Capital Markets & Technology Partners: “The present underlying market is not ready to support listed derivatives.”
- Michael Ravnitzky: “The Commission should refrain from the listing of compute derivatives until a standardized, independently governed benchmark is in place and mandatory, verifiable reporting from power-dependent compute providers is established.”
- Submitter listed as "Nico": “I comment on one narrow question: whether a compute cash market today publishes a price a DCM could honestly settle against under Appendix C and Core Principle 3. Answer: it does not.”
- Secondary position Abaxx Technologies: “a cash-settled compute contract should not be listed until (a) an independently governed, multi-provider benchmark exists and (b) mandatory data sharing (utilization, posted rates, committed capacity) is in place.”
- Secondary position Simeon Bochev: “I do not believe any compute cash price series available today satisfies the Appendix C criteria.”
- → 59 of 92 letters took a position
- → 20 state more than one position
- → Secondary positions are marked in the dropdown below
Commenters and excerpts
Compute-specific guidance under existing rules (35)
- Abaxx Technologies: “Specify the settlement instrument in guidance. In any guidance or rulemaking on compute derivative listing, expressly recognize settlement by transfer of a standardized verifiable claim ticket as physical delivery”
- ACU Markets: “Publish compute-specific guidance under 17 CFR part 38, Appendix C — or an addendum to it — setting out what a compute settlement reference price must demonstrate”
- ACU Markets: “We recommend that the Commission say so, in guidance or in a statement accompanying the disposition of this proceeding.”
- Animica Labs: “We respectfully recommend that, should the Commission provide guidance on compute derivatives, it condition the acceptability of a settlement reference price on properties that are verifiable by the Commission without provider cooperation”
- Architect Financial Technologies: “Architect recommends that Commission staff issue a short, informational advisory on compute derivative listings, similar in form to CFTC Staff Advisory No. 26-16, that creates no new obligations.”
- Babbage: “We ask the Commission to encourage common minimum service disclosures, qualification supported by evidence, and verifiable prices.”
- Compute Capital: “Recognize, in any order or guidance that follows this Request, that a compute future listed in isolation from a documented, reportable OTC book will not produce the robust market the Chairman described.”
- De Silva Law Offices: “The Commission should set out its expectations for compute derivatives in supplemental guidance to Appendix C to Part 38, proposed for comment.”
- DNA Compute: “If no series presently meets these criteria, the Commission's step should be to publish the criteria, measure the candidates against them in the interval before listing, and state the result in a staff advisory”
- Douro Labs: “It should articulate what a robust compute settlement index looks like and let DCMs list contracts that settle to indices meeting that standard.”
- DRW Holdings: “We strongly encourage the Commission to articulate its expectations for administrators of compute settlement indices.”
- Fiftyrivers: “We think the Commission should expect DCMs to build these deliberately rather than assume the cash market will supply them.”
- Futures Industry Association: “Provide guidance or otherwise clarify that DCMs should seek to use index administrators for Compute benchmarks”
- Gemini: “Gemini recommends that the Commission, through guidance rather than rulemaking, identify the following”
- Hivemind Capital Partners: “Provide clear guidance affirming that transparent, volume-weighted median benchmark methodologies meet CEA Core Principle 3 standards.”
- Houman Shadab: “They ask the Commission to say how cryptographic evidence fits the framework it already administers, not to build a parallel one for compute.”
- Hydra Host: “Hydra recommends that the Commission encourage the adoption of compute quality grades.”
- Hyperbolic Labs: “The Commission can accelerate that process by stating the criteria it will apply, by confirming that the Act’s prohibitions on false reporting reach contributions to a compute index”
- Hyperliquid Policy Center: “The Commission should articulate what a reliable compute settlement index looks like and let DCMs list contracts that settle to indices meeting that standard”
- INFEREXUS: “market participants would benefit from early Commission guidance on how AI inference capacity is classified as a commodity under the Commodity Exchange Act”
- InfraSight Software: “In any guidance that follows this request, acknowledge that standardizing compute may require units of delivered work, not only standardized rental terms for specific chips.”
- Liquid Compute (PMEX Markets): “We encourage the Commission to articulate expectations along the following lines for any index used to settle a listed compute contract”
- Mercatus AI: “Appendix C to Part 38, paragraphs (c)(3)(i)–(iii), already covers both cases; compute-specific guidance should explain how a DCM shows these safeguards work in practice.”
- NativX: “We ask the Commission to reflect the following principles in any guidance resulting from this Request and in its evaluation of compute derivatives.”
- Paradigm and Andromeda: “we urge the Commission to issue guidance clarifying how exchanges can demonstrate, under the existing regulatory framework, the quality and transparency of indices underlying listed compute derivatives”
- PTG: “The Commission should instead articulate the standards that any compute settlement reference should satisfy.”
- Samsung Next Fund: “proactive guidance and regulatory supervision within the United States can assist in setting credible global benchmarks for market integrity and price discovery”
- Silicon Data: “We recommend that the Commission make clear, in applying its Appendix C guidance, that a DCM listing a cash-settled compute contract should verify that the settlement index administrator adheres to the IOSCO Principles for Financial Benchmarks.”
- Simeon Bochev: “Step one: state the criteria now, in guidance rather than by rule.”
- Solutions by Jewel: “That sequence offers a practical model for compute: heightened expectations at listing, tied to a scheduled review that recalibrates them as the cash market demonstrates depth and transparency.”
- Steven Skinner: “The Commission should expect a compute settlement reference to count only GPU-hours”
- TCX: “TCX respectfully recommends that the Commission include the following technical principles in any subsequent guidance or policy statements concerning compute derivatives”
- Vandermolen Power Limited: “Publish technology-neutral minimum expectations for benchmark governance”
- Vast.ai: “The Commission should clarify how designated contract markets (“DCMs”) should exercise reasonable, documented diligence on source authorization, methodology and reliability.”
- Voltry: “We ask the Commission to state, in whatever follows this request, that the "grade, quality" clause of Appendix C applies to compute as it applies to other commodities”
- Secondary position Andreessen Horowitz (a16z): “Guidance confirming that an IOSCO-aligned, transaction-based settlement index can satisfy Appendix C to Part 38 notwithstanding a predominantly bilateral cash market would be helpful.”
No new rules or guidance needed (20)
- Andreessen Horowitz (a16z): “The Commission does not require novel statutory authorities to oversee compute derivatives.”
- Arkhai (CoopHive): “These recommendations should be considered within the existing legal framework.”
- Capital Markets & Technology Partners: “We would expect that the existing regulatory authorities (CME, NFA, CFTC) would be responsible for compliance oversight of the Compute Futures market, index providers and market participants (FCMs, CTAs etc.)”
- CCE: “We recommend that the Commission evaluate each product against evidence about the underlying compute rights, available supply and settlement methodology.”
- Clarke Index: “The Guidance already asks for monthly estimates covering at least the most recent three years”
- Crypto Council for Innovation: “Bitcoin offers a modern and highly relevant example of the adaptability of the commodity derivatives framework.”
- Dragonfly: “The Commission should use its existing principles to address genuine manipulation risks without imposing rigid audit mandates, minimum contributor counts, or standardized product designs”
- Fullview: “The Commission should assess whether existing rules and supervisory practices adequately address such conduct before adding compute-specific requirements.”
- Intercontinental Exchange: “we caution the Commission against issuing any “compute-specific” listing guidance as an outgrowth of this RFC.”
- Jump Trading Group: “We encourage the Commission to accommodate a range of underliers and contract designs, preserve competitive access to regulated markets, and apply its existing market integrity requirements in a manner suited to this evolving commodity.”
- Kalshi: “neither new statutory authority nor compute-specific rulemaking is necessary to permit appropriately designed derivatives to trade.”
- Nodal Exchange: “The existing Core Principles and regulations are more than sufficient to manage compute derivatives products.”
- OneChronos: “These characteristics do not place compute derivatives beyond the reach of the existing framework of the Act.”
- Ornn AI: “Ornn respectfully encourages the Commission to apply the existing CEA framework in a principles-based manner”
- Polymarket US: “in our view, no compute-specific rule is needed.”
- PSYCHE: “Both halves already have a home in the Commission’s rules.”
- The Digital Chamber: “There is no need for a separate or additional regulatory regime for compute derivatives.”
- VeloxVFX: “Where the evidentiary basis remains incomplete, the Commission may require additional evidence or controls through the applicable product-submission process under existing authority.”
- Voltry: “Appendix C to Part 38 already supplies the form: a par quality, with differentials for non-par qualities that reflect what the cash market observes.”
- Wei Wang: “This is the right question, and it has a natural statutory home: DCM Core Principle 3”
- Secondary position Architect Financial Technologies: “Architect’s central position is that the Commission’s existing, principles-based framework for DCMs can accommodate compute derivatives.”
- Secondary position Douro Labs: “Compute perpetuals can be evaluated through that existing process.”
- Secondary position DRW Holdings: “Compute does not require a new conceptual framework; it requires the existing one to be applied from the outset rather than retrofitted after a failure.”
- Secondary position Fiftyrivers: “In our view, cash-settled futures cleared through FCMs fit existing AML/KYC frameworks.”
- Secondary position Futures Industry Association: “FIA supports continued adherence to these principles from Appendix C of Part 38 with regards to Compute derivatives.”
- Secondary position Gemini: “Gemini believes the Appendix C Guidance already supplies an adequate framework for contracts settling to an index derived largely over bilateral transactions”
- Secondary position Hivemind Capital Partners: “Addressing the underlying heterogeneity of compute and its short technological shelf-life does not require inventing novel regulatory concepts.”
- Secondary position Houman Shadab: “The recommendations ask for no new regulatory structure.”
- Secondary position Hyperbolic Labs: “This is superior to prescribing a particular series by rule.”
- Secondary position Hyperliquid Policy Center: “HPC accordingly encourages the Commission to permit compute contracts to proceed through its existing review processes”
- Secondary position InfraSight Software: “We do not ask the Commission to prescribe a unit or a methodology.”
- Secondary position Mercatus AI: “Appendix C to Part 38, paragraphs (c)(3)(i)–(iii), already covers both cases”
- Secondary position Paradigm and Andromeda: “But those issues do not require a bespoke regulatory regime.”
- Secondary position PTG: “PTG does not believe compute derivatives require a new or separate listing process.”
- Secondary position Silicon Data: “The Commission Can Promote Innovation Within Existing Core Principles”
- Secondary position Simeon Bochev: “I do not recommend new conduct rules specific to compute.”
- Secondary position Vast.ai: “Vast recommends applying that framework to reasonable diligence on the sourcing and characterization of compute data.”
New rules or a new regime (4)
- 1791 Technologies: “The Commission can preserve innovation while requiring a simple set of controls: a bounded underlier identity, reconstructable benchmark ancestry, scoped representativeness, governed equivalence rules, versioned methodology, explicit degraded/fallback states, and surveillance of both source markets and the transformation layer.”
- Better Markets: “To the extent compute derivatives will be engineered for retail participants, the Commission needs investor protection requirements geared towards these users.”
- Michael Ravnitzky: “the Commission should mandate regulator-accessible reporting of bilateral prices and capacity, vest governance in an independent body”
- Steven Singleton: “Tying any retail-facing compute contract to a specific, dated disclosure requirement, rather than leaving it for unscheduled future guidance, would close this before the first contract lists”
- Secondary position De Silva Law Offices: “The firm also renews its 2025 recommendation that the Commission address perpetual contracts by rule.”
- Secondary position Jump Trading Group: “Jump requests that the Commission consider rulemaking to exclude otherwise countable dealing activity in centrally traded and cleared compute swaps from the de minimis calculation”
- → 59 of 92 letters took a position
- → 3 state more than one position
- → Secondary positions are marked in the dropdown below
Commenters and excerpts
Only if verified, executable, capped or disclosed (43)
- Abaxx Technologies: “registry data as an index input: issuance and redemption events are observed market transactions that should take precedence over asserted prices.”
- ACU Markets: “provider-administered posted rates are not admissible as a primary settlement input, and where used at all are weight-capped and disclosed”
- ACU Markets: “a public, executable posted offer observed directly by the index administrator and independently verified for fillability.”
- Andreessen Horowitz (a16z): “Where posted rates or bids and offers are used as inputs, their weight should be limited and their sources identified”
- Andrew Huff: “Admit posted rates only as a disclosed, capped share of the calculation.”
- Architect Financial Technologies: “The index uses posted rates only if the administrator verifies that they are actionable”
- Arkhai (CoopHive): “Their use for settlement needs evidence that they measure the intended exposure.”
- Babbage: “Advertised rates used in executable-offer settlement references should be verified as available for purchase.”
- Better Markets: “Because the cash market for compute is “fragmented and operationally complex,” hedgers “will not trust” a settlement index that is “based on quotes that are not executable.””
- CCE: “Settlement rules should prioritize verifiable transactions and define any use of posted quotes.”
- Compute Desk: “Posted list prices enter the index only from vetted providers and only where a buyer can transact at the posted rate.”
- Crypto Council for Innovation: “exchanges and benchmark administrators should be able to demonstrate that inputs are independently verifiable”
- De Silva Law Offices: “exchanges can expect index administrators to obtain attestations, from contributors that supply rates under contract, that those rates could be booked when read.”
- DNA Compute: “The lesson for compute is not that posted rates are illegitimate inputs; the IOSCO Principles admit firm, executable quotes where an index is anchored in real transactions”
- Dragonfly: “Practical responses may include comparing administered quotes with actual transactions, diversifying inputs when feasible, or reviewing unusual changes around an observation window.”
- Evelyse Carvalho Ribas: “a stated hierarchy of inputs that prefers observed transaction data over posted or administered rates, and that discloses the proportion of each in every published value”
- FastGPU: “require evidence that capacity was available at the posted price”
- Federico M. Bandi and Yinan Su: “Where transaction liquidity is insufficient, firm, binding quotes from a contributor panel could supplement the inputs; nonbinding indicative quotes should be distinguished”
- Fiftyrivers: “Executed, verifiable transactions first, then firm executable bids and offers, then posted rates, then judgment, with each tier’s daily share published.”
- Fullview: “Where posted rates are used, the exchange should establish whether they represent access realistically available on the stated terms.”
- Futures Industry Association: “encourage methodologies that distinguish executed transactions, executable bids and offers, and indicative posted rates; explain the hierarchy and validation of those inputs”
- Gemini: “(iii) a preference for executed or binding transactions over indicative posted rates”
- GPU Economy: “Consider whether panel members should attest that a posted rate could be booked at the time it was read.”
- Hedge Quantitative Research: “the exchange's surveillance should treat a change in a contributor's posted rate during the settlement window as a flagged event”
- Houman Shadab: “priority to bona fide executed transactions over posted rates where transactions suffice; any fallback inputs defined and disclosed in advance”
- Hyperbolic Labs: “Posted rates should be excluded or, at most, used as a fallback with published weighting when transaction volume falls below a threshold.”
- ICX Global: “The Commission should require evidence that posted rates were commercially available for the specified product, quantity and observation period”
- INFEREXUS: “Settlement price integrity depends on transparent, verifiable reference pricing from underlying providers, since compute cash markets lack the deep, independently verifiable transaction record that exists for agricultural or energy commodities.”
- Kalshi: “Just as in other indices, eligible executed transactions should normally receive greater weight than untested posted rates.”
- Liquid Compute (PMEX Markets): “Inputs should be executed transactions wherever available, with any indicative or assessed component separately disclosed as a percentage of the index.”
- Mercatus AI: “Posted rates and estimates can contribute when validated against such transactions.”
- NativX: “The relevant distinction is visible provenance and controls appropriate to each input, rather than a categorical claim that every benchmark using quotations is legally deficient.”
- Ömer Demirel: “Use executed contracts only, or give posted rates very low weight”
- Paradigm and Andromeda: “(3) are grounded in transaction- based inputs or, where quotes are used, are labeled as such”
- Polymarket US: “Defenses against this are methodological: weighting toward transacted rather than listed prices, capping any one source’s contribution, trimming or capping outliers, and averaging over a window rather than a single print.”
- Redr Labs: “Index methodologies should state how they treat zero-variance pools, pools with too few observations, and provider-administered posted rates, and publish those rules in advance”
- Silicon Data: “executable listings should be admitted only where availability and rates are verified by the benchmark administrator or other third-party entities and should be subject to per-provider weight caps.”
- Simeon Bochev: “I recommend the Commission treat provider-administered posted rates as unsuitable as primary inputs to a settlement reference, and set a minimum proportion of index weight that must derive from executed, arm's-length, reported transactions.”
- Solutions by Jewel: “Each provider whose posted rates enter the series keeps a timestamped log of rate changes, producible to the DCM”
- Squaretower: “market-wide firm quotes from sellers may be a better indication of the market value of the underlying compute, and could support a more representative index if included and distinguished appropriately.”
- Submitter listed as "Nico": “Before a DCM lists such a contract, the Commission should require the submission to state (i) which posted series, (ii) who administers it, (iii) whether transacted volume is captured, and (iv) what independent verification of that data exists (Q1(d)).”
- Vandermolen Power Limited: “Context only unless the administrator demonstrates that rates are actionable and representative.”
- Vast.ai: “Quotes can inform a benchmark if the methodology clearly identifies what they measure and supports the benchmark’s reliability.”
- Secondary position Balthasar Issa: “I would also ensure that quotes during observation windows are genuinely executable.”
- Secondary position Stark Executive Advisory: “Weighted Blended Model: A methodology that incorporates both continuous transaction averages from settled contracts and a clear, explicit weighting of unencumbered capacity rate cards.”
Executed transactions only (13)
- Alec Snelling: “build settlement indexes from verified, actual transactions, not providers' self-reported or posted rates”
- Animica Labs: “Index inputs should be drawn from transactions whose price, timestamp, and counterparty identifiers can be confirmed against a record the reporting party does not control”
- Balthasar Issa: “I would argue that credible, transaction-referenced, audited price- reporting infrastructure needs to be established first.”
- Compute Capital: “Evaluate DCM filings against whether the settlement series is a transaction print with governance the Commission can inspect, not a posted offer the supplier administers.”
- Hivemind Capital Partners: “Settlement indices must be derived exclusively from programmatic, auditable API billing records and cleared bilateral trade prints, explicitly excluding non-binding, advertised list prices or self-reported broker quotes.”
- Hydra Host: “Hydra suggests that an index should be built only from prices in completed rental transactions, as opposed to posted or “list” prices.”
- Kottos AI: “If there are too few eligible transactions, the reference price must be marked unavailable and the contract must specify how settlement proceeds in that situation.”
- OneChronos: “a reliable cash-settlement benchmark requires not only transaction-based inputs, but also sufficient market depth and breadth of participation so that a few trades or posted-rate changes cannot move the settlement price.”
- Ornn AI: “Benchmark administrators should use a transaction-based methodology and clearly define the economic unit being measured and the transactions eligible for inclusion in a benchmark.”
- Stark Executive Advisory: “Settlement mechanisms are more reliable when anchored to continuous transaction averaging of cleared, cash-settled capital flows.”
- Steven Skinner: “Posted or list rates should carry no weight in settlement.”
- TCX: “Require DCMs seeking to list compute derivatives to demonstrate that settlement indices rely on verified, arm's-length execution data rather than unverified supplier rate cards or self-reported surveys.”
- Wei Wang: “Executed prices only. The index should weight only consummated transactions, not indicative listings.”
Posted rates can feed the price (3)
- Douro Labs: “Where the index takes inputs from many providers and from purchasers and brokers, one provider’s posted rate is one input among many, and a median ignores it when it departs from the field.”
- DRW Holdings: “In compute, the conventional distinction between quotes and trades is less meaningful than in other commodities because posted rates are directly executable through API calls, while reported transaction data remains comparatively sparse.”
- Nodal Exchange: “a well-constructed index will not give disproportionate weight to quoted prices; instead, appropriately aggregating and balancing quoted prices, prices from market-based transactions, and prices from independent bilateral transactions”
- Secondary position GPU Economy: “It may suit a contract written on posted on-demand rates, such as a hedge of list-price exposure; it does not stand in for transaction prices.”
- → 57 of 92 letters took a position
Commenters and excerpts
Each exchange or administrator sets and discloses its own (37)
- ACU Markets: “we would encourage the Commission to set structural minima that the administrator must disclose and enforce fail-closed, rather than fixed numbers in guidance that will date quickly”
- ACU Markets: “The index applies a published contributor concentration control in each determination and publishes the largest contributor’s share daily.”
- Alec Snelling: “Caps on any one contributor's share of the index, rules against steering capacity or withholding trades during the observation window, and published methodology and governance would all help”
- Andrew Huff: “A DCM could set thresholds on them, such as a minimum contributor count and a maximum single-contributor share”
- Babbage: “Publish the aggregation formula, weights and caps, unit conversions, quality adjustments, outlier treatment, minimum-data thresholds and fallback order.”
- CCE: “We recommend product-specific thresholds for transaction volume, frequency and contributor diversity, supported by empirical testing and public data-quality indicators.”
- Compute Desk: “An index design that lacks concentration caps, actionability checks, and contributor obligations may be unable to comply with Core Principle 3.”
- Dartmouth Advisory Partners: “We would require each administrator to state its own minimum volume and contributor floor, publish both values for each fixing, and switch to the fallback when either is breached.”
- Douro Labs: “We recommend instead disclosure-based robustness metrics, set by the DCM in its contract rules and monitored continuously by the index administrator”
- Dragonfly: “Do not mandate a uniform minimum number of contributors, source-weight caps, or other fixed quantitative tests for all compute indexes. Let DCMs demonstrate that their particular reference is sufficiently reliable”
- DRW Holdings: “The index should draw from a broad panel of independent venues and capacity providers, and its methodology should include controls that prevent any single contributor from exerting outsized influence on a determination.”
- Federico M. Bandi and Yinan Su: “Capping very large contributor weights may therefore be useful, moving the weighting toward a more equal allocation.”
- Fiftyrivers: “As illustrative starting points, not empirical findings: at least five unaffiliated contributors and no group above one quarter of weighted inputs, calibrated and disclosed by the DCM.”
- Fullview: “we recommend contract-specific evidence rather than a universal numerical rule.”
- Futures Industry Association: “address contributor concentration and susceptibility to short-term rate adjustments.”
- Gemini: “Gemini lacks the data to propose specific volume, frequency, or concentration thresholds and cautions against fixing them in advance.”
- GPU Economy: “Use a median or a trimmed estimator, require a minimum number of independent providers, and publish the largest provider's share and the concentration index with each value.”
- Houman Shadab: “They belong to a DCM’s contract design and an administrator’s methodology, supported by cash-market evidence and subject to Commission review”
- Hydra Host: “no contributor’s weight in the index should exceed a stated limit; each index should require a minimum number of independent contributors”
- Hyperbolic Labs: “No single contributor’s transactions should exceed a published share of the weight in any window.”
- ICX Global: “Benchmarks should impose contributor caps, minimum coverage requirements for providers and observations, related-party controls and heightened surveillance during the settlement period.”
- Kalshi: “In order to further reduce the risk of manipulation, the index should require a minimum number of contributors and caps on concentration as well as a time- or volume-weighted average over a window, as opposed to a single snapshot.”
- Kottos AI: “The DCM discloses market coverage and contributor concentration, justifies minimum transaction and independent-contributor thresholds”
- Liquid Compute (PMEX Markets): “The Commission could instead require each methodology to establish and disclose minim um transaction, notional, and contributor-diversity thresholds appropriate to the relevant bucket”
- Mercatus AI: “A DCM should justify its concentration limits and data-sufficiency thresholds for the specific market, and test the effect of removing the largest contributor.”
- NativX: “They are NativX's benchmark thresholds, not CFTC-prescribed minima; a listing exchange must assess whether they are sufficient for the proposed contract and expected open interest.”
- Ömer Demirel: “The level needs calibration on real data; 20 to 25% is an illustrative starting point.”
- Ornn AI: “For example, data-sufficiency thresholds, concentration measures, sensitivity testing, or contributor limits may be appropriate where supported by observed trading patterns, but should not be prescribed without regard to the structure of the underlying market.”
- Polymarket US: “Rather than fix these thresholds by rule, the Commission should expect a DCM to show in its filing which thresholds its chosen index applies and why they are adequate for that market.”
- PTG: “Contributor-concentration limits, outlier procedures, and controls that prevent any one provider from exercising outsized influence over settlement.”
- Solutions by Jewel: “the DCM sets and publishes a maximum single-contributor weight and a minimum contributor count per determination”
- Squaretower: “An index should disclose the observation count and contributor concentration behind each value and the largest change a single contributor could cause, and should withhold publication when inputs fall below clearly defined criteria.”
- Steven Skinner: “The reference should use executed transactions with delivery records, exclude trades between affiliates, publish contributor counts and concentration for each window, and cap any single provider's share.”
- Vandermolen Power Limited: “DCMs should therefore require conflict controls, affiliate flags, contributor caps, minimum observation windows, transaction-quality screens, publication of restatement policies and surveillance of capacity movements around settlement windows.”
- Vast.ai: “Benchmark review should assess that economic diversity rather than impose uniform caps that could exclude representative data.”
- VeloxVFX: “Thresholds should be derived from empirical evidence and the responsible DCM's regulatory analysis, not from an architecture assumption.”
- Wei Wang: “the contract should define what happens when those metrics fall below stated floors”
Publish counts and concentration, no fixed floor (13)
- 1791 Technologies: “Useful disclosures may include the share of index weight by input class, contributor concentration, observation count, transaction-value coverage where available, and sensitivity to removal of the largest contributors.”
- Crypto Council for Innovation: “disclosures and controls should provide sufficient information about contributor composition, submission quality, validation practices, and material changes in data coverage”
- De Silva Law Offices: “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.”
- DNA Compute: “We do not propose numeric thresholds. We propose the metric and the publication, so that thresholds can be set from evidence.”
- Evelyse Carvalho Ribas: “disclosure, with each published value, of the number of contributors and the concentration of contribution.”
- Hyperliquid Policy Center: “index robustness be measured by disclosed metrics rather than fixed transaction-count thresholds.”
- Lee Dryburgh: “The minimum the Commission can require of a settlement reference is that its contributor count, its input mix, its concentration and its standardization model be published.”
- Lida Liberopoulou: “the number of providers and transactions behind each observation window;”
- Ola Ologunde: “Require per-fixing observation counts and contributor concentration, including whether contributors are large compute providers or users.”
- Paradigm and Andromeda: “An index that discloses contributor concentration, separates transactions from posted rates, and is independently administered is far more difficult for any single provider to move”
- Steven Singleton: “condition self-certification eligibility on public disclosure of the number and market share of rate-contributing providers feeding any settlement index”
- Submitter listed as "Grok": “Each source should disclose the share of executed transactions versus posted rates, contributor count, and the largest contributor's weight in every published value.”
- Trevor Love: “Count affiliated entities as a single benchmark contributor and disclose contributor concentration by unaffiliated contributor or contributor group.”
Set a fixed minimum or cap (7)
- Abaxx Technologies: “maximum share of index input attributable to any single contributor (we suggest a 20–30% cap, calibrated to compute's actual concentration)”
- Animica Labs: “Counterparty-diversity floors, not volume floors. Given the wash-trading exposure in Part 5, thresholds should be expressed in distinct economically independent counterparties within the window.”
- Architect Financial Technologies: “Architect recommends the following minimum listing expectations. A DCM may adopt stricter thresholds.”
- Balthasar Issa: “I would encourage the Commission to require both minimum contributor counts and a cap on a single contributor's share of index volume.”
- Hivemind Capital Partners: “Constituent Concentration Limits: Enforcing strict index diversification rules ensuring that no single cloud provider or data center operator accounts for more than 20% of the total index weight during any settlement determination window.”
- Silicon Data: “we suggest that a settlement index be calculated each business day from no fewer than 5 independent providers, with no single provider accounting for more than 20% of index weight.”
- Simeon Bochev: “Distinct unaffiliated contributors: no fewer than ten in any observation window, with the index suspended to fallback below five.”
- → 64 of 92 letters took a position
- → 23 state more than one position
- → Secondary positions are marked in the dropdown below
Commenters and excerpts
One tightly defined GPU-hour (22)
- 1791 Technologies: “Define the underlier as a bounded economic identity: contract terms should specify the dimensions that materially determine equivalence rather than treating "compute" as a unitary commodity.”
- Alec Snelling: “Agreed definitions for chip type, region, and term should come first, with a narrow set of contracts expanding as data and liquidity grow”
- Andreessen Horowitz (a16z): “Benchmarks should precisely isolate standardized economic units, such as rental costs per GPU-hour for specific chip families, rather than blending dissimilar hardware tiers into an unrepresentative composite.”
- Babbage: “Disclose this limitation and narrow or defer settlement use, rather than pool incompatible services.”
- Balthasar Issa: “a compute derivatives contract would need to show: (i) a precisely and narrowly defined hardware generation, configuration, and location as the reference underlier”
- CME Group: “The underlying indices price the hourly cost of renting standardized compute capacity of NVIDIA H100 and B200 GPUs.”
- Compute Capital: “A named settlement underlier, SKU by SKU, that is a transaction print”
- Dartmouth Advisory Partners: “An hour of compute differs by accelerator generation, memory, interconnect, cluster size, commitment term and region. A contract should name these.”
- DRW Holdings: “it requires a reference rate to be a faithful, reproducible measure of a defined, well-specified pricing point; i.e., the on-demand hourly rental rate for a specified GPU model.”
- GPU Economy: “Different products at different prices; a contract must name one”
- Hedge Quantitative Research: “Any settlement index needs a stated quality basis.”
- Hydra Host: “Hydra’s primary recommendation is that an index publish a separate series for each grade and that a series serve as a settlement reference only once it meets minimum depth and contributor thresholds.”
- ICX Global: “The first generation of compute derivatives should reference observable, precisely specified products, defined by a named GPU, geography and purchase mode (on-demand, spot, reserved or committed).”
- Ömer Demirel: “Fix the grade specification of the contract: chip, memory, interconnect, minimum uptime, region class and contract length band.”
- OneChronos: “Compute has no equivalent of a battery: a data center can store electricity, but an unused accelerator-hour cannot be banked against a future shortage, making contracts defined around narrower, physically fungible capacity the natural starting point.”
- Ornn AI: “A compute contract can specify the GPU type, geographic region, delivery period, configuration requirements, and other characteristics that materially affect value, with separate contracts or reference prices where particular characteristics produce persistent, economically significant differences in value.”
- PTG: “It needs a clear definition, including the chip model, the rental type, the kind of provider, the location, and how it settles.”
- Simeon Bochev: “Fixed USD per GPU-hour for the whole-rack unit, quoted in $0.001 increments.”
- Squaretower: “Indices should, where possible, specify precisely which configurations and services qualify as eligible data points and how variables that are otherwise difficult to measure are normalized or inferred.”
- Steven Skinner: “The Commission should expect a compute settlement reference to count only GPU-hours that are (a) identified by hardware class and configuration”
- Trevor Love: “Define the underlier by operational service tier, including availability, scheduling, and interruption rights.”
- Vandermolen Power Limited: “Launch with a narrow contract scope: one hardware or workload profile, limited regions, fixed expiries and transparent settlement windows.”
- Secondary position ACU Markets: “In the interim, contracts continue to list on single, tightly specified grades.”
- Secondary position Animica Labs: “a compute contract's terms should state the delivered unit (accelerator-hour of a named class, or tokens of a named model at a named context length)”
- Secondary position Federico M. Bandi and Yinan Su: “Contracts can reference specific configurations, with contracts on new configurations introduced as they become available.”
- Secondary position Kalshi: “Such a contract could identify, for example, a particular GPU class such as H100 or B200, a defined quantity of computing capacity over a stated period, or regional geographic and service-quality-based contracts.”
- Secondary position Liquid Compute (PMEX Markets): “Hardware class and configuration , specified to the SKU level, including memory configuration and any relevant firmware or driver constraints.”
A reference grade plus priced differentials (20)
- Abaxx Technologies: “a claim ticket specifies the deliverable's class (e.g., a named accelerator generation or a performance-equivalence class), region, duration, availability grade, and network terms.”
- Andrew Huff: “Agricultural futures solved a version of this problem with published grade differentials; compute needs the same discipline.”
- Architect Financial Technologies: “Compute units can become fungible through published quality adjustments, as grain and crude oil contracts use grade differentials.”
- CCE: “We recommend defining fungibility within compute grades that specify hardware, allocation, connectivity, location and service requirements. Quality adjustments should use disclosed, reproducible methods.”
- Compute Desk: “Compute Desk isolates the pure, standardized par-equivalent price of bare metal compute, establishing fungibility across heterogeneous physical infrastructure.”
- Federico M. Bandi and Yinan Su: “A clearly specified reference, or “par”, service may define the”
- Fiftyrivers: “Specify a “par” grade in the contract terms (accelerator, interconnect/cluster, region/jurisdiction, commitment term, SLA) with published differentials.”
- Hivemind Capital Partners: “DCMs can standardize around rolling, annual “on-the-run” flagship GPU-hour benchmarks, pricing variations in interconnect, provider tier, and region via transparent basis differentials.”
- Hyperbolic Labs: “a benchmark grade defined by chip, node configuration and interconnect, region, and uptime commitment, with a published schedule of adjustment factors for deliverable substitutes”
- Lee Dryburgh: “The commodity is the usable accelerator-hour, graded by chip and by the capacity behind the chip, and not storable.”
- Michael Ravnitzky: “DCMs should be precluded from listing contracts until a benchmark that adjusts for region and hardware generation is established”
- Ola Ologunde: “Require any index used to settle a listed compute contract to publish its specification and adjustment rules.”
- OrchestraOS: “An index could restrict inputs to a narrow performance class or use a reference grade with published differentials.”
- PSYCHE: “A par quality for compute would be stated in the dimensions Exhibit 2 shows two filed exhibits stating and eight withholding.”
- Silicon Data: “Applying a deterministic, rules-based econometric basis adjustment to normalize every admitted record to a unified baseline configuration prior to aggregation”
- Simeon Bochev: “That is the correct model for different silicon families and different facility classes: not one universal compute unit, but defined grades with observable spreads.”
- Stark Executive Advisory: “Mature commodity markets achieve fungibility not through complete homogeneity, but by establishing a Par Grade and utilizing Standardized Basis Tables for structural deviations”
- Taylor Holland: “However, standardized contract specifications and quality adjustments could establish fungibility within defined product grades, as occurs in energy and agricultural markets.”
- Voltry: “That lets a contract specify "reference condition or better" the way a grain contract specifies "No. 2 or better."”
- Voltry: “Appendix C to Part 38 already supplies the form: a par quality, with differentials for non-par qualities that reflect what the cash market observes.”
- Secondary position ACU Markets: “Require the underlier to be defined either on a normalised unit of delivered work or on a tightly specified single-grade, single-package basis with a published grade-adjustment table and a governed process for adding and retiring grades.”
- Secondary position ACU Markets: “Require any contract that references more than one accelerator grade to publish, as a term of the contract, an eligible grade list and the basis on which each differential is observed and settled”
- Secondary position Andreessen Horowitz (a16z): “DCMs listing compute derivatives can apply the same discipline by defining standardized contract specifications, specifying precise GPU models, minimum network throughput, cluster scale, and temporal delivery blocks, while establishing separate contract symbols or basis spreads for distinct architectural generations.”
- Secondary position Hydra Host: “it should be anchored to a stated reference grade (e.g., AAA), which Hydra suggests should be the highest grade for which the series meets those thresholds, with differentials for other grades”
- Secondary position Jump Trading Group: “The Commission should allow DCMs to develop grades and support trading in the price differentials between them.”
- Secondary position Paradigm and Andromeda: “The market may also benefit from multiple “grades” or tiers of deliverable compute capacity, based on the quality or demand for the specifications of a given cluster.”
- Secondary position The Digital Chamber: “Compute grades should likewise be able to evolve as hardware generations turn over, with separate contracts, premiums, discounts, or basis trading between grades”
- Secondary position Trevor Love: “Separate firm, scheduled, and interruptible transactions in the benchmark methodology, or justify and disclose any grade adjustments used to combine them.”
- Secondary position Vandermolen Power Limited: “fungibility can be created within a defined grade or delivery profile, with transparent differentials or conversion rules between grades.”
Other: leave it to exchanges and the market (12)
- Arkhai (CoopHive): “The choice depends on the intended use, not granularity alone. Declare the target, sample, weights and adjustment tests so users can assess that choice.”
- DNA Compute: “The same criteria apply whatever the unit of account (a GPU-hour of a stated grade, an energy-normalized unit or a quantity of inference measured in tokens)”
- Douro Labs: “What the Commission can do is expect a DCM to disclose which unit its settlement index defines”
- Dragonfly: “Permit DCMs to develop commercially useful contract units, including GPU-hours with hardware and service characteristics necessary for economic comparability, without prescribing a single Commission-wide compute specification.”
- Gemini: “The Commission should encourage, but not itself select, standard definitions for underliers such as hardware classes, units, and service types”
- Houman Shadab: “for compute those may include hardware class, performance and latency, location, and the applicable model or software specification, drawn from cash-market practice rather than prescribed generically”
- Hyperliquid Policy Center: “The Commission should encourage the introduction of derivative products referencing both classes of compute rather than treat one as the sole template for the market.”
- Jump Trading Group: “the Commission should accommodate multiple underliers rather than seek to establish a single canonical compute product.”
- Kalshi: “The Commission need not define one ubiquitous standard compute unit or product before permitting this important market to develop in the United States”
- Liquid Compute (PMEX Markets): “We do not suggest th at the Commission should specify this by rule; contract design is properly the province of registered entities under Core Principle 3.”
- Paradigm and Andromeda: “The Commission should therefore not seek to set quality standards or otherwise limit listed compute derivatives based on quality, provided that the terms of such listed derivatives are fully disclosed.”
- The Digital Chamber: “the market should be permitted to determine which underliers, contract grades, settlement structures, and reference prices prove most useful.”
- Secondary position Andreessen Horowitz (a16z): “the Commission need not prescribe which chip generations or hardware configurations may underlie a contract.”
- Secondary position Fullview: “The framework should allow markets to develop around each where the relevant conditions can be demonstrated.”
- Secondary position InfraSight Software: “We do not ask the Commission to prescribe a unit or a methodology.”
A unit of performance or delivered work (7)
- ACU Markets: “The Commission should require either a normalised unit of delivered work or a tightly specified single-grade delivery basis with a published grade-adjustment table.”
- ACU Markets: “Compute should be priced on a unit of delivered work, defined by an independent, audited and published measurement standard, and not on the rental hour of a physical device.”
- Anonymous: “there needs to be a clear designation of what and how a unit in this commodity class is measured and standardized across hardware architectures”
- Fullview: “The framework should accommodate outcome-based commercial arrangements where the relevant market and regulatory requirements can be satisfied.”
- InfraSight Software: “Define the unit by delivered work, not hardware. A unit should represent a defined quantity of useful work rather than time on a named chip.”
- NativX: “Recognize an energy-normalized unit of qualifying compute delivery as a viable denomination for a broad compute benchmark.”
- TCX: “Encourage standardized performance-grading parameters based on realized workload throughputs specifically actual execution yield (Tokens/Sec per watt)”
- Secondary position ICX Global: “Before a normalized index is used for settlement, its performance factors should be independently measured, auditable, workload-specific, versioned and subject to governance.”
- Secondary position Michael Ravnitzky: “Prices should be weighted based on actual performance-per-watt efficiency relative to current market standards, not just raw server counts.”
- Secondary position OneChronos: “A mature index could also reference the measured output of specific accelerators, adjusted for location, data center tier, and workload, which is harder to manipulate than a posted rate.”
- Secondary position Simeon Bochev: “A contract denominated in a vendor's part number is denominated in a nameplate, not in work delivered.”
- Secondary position Stark Executive Advisory: “Bare-metal telemetry introduces an auditable performance standard, shifting the index measurement from nominal uptime to actual realized execution yield.”
Other: tokens or inference output (3)
- Animica Labs: “a compute contract's terms should state the delivered unit (accelerator-hour of a named class, or tokens of a named model at a named context length)”
- INFEREXUS: “the “commodity” underlying these contracts — inference capacity at a given model version — is issued and priced unilaterally by a small number of frontier AI labs”
- Kottos AI: “I therefore propose starting with one clearly specified open-weight inference grade and expanding as transaction volume increases.”
- Secondary position Dragonfly: “Treat inference throughput as a separate potential contract class where model choice, version, and measurement protocol can be standardized reliably.”
- Secondary position Hivemind Capital Partners: “the Commission’s immediate focus should center on Datacenter Fabric Accelerators (GPU/TPU hours) and Cloud Fabric (Inference Tokens)”
- Secondary position Hyperliquid Policy Center: “Both forms of compute are suitable underliers for derivatives contracts, and they serve different participants.”
- Secondary position Jump Trading Group: “Relevant underliers may include accelerator capacity by hardware grade, inference or token throughput, memory- or bandwidth-based measures, and capacity available in specified regions.”
- Secondary position The Digital Chamber: “Compute Can Also Be Measured at the Service Layer Through Inference Tokens”
- → 33 of 92 letters took a position
- → 2 state more than one position
- → Secondary positions are marked in the dropdown below
Commenters and excerpts
Conflicts allowed, disclosed and managed (18)
- 1791 Technologies: “how conflicts of interest are managed, and what review or challenge process applies.”
- ACU Markets: “We would encourage the Commission to require, as a condition of settlement use, disclosure of the administrator's ownership, its commercial relationships with contributors, its conflicts register and its independence arrangements”
- Andrew Huff: “Separate the people and systems that contribute data from those that administer the index.”
- Anonymous: “clear separation of index providers business functions and the interests of its investors, of whom many are trading firms”
- Arkhai (CoopHive): “Venues, administrators and clearing entities also have revenue and continuity incentives. These warrant conflict review, not an assumption of misconduct.”
- Babbage: “Use documented responsibilities, recusals and independent oversight to separate data selection, calculation, corrections and methodology decisions from conflicted sales, trading and investor interests.”
- Dartmouth Advisory Partners: “Any tie between the administrator and a contributor or venue”
- De Silva Law Offices: “The administrator should disclose any ownership by, or affiliation with, a firm that trades compute derivatives or a venue whose transactions enter the index.”
- Dragonfly: “Address credible manipulation or material conflicts through the DCM's existing market-integrity obligations; do not disqualify a supplier's price solely because the supplier is also a legitimate derivatives hedger.”
- Fullview: “The Commission should consider dependencies shared by apparently separate suppliers and conflicts spanning providers, venues, contributors and administrators.”
- ICX Global: “The administrator should disclose its ownership and commercial relationships with the listing venue, its clearing house and any provider or venue whose prices are used in the calculation.”
- Mercatus AI: “Governance should make those interests visible and limit their influence”
- Ömer Demirel: “Index providers should also follow the IOSCO Principles for Financial Benchmarks, which cover governance, conflicts of interest, methodology and audit.”
- Polymarket US: “the administrator should keep its index business separate from any trading interest and disclose its ownership”
- Simeon Bochev: “An oversight function independent of the administrator's commercial operations, with contributors bound by a code of conduct, a conflicts policy reaching affiliates”
- Solutions by Jewel: “These affiliations are ordinary features of an early market, and they define the conflicts a listing review should identify and manage.”
- Vast.ai: “The DCM should require the administrator to disclose material ownership and commercial affiliations, policies governing index-linked derivatives positions held by the administrator, its affiliates and staff, and controls separating calculation from trading decisions.”
- Wei Wang: “The reference price for a regulated product should not be computed by an entity with an undisclosed interest in where that price lands.”
- Secondary position DNA Compute: “That is also the remedy the IOSCO Principles contemplate for an administrator whose ownership creates conflicts: disclosure and independent oversight with balanced representation, not exclusion.”
- Secondary position Vandermolen Power Limited: “Require benchmark administrators and DCMs to disclose input-data coverage, contributor concentration, methodology, conflicts, exclusions, restatements and surveillance access.”
Administrator may not trade the contract or hold compute (15)
- Architect Financial Technologies: “An index administrator and its affiliates should not trade derivatives that settle to the administrator’s index.”
- Capital Markets & Technology Partners: “The independence of the reporting authority from the entities that it measures will be key, going forward.”
- CCE: “We support commercial hedging with benchmark governance independent of supplier pricing and trading interests.”
- Compute Desk: “An affiliate of a compute derivatives market-maker acting as the index provider underlying the compute derivatives in question may be faced with a conflict of interest in the form of an incentive to manipulate the market.”
- DNA Compute: “an administrator should neither trade the referencing contract nor hold balance-sheet exposure to capacity whose price it publishes”
- DRW Holdings: “An administrator should not take proprietary positions in the derivatives that reference its index, nor in the underlying commodity beyond its own internal needs.”
- Evelyse Carvalho Ribas: “an administrator that is independent of the contributors and does not itself supply compute capacity”
- Fiftyrivers: “Index administrators should be independent, in ownership and governance, of hyperscalers and other major capacity suppliers.”
- Hyperbolic Labs: “the administrator should be independent of any contributor whose weight is material.”
- Liquid Compute (PMEX Markets): “In PMEX’s view, an administrator of a se ttlement reference generally should not take positions in contracts that reference it.”
- Michael Ravnitzky: “A benchmark must be governed by an entity that is structurally independent from the firms whose behavior it measures.”
- PTG: “an administrator should not trade the derivative or hold material physical exposure to the compute capacity its index measures.”
- Silicon Data: “a benchmark administrator must be strictly prohibited from engaging in proprietary trading or holding physical risk exposure in compute assets.”
- Submitter listed as "Grok": “Settle only when two independently administered price sources, drawn from separate input populations and run by parties that do not supply the capacity being priced, agree within a published tolerance in the observation window.”
- Vandermolen Power Limited: “Separate capacity registration, benchmarking, certification, registry operation, trading and clearing so that no producer can unilaterally create, price and validate its own deliverable supply.”
- → 46 of 92 letters took a position
- → 2 state more than one position
- → Secondary positions are marked in the dropdown below
Commenters and excerpts
Required as a condition (31)
- ACU Markets: “Require deterministic replayability: an independent party, given the pinned inputs, must be able to regenerate any published settlement value bit-for-bit without access to the administrator's systems.”
- ACU Markets: “Treat alignment with the IOSCO Principles for Financial Benchmarks, confirmed by published independent assurance, as a condition of a compute settlement reference.”
- Andrew Huff: “Require periodic independent audit of the methodology and of a sample of inputs against source records.”
- Animica Labs: “Index inputs should be drawn from transactions whose price, timestamp, and counterparty identifiers can be confirmed against a record the reporting party does not control — a public ledger, or an exchange record subject to Commission access.”
- Architect Financial Technologies: “Architect recommends that each administrator of a compute settlement index obtain an independent external assurance review of its compliance with the IOSCO Principles.”
- Arkhai (CoopHive): “In the recommendations, independent verifiers audit these inputs and processes, while independent reviewers assess incentive and conflict risks.”
- Babbage: “Retain traceable evidence and independently verify material inputs.”
- Balthasar Issa: “I would argue that credible, transaction-referenced, audited price- reporting infrastructure needs to be established first.”
- CCE: “Settlement inputs should be independently verifiable.”
- De Silva Law Offices: “The administrator should have its methodology and data controls reviewed independently at regular intervals.”
- DNA Compute: “independent assurance against the IOSCO Principles for Financial Benchmarks, obtained within a stated period after first listing and renewed at regular intervals”
- DRW Holdings: “The index administrator’s adherence to its published methodology and governance should be subject to periodic independent external assurance, with the statement of adherence made public.”
- Fiftyrivers: “Require independent external audit, a full audit trail, and cooperation with regulators”
- Gemini: “Auditability: periodic independent assurance that the methodology is applied as published”
- Houman Shadab: “independently verifiable evidence of the provenance, integrity, and eligibility of each input”
- Hydra Host: “price reports should be audited against the underlying transaction and delivery records”
- ICX Global: “the Commission should require an auditor or independent control function to reproduce historical values using the underlying eligible observations and approved methodology.”
- Kottos AI: “Before listing, the DCM should demonstrate which service attributes can be independently verified and how remaining verification gaps are handled.”
- Michael Ravnitzky: “Requiring providers to report audited, region-specific capacity data ensures that listed contracts reflect true underlying supply rather than artificial operational shifts.”
- Ömer Demirel: “Each period, an independent auditor should pick transactions at random after the reporting window closes.”
- Silicon Data: “the administrator’s adherence should be subject to formal, periodic independent third-party assurance.”
- Simeon Bochev: “The Commission should require, as a condition of a series being used as a settlement reference, an annual independent assurance report”
- Simeon Bochev: “Require attributable records, protected access to the underlying evidence, independent review and governed publication.”
- Solutions by Jewel: “SBJ recommends that the Commission request, for each series proposed as a settlement reference, the scope, provider and findings of any independent review, and treat a periodic external assurance engagement as an element of the Appendix C demonstration.”
- Stark Executive Advisory: “Independent auditors and regulators must possess the structural capability to reproduce the results completely”
- Steven Skinner: “so the exchange or an independent verifier can confirm it later without relying on either party's word. Transactions without such a record should not enter the settlement calculation.”
- Submitter listed as "Grok": “A transaction should enter the settlement input only if the capacity passed independent checks of hardware identity, benchmarked performance, and delivered uptime, and hosts that fail should be removed.”
- TCX: “Hardware-level telemetry addresses this vulnerability by requiring that any transaction or capacity contribution entering a settlement index be validated by cryptographic hardware attestation proving unencumbered operational capability and actual workload execution.”
- Vandermolen Power Limited: “Commercial terms may remain confidential, while the Commission, the relevant DCM and an authorized independent auditor receive the fields needed to test representativeness, concentration and delivery.”
- Voltry: “as measured from the card and verified by an independent reader”
- Wei Wang: “I respectfully ask the Commission to make "recalculable, auditable, with published fallback settlement rules" an explicit listing precondition for compute futures contracts.”
- Secondary position InfraSight Software: “independent verification of delivered performance, performed by a party other than the contributor;”
Encouraged (13)
- Crypto Council for Innovation: “Although commercially sensitive transaction information may warrant confidentiality protections, exchanges and benchmark administrators should be able to demonstrate that inputs are independently verifiable, sufficiently representative of the defined compute exposure, and accessible to appropriate regulatory oversight.”
- Dartmouth Advisory Partners: “An index whose values can be rebuilt from retained inputs can be audited.”
- FastGPU: “It lets market participants, journalists and the Commission's staff check any index against what providers were posting on the same day”
- Fullview: “Transaction reporting, contractual information sharing and independent review provide a conventional starting point.”
- Futures Industry Association: “DCMs should seek to use index administrators for Compute benchmarks that adhere to governance, methodology, conflict-of-interest, and audit standards comparable to the PRA Principles and Financial Benchmark Principles.”
- InfraSight Software: “consider whether the contributed capacity's delivered performance is independently verified, not only whether its price is observable.”
- Kalshi: “Contributors should also be subject to a code of conduct, recordkeeping requirements, and administrator audits of their submissions.”
- Lee Dryburgh: “the share of index weight held by the largest one, three and five contributors; and the standardization model, or an audit of it.”
- Liquid Compute (PMEX Markets): “the Commission should be slow to treat publication as evidence of verification”
- NativX: “We recommend independent checks of sampled sale evidence, affiliate treatment, exclusions, source completeness and publication against the governed specification.”
- OrchestraOS: “a performance reference measured on a representative sample of delivered instances under published test conditions, with results subject to independent verification”
- Vast.ai: “The DCM should retain its benchmark-selection and review records and obtain contractual assurances that the administrator retains the inputs and calculations it uses, in a form sufficient to reproduce published values.”
- Voltry: “A per-card condition grade supplies the hardware component of each of those proposals, from a reader with no interest in the price.”
- Secondary position Hyperbolic Labs: “Hyperbolic would be willing to voluntarily submit its index contributions for independent verification under a benchmark administrator’s audit program”
Not a listing condition (2)
- Dragonfly: “Do not make a mandatory independent audit or marketplace-wide reconciliation a condition of eligibility.”
- Hyperbolic Labs: “the existing protections described above in the context of natural gas are sufficient.”
- → 61 of 92 letters took a position
- → 2 state more than one position
- → Secondary positions are marked in the dropdown below
Commenters and excerpts
Detail to the exchange and CFTC; summaries public (51)
- 1791 Technologies: “The regulatory standard should be reconstructability, not public disclosure of every proprietary input.”
- ACU Markets: “Confidential inputs are acceptable where the administrator is bound to retain them, audit them, and produce them to the listing DCM and to the Commission on request.”
- Alec Snelling: “Fourth, require information-sharing arrangements between each listing exchange and every compute venue and provider whose data feeds the settlement price, along with surveillance that can spot abnormal moves (Question 2(d)).”
- Andreessen Horowitz (a16z): “share non-public transactional granularity with DCM surveillance teams, which in turn may share such information with the Commission under appropriate confidentiality protections, without requiring proprietary trade secrets to be published”
- Andrew Huff: “I would not permit a contract to settle to a price computed from data that the DCM and the Commission cannot observe or verify on request.”
- Architect Financial Technologies: “the Commission must have assured access to the data, through the DCM, under binding contributor and information-sharing agreements.”
- Arkhai (CoopHive): “Any refinement should distinguish public benchmark information from confidential evidence accessible to authorized supervisors and respond to a demonstrated shortcoming.”
- Babbage: “the administrator should hold enforceable rights to audit, retain and produce the necessary records to the listing DCM and the Commission.”
- CCE: “We support confidential access to transaction records for the Commission and exchange surveillance, alongside public benchmark methodologies and aggregate prices.”
- Compute Capital: “unless the DCM can see, and the Commission can reconstruct, the bilateral book that is the true cash market.”
- Compute Desk: “Counterparties remain fully anonymized in public index publications to protect commercially sensitive competitive terms, while providing access to unfiltered regulatory visibility to the DCM, and indirectly the Commission.”
- Crypto Council for Innovation: “Although commercially sensitive transaction information may warrant confidentiality protections, exchanges and benchmark administrators should be able to demonstrate that inputs are independently verifiable, sufficiently representative of the defined compute exposure, and accessible to appropriate regulatory oversight.”
- Dartmouth Advisory Partners: “The exchange holds an information-sharing arrangement with the administrator, and through the administrator with each contributor, sufficient to examine the inputs behind any settlement value.”
- De Silva Law Offices: “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.”
- Douro Labs: “The submission is confidential as to counterparty and volume, the aggregate is public, and the input is attributable and auditable by the DCM and the Commission”
- Dragonfly: “Permit confidential contractual access to supporting bilateral-price records where needed to assess integrity; do not require public disclosure of proprietary contracts.”
- DRW Holdings: “make that data available to the listing DCM’s market regulation function and to the Commission on request”
- Fiftyrivers: “Inputs may be confidential to the public but must be accessible to the DCM and the Commission.”
- Fullview: “Confidential underlying transactions need not become public in full. The exchange and Commission should nevertheless have appropriate access to assess authenticity, influence and representativeness.”
- Futures Industry Association: “Consider whether the Commission itself should have a right of access to index input data for Compute benchmarks used to settle listed contracts”
- Gemini: “Requiring that every input be publicly disclosed would likely be counterproductive: contributors that value confidentiality would withdraw, and index quality would fall.”
- Hedge Quantitative Research: “A contract settling to such an index can still be listed if the index administrator's inputs are available to the exchange's surveillance under an information-sharing arrangement”
- Houman Shadab: “with source transaction records and supporting identity and affiliation data retained and promptly available to the DCM, and through it the Commission, on request”
- Hydra Host: “Hydra recommends that the listing exchange maintain its information-sharing arrangement with the index administrator”
- Hyperbolic Labs: “The listing DCM and the CFTC should have access to the transaction-level inputs under information sharing arrangements.”
- Hyperliquid Policy Center: “provided the DCM and Commission retain access to the underlying information needed for surveillance”
- ICX Global: “Individual transactions need not be public. The listing exchange, the Commission and an independent auditor should, however, have access to the underlying evidence and be able to inspect and validate it.”
- Kalshi: “support confidence in the robustness of the index even though the individual component data remains private.”
- Kottos AI: “Public reporting can use anonymized aggregates, with detailed access for benchmark administrators, designated contract market (DCM) and the Commission.”
- Liquid Compute (PMEX Markets): “Confidential transac tion data may still be usable if the index administrator and contributing venues or parties are subject to arrangements that make material input data available to the listing DCM and the Commi ssion”
- Mercatus AI: “Information-sharing arrangements should be enforceable, give timely access to contributor and counterparty details, and reach through any administrator to the underlying evidence.”
- Michael Ravnitzky: “A benchmark that incorporates regional energy constraints and requires verifiable, regulator-accessible data from power-dependent compute providers is essential to preventing these dynamics from manifesting in listed markets.”
- NativX: “A listing package should therefore establish enforceable access to underlying reporter evidence, including counterparty and affiliation information when needed for an investigation, through arrangements that protect confidentiality.”
- Ömer Demirel: “The DCM should have an information-sharing agreement with the index provider and basic information access to every reporter, so that coordinated reporters cannot each stay below a threshold.”
- OneChronos: “information-sharing arrangements with cash venues, price-reporting agencies, and index contributors, as Question 2(d) contemplates, providing the cash-price reference data needed to monitor convergence”
- Ornn AI: “The DCM and the Commission may appropriately receive more granular information on a non-public basis as necessary for oversight.”
- Polymarket US: “the Commission does not require that the settlement inputs for other commodity benchmarks, such as freight or assessed energy benchmarks, be individually observable, and it should not impose that requirement on compute.”
- PSYCHE: “Proprietary to the public is not withheld from the designated contract market or the Commission”
- PTG: “A clear obligation for the administrator to retain constituent -level data and make it available to the listing DCM and the Commission on request.”
- Silicon Data: “with the full model and parameters available to the DCM and the Commission on a confidential basis so that any published value can be independently reconstructed.”
- Simeon Bochev: “I would draw the line at observability by the Commission and the listing venue, not at public disclosure.”
- Simeon Bochev: “Customer workloads and commercially sensitive prices are protected; holders, obligated parties, funds agents, authorized verifiers, adjudicators and regulators have defined access within their remit.”
- Solutions by Jewel: “A compute reference price qualifies for settlement to the extent the DCM can produce to Commission staff, on request, the input-level record behind every determination in the settlement window”
- Stark Executive Advisory: “This allows regulators and clearinghouses to ingest and aggregate 100% of relevant transaction flows into a centralized data pool.”
- Steven Skinner: “Information-sharing arrangements with contributing providers should cover these records, not only price feeds.”
- The Digital Chamber: “Information - sharing with the benchmark administrator, together with large -trader reporting and DCM surveillance capable of identifying relevant compute -provider accounts, should provide a more workable framework for monitoring activity affecting the benchmark.”
- Trevor Love: “Bind each settlement value to a stable methodology version and retain the inputs and rules needed to reconstruct it.”
- Vandermolen Power Limited: “Use confidential regulatory reporting for bilateral transactions and capacity commitments where public disclosure would expose commercial terms, while ensuring the Commission and relevant DCM can verify the economic population behind a benchmark.”
- Vast.ai: “The administrator should be able to respond to DCM questions about settlement inputs, subject to appropriate confidentiality protections”
- VeloxVFX: “Before material reliance on an external reference source, a DCM should be able to demonstrate surveillance access and information availability sufficient for its applicable obligations, taking account of materiality, legal restrictions, confidentiality, and the character of the source.”
- Voltry: “Aggregated condition statistics, shared under an information-sharing agreement of the kind Appendix C paragraph (c)(3)(i) contemplates, give surveillance something to surveil.”
- Secondary position DNA Compute: “the identity of each contributor, pseudonymized in anything published, the number and classification of its observations, and the weights applied.”
Input-level detail published (10)
- Abaxx Technologies: “First, require the data that creates one to be publicly available.”
- Animica Labs: “settlement is recorded on a public ledger, so the transaction record is observable to the Commission, to a designated contract market, and to any member of the public, without an information-sharing agreement with the provider”
- DNA Compute: “A consumer of these prices who runs positions against them needs to be able to reproduce them; that is the whole of the point.”
- FastGPU: “publish every constituent observation with its timestamp, so that a change to a posted rate during an observation window is visible to everyone at the time it happens.”
- Federico M. Bandi and Yinan Su: “This includes which providers are surveyed, which parts of their rental business enter the index, and how the surveyed observations are used to measure prices in that segment.”
- GPU Economy: “Publish every input of every settlement value with its source, read time and value, so the contract market, the Commission and the public can recompute it.”
- Lida Liberopoulou: “enough information for an outside party to reproduce the settlement price.”
- Setara Financial: “A market intended to support institutional hedging requires independently governed methodology, transparent inputs, authenticated data, consistent normalization, documented calculation procedures, and sufficient lineage to allow participants to understand how an index was produced.”
- Taylor Holland: “Cash settled compute futures based on an index should disclose, in addition to index data sources, calculation methodology, and price history, the index’s limitations and reasons it may not fully proxy and converge to compute’s underlying economic cost.”
- Wei Wang: “Methodology and raw panel data must be public and independently recalculable.”
- Secondary position Hyperbolic Labs: “a DCM relying on an exclusively licensed compute index should be expected to demonstrate that the index’s level, inputs, and methodology are nonetheless available to the market.”
- → 48 of 92 letters took a position
- → 5 state more than one position
- → Secondary positions are marked in the dropdown below
Commenters and excerpts
One arrangement with the index administrator (26)
- ACU Markets: “The workable alternative is to place the obligation on the benchmark administrator rather than the DCM, and to make contributor access a condition of input admissibility rather than of listing.”
- Andreessen Horowitz (a16z): “Administrators should maintain full audit trails of source inputs and, ideally and where possible, develop and maintain mechanisms to share non-public transactional granularity with DCM surveillance teams”
- Architect Financial Technologies: “Accept an information-sharing agreement with the index administrator, together with contributor identification and trader recordkeeping rules, as sufficient for Core Principle 4 in most cases, without an agreement with each compute provider.”
- Babbage: “For contributed non-public records supporting settlement, the administrator should hold enforceable rights to audit, retain and produce the necessary records to the listing DCM and the Commission.”
- Compute Desk: “By requiring contributing providers to sign binding information-sharing agreements with Compute Desk as a condition of benchmark participation, and with information sharing agreements between Compute Desk and any DCM listing futures”
- De Silva Law Offices: “Where contributors submit data, the administrator should obtain by contract the right to pass that data to the exchange.”
- DNA Compute: “An exchange should maintain an information- sharing arrangement with each administrator, and through it with contributors, sufficient to reconstruct any print used in a final settlement”
- Douro Labs: “We therefore recommend against requiring a DCM to maintain an information-sharing arrangement with each compute venue and each capacity provider whose transactions or posted rates enter a settlement reference price.”
- DRW Holdings: “Place the surveillance information-sharing obligation on the administrator, rather than requiring a DCM to contract bilaterally with every venue and capacity provider.”
- Fiftyrivers: “DCMs lack cross-market visibility, which we would address through (i) information-sharing agreements with the index administrator as a listing condition”
- Futures Industry Association: “a DCM listing a cash-settled Compute derivative contract should, where practicable, maintain an information-sharing arrangement with the index administrator”
- Gemini: “The better structure is a single arrangement between the DCM and the index administrator, under which the administrator requires contributors by contract to furnish records on request”
- Hedge Quantitative Research: “A contract settling to such an index can still be listed if the index administrator's inputs are available to the exchange's surveillance under an information-sharing arrangement”
- Houman Shadab: “Clarify that the information-sharing arrangement between a DCM and a third-party compute index administrator may use layered access”
- Hydra Host: “Hydra recommends that the listing exchange maintain its information-sharing arrangement with the index administrator (the entity that calculates and publishes the index and that holds the contributor agreements and audit rights) rather than with each individual compute provider.”
- Hyperliquid Policy Center: “that a DCM be permitted to satisfy Core Principle 4 through an information-sharing arrangement with the index administrator rather than separate arrangements with every data source”
- Liquid Compute (PMEX Markets): “One workable altern ative is for the index administrator to require contributors, as a condition of inclusion, to per mit relevant transaction-level data to be made available to the Commission and the listing DCM on request”
- Mercatus AI: “Information-sharing arrangements should be enforceable, give timely access to contributor and counterparty details, and reach through any administrator to the underlying evidence.”
- NativX: “We favor a comprehensive agreement with the administrator, supported by enforceable obligations binding venues and reporters, over a mandatory separate contract with every provider regardless of the access already secured.”
- Ornn AI: “Where the benchmark administrator maintains appropriate rights to obtain contributor information, preserves auditable records, and can provide the DCM and Commission with information necessary for oversight, that structure may provide a functionally equivalent and more workable means”
- Polymarket US: “We support requiring a DCM to hold an information-sharing arrangement with the index administrator covering inputs, contributor identities, and methodology changes, consistent with the existing Appendix C guidance.”
- PTG: “A clear obligation for the administrator to retain constituent -level data and make it available to the listing DCM and the Commission on request.”
- Silicon Data: “Because the compute supply landscape evolves rapidly and a reliable benchmark must aggregate data across a large number of independent providers, a neutral benchmark administrator is positioned to serve as a centralized surveillance conduit.”
- Solutions by Jewel: “An information-sharing arrangement between the DCM and the index administrator, and with each venue or provider whose inputs exceed the disclosed weight threshold, covering contributor identity, price, volume, timestamp and posted-rate changes.”
- The Digital Chamber: “Information-sharing with the benchmark administrator, together with large-trader reporting and DCM surveillance capable of identifying relevant compute-provider accounts, should provide a more workable framework for monitoring activity affecting the benchmark.”
- Voltry: “Aggregated condition statistics, shared under an information-sharing agreement of the kind Appendix C paragraph (c)(3)(i) contemplates, give surveillance something to surveil.”
- Secondary position CCE: “We support enforceable, timely access throughout the data-contributor chain, directly or through a benchmark administrator.”
- Secondary position Dartmouth Advisory Partners: “The exchange holds an information-sharing arrangement with the administrator, and through the administrator with each contributor, sufficient to examine the inputs behind any settlement value.”
- Secondary position Fullview: “The arrangement may operate through an administrator where equivalent access is assured.”
- Secondary position Ömer Demirel: “The DCM should have an information-sharing agreement with the index provider and basic information access to every reporter, so that coordinated reporters cannot each stay below a threshold.”
An arrangement reaching every source (15)
- Abaxx Technologies: “the answer is affirmative: the registry is the arrangement, without separate agreements with each venue or provider.”
- Alec Snelling: “require information-sharing arrangements between each listing exchange and every compute venue and provider whose data feeds the settlement price”
- Arkhai (CoopHive): “Enable independent review of inputs and decisions:identify lawful, timely access to necessary records for every source of index inputs.”
- CCE: “We support enforceable, timely access throughout the data-contributor chain, directly or through a benchmark administrator.”
- Compute Capital: “Information-sharing with venues is necessary and not sufficient.”
- Dartmouth Advisory Partners: “Yes, a designated contract market should be expected to hold an information-sharing arrangement covering every venue and provider whose transactions or posted rates enter a settlement reference.”
- Evelyse Carvalho Ribas: “I therefore suggest that legal feasibility be treated as a condition to be established provider by provider and jurisdiction by jurisdiction, rather than as an undertaking a designated contract market can give in the abstract.”
- Fullview: “Every source whose transactions or posted rates enter the reference price should be covered by an enforceable arrangement providing the exchange with necessary information.”
- Hyperbolic Labs: “Hyperbolic would enter into such an arrangement with any DCM listing a compute contract that references its transactions.”
- Michael Ravnitzky: “Without mandatory data-sharing agreements, a designated contract market would be forced to rely on posted rates that can be changed at will and on partial transaction data that may not reflect true economic value.”
- Ömer Demirel: “The DCM should have an information-sharing agreement with the index provider and basic information access to every reporter, so that coordinated reporters cannot each stay below a threshold.”
- OneChronos: “information-sharing arrangements with cash venues, price-reporting agencies, and index contributors, as Question 2(d) contemplates”
- Simeon Bochev: “Yes, with each venue and each capacity provider whose transactions or posted rates enter the settlement reference, and I recommend the Commission treat this as a condition precedent to listing”
- Steven Skinner: “Information-sharing arrangements with contributing providers should cover these records, not only price feeds.”
- Vandermolen Power Limited: “A DCM should maintain enforceable information-sharing arrangements with every material venue and provider whose transactions or rates enter the settlement reference.”
- Secondary position Fiftyrivers: “information-sharing agreements with the index administrator as a listing condition, and with any contributor or venue that supplies a material share of inputs”
No mandatory arrangement (7)
- Animica Labs: “the transaction record is observable to the Commission, to a designated contract market, and to any member of the public, without an information-sharing agreement with the provider”
- Dragonfly: “Recognize publicly accessible transaction records as useful evidence for cross-market surveillance. Permit DCMs to combine them with existing trader-identification and information-access arrangements to meet 17 C.F.R. §§ 38.253(b) and 38.254, without imposing duplicative new reporting systems.”
- GPU Economy: “Every input of a posted-rate index is published by the provider, so a contract market and the Commission can check the inputs and recompute the arithmetic without an information-sharing agreement, at the time of reading.”
- Intercontinental Exchange: “Further, ICE does not believe that any “compute-specific” surveillance practices, such as mandatory information sharing agreements , are necessary .”
- Kalshi: “These safeguards will help reduce the risk of insider manipulation without the need for exchanges to enter an information-sharing agreement with every compute provider, which is likely impracticable and unworkable.”
- Paradigm and Andromeda: “requiring such arrangements would give compute suppliers an effective veto over whether a contract can be listed at all”
- VeloxVFX: “VeloxVFX does not recommend a categorical rule requiring an identical bilateral agreement with every provider regardless of materiality.”
- → 51 of 92 letters took a position
- → 11 state more than one position
- → Secondary positions are marked in the dropdown below
Commenters and excerpts
Allow with added safeguards (25)
- Abaxx Technologies: “For any compute contract — and especially any perpetual — require the DCM to assess the channel between derivative price signals and automated physical capacity management, with rate limits or confirmation gates on redemption-driven provisioning”
- Alec Snelling: “I also urge caution about offering highly leveraged perpetual-style contracts to retail customers”
- Andreessen Horowitz (a16z): “Subject to those considerations, perpetual contracts with periodic funding, dynamic margin, and DCM surveillance can concentrate liquidity and provide a continuous hedging instrument for ongoing workloads.”
- Balthasar Issa: “I would encourage the Commission, if it entertains perpetual compute futures, to require the same reference-price integrity standards”
- CCE: “We recommend transparent funding rules, reliable inputs at each calculation, proportionate leverage limits and tested disruption procedures for perpetual compute contracts.”
- Compute Desk: “Managing these unique hazards, including the hardware lifecycle, demands rigorous index governance, sophisticated margin requirements, and active market surveillance rather than dismissing the perpetual format as unviable for the compute asset class.”
- Dartmouth Advisory Partners: “We recommend that a perpetual list only against a reference that already meets the standard set out above.”
- De Silva Law Offices: “Perpetual compute futures should be listed only if their reference index meets the standards in Section VI.”
- Fiftyrivers: “We support perpetual compute futures, subject to conditions.”
- Fullview: “Safeguards should include understandable funding rules, position and leverage controls, stress testing, benchmark contingency arrangements and orderly termination procedures.”
- Gemini: “Permit perpetual compute futures through existing Part 40 pathways and publish the criteria now.”
- Hyperbolic Labs: “Of course, a perpetual referencing a specific hardware generation will decline toward obsolescence as that generation is retired, and the contract’s terms must specify in advance how the reference rolls to a successor grade.”
- ICX Global: “Perpetual contracts (Question 4) introduce an additional funding-rate mechanism that also depends on benchmark integrity, making a credible underlying benchmark especially important.”
- Liquid Compute (PMEX Markets): “These considerations do not ar gue against perpetual compute futures, but th ey support a more demanding review of reference-price integrity and funding-rate design before such products are listed.”
- NativX: “We recommend explicit funding intervals and bounds, reference-age limits, treatment of carried values, emergency suspension rules and transparent reopening procedures.”
- Ola Ologunde: “For perpetual or long-dated contracts, require a published rule for rolling between chip generations.”
- Ömer Demirel: “Perpetual compute futures should meet the five safeguards above.”
- Ornn AI: “Risks may be appropriately addressed through contract design, benchmark governance, margin and liquidation controls, and surveillance rather than by treating the perpetual structure itself as unsuitable for compute.”
- Polymarket US: “Funding should reference the most recently finalized index value, the funding formula should be published, and price bands should be set relative to the last index print.”
- Setara Financial: “We also support the development of perpetual compute futures where the product structure is appropriate to the underlying market and accompanied by appropriate risk controls.”
- Silicon Data: “Perpetual structures rely continuously on the reference index for funding calculations, requiring heightened ongoing index resilience”
- Simeon Bochev: “Condition perpetual listings on the index standards, rather than on the passage of time.”
- Solutions by Jewel: “For any perpetual compute listing, SBJ recommends that the Input Evidence Profile, conflict controls and determination records described in Comments 1, 2 and 4 apply at the frequency of the funding calculation”
- VeloxVFX: “Their safeguards should therefore address continuing, not merely initial, benchmark and market conditions.”
- Wei Wang: “I recommend the Commission require: (a) disclosure of the index update frequency and the maximum permissible staleness; (b) funding-rate caps or automatic pauses when the index is stale beyond that bound;”
- Secondary position ACU Markets: “If perpetual compute derivatives are permitted, the funding mechanism should be specified in the contract terms and analysed under Core Principle 3, because it is a settlement mechanism.”
- Secondary position Andrew Huff: “any later perpetual product begin with conservative leverage limits and enhanced retail disclosures”
- Secondary position Architect Financial Technologies: “Initially review perpetual compute contracts under 17 CFR 40.3, with the safeguards in Section VIII.”
- Secondary position Compute Capital: “If the Commission permits listed perpetuals, it should do so only against a print it can surveil, with funding-rate and liquidation safeguards sized for a market that is still a handful of names.”
- Secondary position Crypto Council for Innovation: “The Commission should focus on the integrity of the reference benchmark and the mechanism connecting the perpetual contract to it.”
- Secondary position Douro Labs: “We repeat our recommendation from the Perpetuals Letter that a DCM listing a perpetual contract should have access, through the index administrator, to contributor-level data sufficient to surveil the funding reference.”
- Secondary position Hedge Quantitative Research: “If a perpetual is listed anyway, the safeguards that matter are a published funding formula with a cap, a published index update schedule”
- Secondary position Hyperliquid Policy Center: “Three risks are specific to compute perpetuals, and each has a corresponding safeguard”
- Secondary position PTG: “PTG also encourages the Commission to ensure that perpetual contract specifications cannot change while positions remain open.”
- Secondary position Vandermolen Power Limited: “A DCM seeking to list a perpetual compute future should additionally demonstrate:”
Not at first (15)
- ACU Markets: “We recommend the Commission not authorise perpetual compute futures at inception.”
- Andrew Huff: “I recommend deferring perpetual compute futures until a fixed-date contract on the same index has operated with its vital signs published and within threshold”
- Architect Financial Technologies: “The Commission should require all initial compute perpetual contracts to be submitted for approval under 17 CFR 40.3, consistent with its May 2026 policy statement, and should consider delaying approval until dated futures have been listed and properly studied.”
- Capital Markets & Technology Partners: “The structure of perpetual Compute Futures behaves more like swaps products and warrants further investigation.”
- Compute Capital: “a perpetual can be a useful listed complement once a daily print exists. It does not replace tenor.”
- DNA Compute: “A perpetual should also follow, not precede, the dated contracts on the same reference”
- DRW Holdings: “We recommend the Commission allow conventional listed, financially settled, fixed-expiry compute futures to proceed on the ordinary product-listing path, and treat perpetual-style compute contracts as a separate, later question.”
- Evelyse Carvalho Ribas: “Establish whether a compute index can be those things before assessing the product's advantages.”
- Hedge Quantitative Research: “I do not see a commercial need that a perpetual meets and a monthly contract does not.”
- Michael Ravnitzky: “Until these risks are better understood, perpetual compute futures should not be listed.”
- Steven Singleton: “Before any DCM lists a perpetual compute contract, the request for comment should ask directly whether the payment mechanism proposed for keeping a perpetual contract tethered to its index would rely on the same data inputs”
- Steven Skinner: “Until a compute reference has a history of verified, delivered inputs and adequate depth, I would not approve perpetual compute futures, particularly for retail participants.”
- Submitter listed as "Grok": “Sequence perpetuals after short-dated fixed-date contracts and principal forwards have a settlement record.”
- Trevor Love: “A perpetual future referencing an undefined underlier doesn't resolve the problems above — it bakes them into market infrastructure.”
- Vandermolen Power Limited: “Defer broad retail access and perpetual compute futures until fixed-date markets, reference prices and surveillance arrangements have demonstrated resilience”
- Secondary position Simeon Bochev: “I would encourage the Commission to regard perpetuals as the right destination and the wrong starting point, and to condition their availability on the index standards at 2(a) rather than on the passage of time.”
Allow perpetuals (10)
- Crypto Council for Innovation: “The Commission should therefore permit exchanges to develop both structures where the contracts satisfy applicable requirements under the Commodity Exchange Act and Commission regulations”
- Douro Labs: “Permit perpetual compute contracts.”
- Dragonfly: “Dragonfly supports permitting perpetual compute futures alongside traditional fixed-date contracts.”
- Hyperliquid Policy Center: “The Commission should permit perpetual compute contracts to proceed through its existing case-by-case review under Regulation 40.3”
- INFEREXUS: “We raise this not to argue against perpetual products, but to encourage the Commission's forthcoming framework to leave room for both structures to compete on their merits.”
- Jump Trading Group: “Support for physical settlement is consistent with support for perpetual compute futures, which may serve different risk management needs and can be listed alongside physically settled contracts.”
- Kalshi: “The Commission should permit well- designed perpetual compute contracts that satisfy the Core Principles and existing CFTC regulations”
- Paradigm and Andromeda: “We believe perpetual futures are a natural fit for compute, but that they are complementary to fixed -date contracts, rather than a replacement for them.”
- PTG: “PTG sees genuine promise in perpetual contracts.”
- The Digital Chamber: “The Commission should therefore permit DCMs to develop perpetual compute products where there is commercial demand and evaluate those products under the same existing framework applicable to other derivatives products.”
Do not allow (1)
- Better Markets: “The Commission should focus on making traditional, fixed-date futures contracts with compute as the underlying commodity workable rather than promoting perpetual futures on compute.”
- → 16 of 92 letters took a position
Commenters and excerpts
Self-certification under existing rules (8)
- Andreessen Horowitz (a16z): “The self-certification and approval procedures in Part 40 are sufficient for these products”
- Architect Financial Technologies: “Confirm that a DCM may list a compute derivative under the existing principles-based framework, on a product-by-product basis, including by self-certification under 17 CFR 40.2 for a dated, cash-settled contract.”
- DRW Holdings: “We recommend the Commission allow conventional listed, financially settled, fixed-expiry compute futures to proceed on the ordinary product-listing path, and treat perpetual-style compute contracts as a separate, later question.”
- Gemini: “A DCM should be expected to demonstrate, in its submission under 17 CFR §§ 40.2 or 40.3, using the data available to it:”
- Jump Trading Group: “DCMs should also retain access to the Part 40 self-certification process for new compute contracts.”
- Polymarket US: “A new underlying commodity should not, by itself, turn the ordinary self-certification process into a requirement for Commission approval.”
- PTG: “PTG does not believe compute derivatives require a new or separate listing process.”
- The Digital Chamber: “Self -certification should remain the ordinary listing path”
Commission approval or added review (8)
- ACU Markets: “Direct that first-of-kind compute contracts proceed by Commission approval under 17 CFR 40.3 rather than self-certification under 17 CFR 40.2 for a defined initial period”
- Compute Desk: “Compute Desk also asks the Commission to apply heightened review, rather than self-certification, to any settlement methodology designed or controlled by a firm with significant market-making, proprietary trading or capacity-ownership interests”
- Crypto Council for Innovation: “Accordingly, compute contracts may warrant particularly careful review of benchmark representativeness, source independence, manipulation resistance, and settlement methodology before listing.”
- De Silva Law Offices: “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.”
- DNA Compute: “Regulation 40.3 review for first-of-kind compute contracts”
- Simeon Bochev: “The Commission should state that it will review self-certified compute filings against the same criteria.”
- Steven Singleton: “The concern here is sequencing: the diagnostic questions in Sections I, II.2, and II.4 need an enforceable link to the self-certification process, not just a comment docket, before the market moves faster than the answers do.”
- Submitter listed as "Grok": “The Commission extended review of the CME H100 and B200 compute futures filing to November 9, 2026, citing novel or complex issues and this request. Take that time.”
- → 28 of 92 letters took a position
- → 7 state more than one position
- → Secondary positions are marked in the dropdown below
Commenters and excerpts
Build it on purpose: market makers, one instrument, fewer months (13)
- Andreessen Horowitz (a16z): “Subject to those considerations, perpetual contracts with periodic funding, dynamic margin, and DCM surveillance can concentrate liquidity and provide a continuous hedging instrument for ongoing workloads.”
- Architect Financial Technologies: “Liquidity programs. Market maker programs with quoting obligations for size and spread during the first contract months.”
- DNA Compute: “In each case the index became the benchmark because the grade was defined narrowly enough to hedge and because physical business settled against it through an exchange-for-physical or index-plus-differential mechanism.”
- Fiftyrivers: “dynamic price bands, settlement procedures for thinly traded periods (Q2(e)), and minimum-liquidity criteria before deferred months or new grades are listed.”
- Gemini: “In an early-stage market, perpetual contracts also concentrate liquidity in a single contract rather than dividing it across many expiries, which is a meaningful benefit when liquidity is thin.”
- Hivemind Capital Partners: “Listings should concentrate liquidity on high-granularity tenors within a 12-month window.”
- Hydra Host: “Pricing by provider would also fracture trading into pools too small to support liquid hedging.”
- Jump Trading Group: “Jump requests that the Commission consider rulemaking to exclude otherwise countable dealing activity in centrally traded and cleared compute swaps from the de minimis calculation, subject to appropriate eligibility conditions and safeguards.”
- Kottos AI: “I therefore propose starting with one clearly specified open-weight inference grade and expanding as transaction volume increases.”
- OneChronos: “Accordingly, a market design that accommodates multiple attributes directly can concentrate liquidity that fixed contracts would divide, and can support narrower, manipulation-resistant indices as transaction data matures under the existing framework.”
- Ornn AI: “That structure may also concentrate liquidity that otherwise could be divided among successive contract months.”
- PTG: “PTG respectfully urges the Commission to confirm that bona fide market -making in exchange -traded, cleared compute derivatives should not be counted toward swap dealer de minimis thresholds.”
- Stark Executive Advisory: “Splitting a compute index by every hardware variable fragments liquidity until the contract becomes thin and untradable.”
- Secondary position Crypto Council for Innovation: “A perpetual structure may also concentrate trading and hedging activity in a single continuously available instrument rather than dispersing liquidity across many expiries. This may be particularly valuable while compute derivatives markets are developing.”
- Secondary position Douro Labs: “Perpetual structures also let a market develop before the term structure has enough liquidity to support each contract month.”
- Secondary position Hyperliquid Policy Center: “Because a perpetual contract does not expire, trading interest gathers in a single instrument rather than fragmenting across a ladder of contract months, which matters most in a new market where liquidity will be scarce.”
- Secondary position The Digital Chamber: “As the compute market develops, however, the Commission should consider whether application of the minimis calculation could unintentionally discourage firms from providing liquidity in standardized, centrally traded and cleared compute swaps.”
A listed contract creates it (10)
- Crypto Council for Innovation: “The evolution of energy commodities and, more recently, Bitcoin demonstrates how derivatives markets can emerge alongside developing spot markets to provide price discovery, liquidity, and tools for managing price risk.”
- Douro Labs: “In our experience, a liquid listed contract is what standardizes and deepens an underlying cash market—not the reverse.”
- DRW Holdings: “Without an exchange-traded listing, hedges will be made via bilateral, privately negotiated swaps referencing whatever index two counterparties are willing to accept”
- Hyperbolic Labs: “A listed contract settling to a published index would give the cash market a public reference price.”
- Hyperliquid Policy Center: “Listed markets can themselves improve standardization, price discovery, and disclosure, as RFC Question 1(f) recognizes in asking how listing might affect provider incentives.”
- Kalshi: “A well-regulated compute futures market will strengthen the underlying cash market by promoting common commercial specifications, supporting transparent price discovery, and expanding hedging opportunities.”
- Paradigm and Andromeda: “We believe that liquidity and significant market-initiated changes will follow, which would not only benefit the industry, but also address many of the Commission’s concerns expressed in the RFC.”
- Polymarket US: “That lowers the cost of building capacity and lets operators keep more capacity on the open market, which deepens the underlying cash market.”
- Silicon Data: “With an IOSCO-aligned index settling on a DCM, commercial participants will be able to structure and evaluate bespoke, long-dated OTC swaps referencing the same or related indices, greatly increasing financing flexibility.”
- The Digital Chamber: “The derivatives market itself can help generate the transparency and standardization necessary for the broader market to mature.”
- Secondary position Architect Financial Technologies: “Opacity is also not necessarily permanent. Spot and short-tenor venues for compute are growing, and listed derivatives likely accelerate that growth.”
- Secondary position PTG: “PTG also expects a robust listed, cash -settled market to help physically settled contracts develop, because a transparent forward curve gives buyers and sellers the confidence to contract for delivery.”
The cash or OTC market has to come first (5)
- Better Markets: “the “danger is excessive speculation without a strong underlying cash market, which can create noisy price discovery and poor hedging performance.””
- Compute Capital: “The lesson is that listed futures inherit liquidity from the OTC market they sit next to. They do not invent it.”
- Liquid Compute (PMEX Markets): “In each case, the reference came first and the derivative followed.”
- Simeon Bochev: “This is the most important structural absence in the market and it is the reason I have argued elsewhere that the ladder from standard to spot to index to hedge to contract is being climbed out of order.”
- Simeon Bochev: “Together they make compute transactions comparable and auditable, which is the foundation of market integrity and, later, of liquidity.”
- Secondary position OneChronos: “an active, well-functioning cash market is a precondition of success.”
- → 21 of 92 letters took a position
- → 3 state more than one position
- → Secondary positions are marked in the dropdown below
Commenters and excerpts
Hard limits sized to the cash market (14)
- ACU Markets: “R9. Calibrate position limits and accountability to a flow-based deliverable-supply measure incorporating a transferability factor for physically-settled contracts, and to input-population size and diversity for cash-settled contracts.”
- ACU Markets: “Position limits Calibrated to the size and diversity of the input population for the contract month, under the deliverable supply methodology at [reference]”
- Architect Financial Technologies: “Accept spot-month limits based on uncommitted capacity and transaction depth, with accountability outside the spot month”
- CCE: “We recommend basing position limits on capacity available for the contract's grade, location and delivery period.”
- Compute Capital: “Deliverable supply for position-limit design should be measured as installed, energized, contracted capacity at a named node and SKU”
- De Silva Law Offices: “Exchanges should use that alternative to set conservative spot-month limits based on index-eligible volume.”
- DNA Compute: “We recommend that the exchange publish V and its method in summary, and set accountability levels and any spot-month limit as a fraction of V/730”
- Fiftyrivers: “Spot-month limits must be no greater than 25 percent of estimated spot-month deliverable supply.”
- Hedge Quantitative Research: “position limits are set against uncontracted capacity rather than installed capacity”
- NativX: “Cash-settled limits also need to reflect the size and influenceability of the price-forming market.”
- Simeon Bochev: “The Commission should size position limits against transferable capacity, not against the market's apparent scale, precisely because the apparent scale is so much larger than the corner-relevant quantity.”
- Stark Executive Advisory: “This segment represents the metric for calculating deliverable market depth and position limits.”
- The Digital Chamber: “Because unused compute cannot be stored for later delivery, spot month limits should be set conservatively against estimated deliverable supply.”
- Wei Wang: “Position limits and large-trader reporting, calibrated to the thinness of the underlying market rather than copied from energy or metals templates.”
Accountability levels at the exchange’s discretion (7)
- DRW Holdings: “Apply position accountability rather than hard position limits at this stage, calibrated to the economic service flow at the pricing point rather than to a deliverable-supply analogy.”
- Futures Industry Association: “Thinly traded markets may require more robust surveillance and safeguards such as calibrated position accountability levels and real-time monitoring for anomalous price movements.”
- Gemini: “Preserve DCM discretion over accountability level methodology and thin-market controls”
- Kalshi: “DCMs should be able to use position accountability in place of fixed position limits, particularly during the market’s early development.”
- Liquid Compute (PMEX Markets): “On position limits and accountability, PMEX believes that, given current data limitations, position accountability may be more readily calibrated than a hard numerical limit.”
- Polymarket US: “We ask the Commission to confirm that position limits and accountability levels for compute”
- Vandermolen Power Limited: “Apply conservative position-accountability levels until verified deliverable supply and benchmark depth are demonstrated.”
- Secondary position Architect Financial Technologies: “Position accountability outside the spot month. Accountability levels, rather than hard limits, outside the spot month.”
- Secondary position De Silva Law Offices: “Outside the spot month, accountability levels backed by large -trader reporting fit better than hard limits until the market has more history.”
- Secondary position Fiftyrivers: “we would set limits conservatively within that ceiling and use accountability levels in deferred months.”
- → 32 of 92 letters took a position
- → 6 state more than one position
- → Secondary positions are marked in the dropdown below
Commenters and excerpts
Open to retail with standard protections (19)
- Abaxx Technologies: “Retail participants should see, in the contract terms, exactly which claims are deliverable to whom.”
- Architect Financial Technologies: “Architect believes that a supplemental, plain-language compute risk disclosure (Section VII.B) would be useful for retail customers.”
- Dartmouth Advisory Partners: “A retail participant should be told that plainly, in the contract terms and not only in general risk language.”
- De Silva Law Offices: “A retail customer should be told what the index measures and how it is built.”
- DNA Compute: “We recommend a standardized risk disclosure covering the items in 3(c), published by the exchange and delivered by every intermediary, rather than access restrictions.”
- Douro Labs: “Retail and institutional participants alike should be able to see, at any time, how many independent inputs stand behind the price they are trading and how much those inputs disagree.”
- Evelyse Carvalho Ribas: “disclosure to retail participants should convey that the relevant risk includes discontinuous changes in the availability of the underlying arising from administrative action, and not only price volatility.”
- Gemini: “Do not create a bespoke retail regime keyed to the underlier.”
- Hyperbolic Labs: “Hyperbolic sees no basis for restricting access to compute derivatives by participant category.”
- Hyperliquid Policy Center: “These protections matter most for the participants the AI Action Plan singles out, the startups and academics that are least able to negotiate bilateral alternatives, and registered venues can implement each of them under the Commission’s existing framework.”
- Kalshi: “the Commission should not impose any compute-specific limitations on retail access to these markets.”
- NativX: “We do not recommend a categorical retail exclusion merely because the underlier is compute.”
- Paradigm and Andromeda: “We do not believe retail -specific protections are necessary.”
- Polymarket US: “Restricting access by participant category would also shut out many of the firms with the most direct exposure.”
- Steven Singleton: “Tying any retail-facing compute contract to a specific, dated disclosure requirement, rather than leaving it for unscheduled future guidance, would close this before the first contract lists rather than after a problem surfaces.”
- Steven Skinner: “Retail disclosures should state plainly that the index reflects a private, concentrated market and explain how its inputs are verified.”
- Taylor Holland: “With consumers like startups and small businesses, academics and universities, and retail customers conceivably among the buyers seeking to manage price risk, open access is key.”
- Trevor Love: “Retail-facing materials should not describe a contract simply as exposure to “compute prices” if its settlement series materially reflects discounted interruptible capacity or another narrower service tier.”
- VeloxVFX: “The record presently available to VeloxVFX does not support a categorical access prohibition or other universal restriction.”
Retail with added limits (8)
- ACU Markets: “We suggest an appropriateness or knowledge assessment reflecting the first-of-kind nature of the underlier; plain-language disclosure covering the items at 3(b); conservative initial leverage limits pending an observed liquidity and volatility history; position limits that bind at retail scale”
- Alec Snelling: “Retail access is fine, but as a modest, well-disclosed market, not a high-leverage casino attached to a critical input.”
- Andrew Huff: “I recommend deferring perpetual compute futures until a fixed-date contract on the same index has operated with its vital signs published and within threshold, and that any later perpetual product begin with conservative leverage limits and enhanced retail disclosures.”
- Better Markets: “Perpetual futures, with their use of leverage, are especially risky for retail investors who may not understand the risk of loss that they entail.”
- CCE: “We support proportionate retail protections based on leverage, complexity, liquidity and delivery obligations. Interfaces should distinguish purchasing compute from trading a derivative, with clear loss scenarios and appropriate access checks and limits.”
- Fiftyrivers: “For retail users we would add leverage limits on perpetuals and a period of institutional trading before retail-sized contracts are offered”
- Fullview: “Exchanges and intermediaries should consider appropriately sized contracts, clear cost disclosures, realistic stress illustrations and access controls consistent with applicable requirements.”
- Liquid Compute (PMEX Markets): “Additional leverage or access controls may be appropriate depending on the robustness of the reference market and the design of the product.”
- Secondary position Architect Financial Technologies: “Initial margin calibrated to a multi-day liquidation horizon and to jump risk, with leverage limits for retail customers.”
- Secondary position NativX: “Access, leverage and suitability controls should follow the product's demonstrated liquidity and customer risks.”
- Secondary position Simeon Bochev: “Position limits for non-commercial participants set conservatively, in light of the thin-liquidity concerns at 2(e).”
- Secondary position Steven Skinner: “Until a compute reference has a history of verified, delivered inputs and adequate depth, I would not approve perpetual compute futures, particularly for retail participants.”
- Secondary position Vandermolen Power Limited: “Retail access should be approached cautiously. A contract labeled as a GPU-hour can appear simpler than it is. Until product behavior is established, enhanced appropriateness, leverage, disclosure and concentration controls would be justified.”
Institutions only at first (5)
- Balthasar Issa: “I would be hesitant to make compute derivatives available to retail before any reference-pricing structure discussed under Question 2(g) exists.”
- DRW Holdings: “We would support an initial market oriented to institutional and commercial hedgers and professional liquidity providers, with retail access, if any, subject to clear risk disclosure”
- Simeon Bochev: “My view is that compute derivatives are not presently an appropriate retail product, and I hold that view more firmly than any other in this section.”
- Simeon Bochev: “Initial participation is proposed for commercial counterparties with verified ability and intent to receive or resupply the capacity.”
- Vandermolen Power Limited: “Defer broad retail access and perpetual compute futures until fixed-date markets, reference prices and surveillance arrangements have demonstrated resilience across hardware transitions and periods of stressed capacity.”
- Secondary position Fiftyrivers: “For retail users we would add leverage limits on perpetuals and a period of institutional trading before retail-sized contracts are offered”
- → 27 of 92 letters took a position
Commenters and excerpts
Raises export controls (27)
- Abaxx Technologies: “Compute is export-controlled: accelerator availability, interconnect, and data residency are subject to the Export Administration Regulations (EAR)”
- ACU Markets: “Consult BIS and OFAC on the export-control and sanctions implications of synthetic exposure to restricted compute capacity before the first compute contracts are listed.”
- Andrew Huff: “Consult the Department of Commerce so that no compute contract, particularly a physically settled one, becomes a channel for parties restricted under U.S. export controls to obtain compute.”
- Architect Financial Technologies: “The disclosure should address seven risks: geopolitical and export-control risk; supplier concentration; non-storability; technology obsolescence; index methodology risk; basis risk; and liquidity risk in a new market.”
- Balthasar Issa: “I would encourage the Commission to require disclosure that a compute derivative's value and deliverable supply may be materially affected by export restrictions.”
- CCE: “We recommend disclosing geographic exposure and how sanctions, export restrictions, power disruptions and supply concentration could affect price or delivery.”
- Dartmouth Advisory Partners: “A change in accelerator export rules can shift compute supply and price quickly, and it may fall unevenly across the regions an index draws from.”
- De Silva Law Offices: “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.”
- DNA Compute: “The hardware underlying these contracts is subject to United States export controls.”
- Douro Labs: “export-control and other geopolitical constraints on the underlying”
- DRW Holdings: “To the extent the Commission has geopolitical concerns, they attach to the physical supply chain and to export-control regimes, not to the derivative.”
- Evelyse Carvalho Ribas: “For advanced compute the transmission is different in character, because access to the underlying is itself subject to export control.”
- Fiftyrivers: “The greater risk is at any physical-delivery or EFRP leg, where delivered compute could be used to evade export controls or sanctions, so those legs should require identification of the ultimate capacity user and jurisdictional screening (Q3(a)).”
- Fullview: “Financial-crime controls should not be conflated with separate export-control or service-access obligations, although the same transaction may require attention to more than one regime.”
- Gemini: “Compute prices may be affected by export controls, trade restrictions, tariffs, and energy supply conditions affecting semiconductors and data center capacity.”
- Hivemind Capital Partners: “Formally affirm the legal distinction between cash-settled financial hedging (CFTC exclusive jurisdiction) and physical capacity provisioning (BIS/CFIUS compliance at delivery).”
- Hyperbolic Labs: “Hyperbolic agrees that a contract’s terms and conditions and the associated risk disclosures should address, at a minimum, that: (1) the underlying hardware is subject to export controls and that changes in those controls can move prices sharply”
- Hyperliquid Policy Center: “disclosures should explain how export restrictions or other geopolitical developments may affect compute prices and access.”
- Liquid Compute (PMEX Markets): “PMEX therefore supports contra ct-specific disclosure of material export-control and supply-restriction risks where those risks co uld affect the underlying reference market, deliverability, or price formation.”
- NativX: “Relevant risks include export restrictions, sanctions, data-residency requirements and restrictions on remote access, as well as changes in power or network access.”
- Ola Ologunde: “Ask venues to state plainly which risks a compute contract hedges and which it does not: residual value, delivery, renewal, lien ranking and control (export limits and remote restriction).”
- Polymarket US: “These same risk mitigation tools, together with applicable sanctions and export-control compliance, are available to address potential risks inherent in compute.”
- PSYCHE: “A disclosure regime for a contract settling against a commodity whose supply is subject to jurisdictional restriction presupposes that the contract states which grade of the commodity it references.”
- Simeon Bochev: “The Commission should coordinate with the Bureau of Industry and Security at the Department of Commerce before finalizing any guidance on compute derivatives.”
- Simeon Bochev: “Screening, sanctions, export and jurisdiction conditions govern suspension, lawful recipient replacement or consensual transfer, mitigation and termination.”
- Squaretower: “Since 2022, the U.S. government has restricted exports of Nvidia and rival AMD chips to destinations including China and Russia”
- Vandermolen Power Limited: “Require clear fallback and disruption procedures for outages, cyber incidents, sanctions changes, export controls, index failure and rapid hardware obsolescence.”
- → 27 of 92 letters took a position
Commenters and excerpts
Physical delivery as well, or first (21)
- Abaxx Technologies: “We recommend the Commission treat settlement by transfer of a claim ticket as the presumptively appropriate settlement form for compute futures, with cash settlement permitted only where the Appendix C index standard is met.”
- Andreessen Horowitz (a16z): “Physically-settled contracts may also become viable.”
- Architect Financial Technologies: “The Commission can instead encourage the conditions that produce qualifying series: transaction reporting to independent administrators, standard contract terms in the cash market, and exchange-for-physical mechanisms that connect futures to physical delivery.”
- Compute Desk: “An institutional market participant holding a short futures position against physical GPU compute can exchange a futures contract based on Compute Desk’s index for an equivalent OTC bilateral bare metal GPU compute agreement with a vetted counterparty.”
- Fiftyrivers: “Encourage exchanges of futures for related positions (EFRPs) against grade-conforming capacity, and develop a physically delivered pilot.”
- Hivemind Capital Partners: “physically deliverable mechanisms—specifically Exchange for Physical (EFP) protocols and compute depositary receipts—are necessary during early market phases to anchor index integrity to real data center capacity.”
- Houman Shadab: “Write physical-delivery standards so that attested delivery records may form part of a DCM’s demonstration under the quality, delivery-instrument, and inspection provisions of appendix C”
- Hyperbolic Labs: “Hyperbolic agrees that cash settlement will come first but urges the CFTC to recognize that physical delivery is practical.”
- INFEREXUS: “We further believe that physical delivery — that is, settlement in the form of actual convertible access to inference capacity, rather than cash-only settlement — provides an important customer protection benefit”
- Jump Trading Group: “We encourage the Commission to accommodate physical delivery alongside cash settlement, with enforceable access rights, clear grading, and realistic deliverable supply estimates.”
- Liquid Compute (PMEX Markets): “Treat physical deliverability as a favor ed anti-manipulation feature. Convergence discipline is worth the operational cost”
- Ömer Demirel: “Physical delivery reduces reliance on a reported-price index.”
- OneChronos: “Physically settled futures, listed on a DCM and cleared through a derivatives clearing organization, carry a favorable Core Principle 3 and 4 profile when paired with clear delivery rules and cash-price reference data.”
- Paradigm and Andromeda: “the deployment of physically- delivered contracts would resolve most of the remaining concerns expressed in the RFC”
- PSYCHE: “The first is to list a physically-settled contract, which carries the specification described at note 6 and at Request 4.”
- PTG: “The Commission should not presume that cash settlement is the only viable design.”
- Silicon Data: “Cash-settled, physically settled, and perpetual contracts serve distinct hedging contingencies and require different operational safeguards.”
- Simeon Bochev: “This appendix sets out a bilateral, physically performed forward purchase of a defined unit of compute for a fixed future service period.”
- Submitter listed as "Grok": “I write on short-tenor GPU forwards where the seller stands as principal, takes the buyer's price risk, and must deliver the capacity.”
- The Digital Chamber: “The Commission should permit the natural development of multiple settlement models under its existing framework rather than require the market to choose a single structure or treat cash settlement as the preferred endpoint.”
- Vandermolen Power Limited: “For physical contracts and deliverable-supply analysis, require a versioned record of capacity, configuration, location, service attributes, encumbrances, delivery telemetry and retirement or consumption.”
Cash settlement only (6)
- ACU Markets: “Absent a provider consent regime or a genuinely transferable capacity instrument, physical settlement is not merely operationally difficult — for much of the market it is contractually prohibited.”
- CME Group: “CME’s Compute Futures have been structured as financially-settled futures based on the arithmetic average of the settlement prices for the Silicon Data H100 Rental Index”
- DRW Holdings: “Confirm that cash settlement to a robustly constructed, independently administered index can satisfy DCM Core Principle 3 for compute”
- Hydra Host: “the Commission should concentrate its framework on contracts that settle in cash against a published compute price index.”
- Michael Ravnitzky: “default to cash settlement until the cash market proves robust”
- Nodal Exchange: “Utilizing benchmark indices from independent third-party index providers is a prudent path to managing the introduction of financially settled compute futures.”
- → 11 of 92 letters took a position
Commenters and excerpts
Allow token or inference contracts now (7)
- Animica Labs: “Where the architecture publishes, the Appendix C predicates are satisfiable today.”
- Hivemind Capital Partners: “the Commission’s immediate focus should center on Datacenter Fabric Accelerators (GPU/TPU hours) and Cloud Fabric (Inference Tokens)”
- Hyperbolic Labs: “Inference token contracts, if listed, should be treated as a separate product class that references a specified model with a settlement methodology adapted to it.”
- Hyperliquid Policy Center: “Both forms of compute are suitable underliers for derivatives contracts, and they serve different participants.”
- INFEREXUS: “The CIF is designed to be listed and traded on existing regulated derivatives exchanges and cleared through existing derivatives clearing organizations”
- Jump Trading Group: “Relevant underliers may include accelerator capacity by hardware grade, inference or token throughput, memory- or bandwidth-based measures, and capacity available in specified regions.”
- The Digital Chamber: “A contract may reference a particular GPU model, a defined quantity of GPU -hours, a specified provider or geographic market, a service-level measure such as inference tokens, or another defined form of compute capacity.”
Later, after GPU-hour contracts or once measurable (4)
- Architect Financial Technologies: “Token-based underliers can become suitable as real-time inference venues publish cleared prices.”
- Dragonfly: “Treat inference throughput as a separate potential contract class where model choice, version, and measurement protocol can be standardized reliably.”
- Hydra Host: “Hydra recommends against listing token- referenced contracts until one can be shown not to be readily susceptible to manipulation, the test set by Core Principle 3.”
- Kottos AI: “For contracts based on inference service we are missing a common specification of the deliverable, followed by evidence of transactions and delivery within that specification.”
10 What the Letters Raise Most
A letter counts toward a subject when it makes an ask or an argument about it.
- Manipulation and false reporting Question 2(a), 2(b) 78
- Defining a grade of compute Question 1(e) 77
- Published method and inputs Question 2(g) 68
- Concentration among contributors Question 2(c) 65
- Data access for the exchange and the CFTC Question 1(c), 2(d) 59
- Proof that compute was delivered Question 1(d), 2(d) 33