The U.S. Commodity Futures Trading Commission asked the public how designated contract markets (DCMs), the CFTC-registered futures exchanges, should think about listing derivatives that settle to the price of computing power. The request is RIN 3038-AF77, issued in Washington, D.C., on August 19, 2026. It is a request for comment, not a product approval and not a live GPU futures board.
That distinction matters for crypto desks because miners, data-center landlords, and perpetual futures venues already treat “compute” as a story. A comment file is not the same thing as a hedge you can click.
Bitcoin (BTC) traded at $69,228 on our CoinGecko snapshot at 2026-08-19T23:51:45+00:00, up 7.2% over 24 hours. Ether (ETH) was $2,253.08, up 17.6%. Those prints tell you the tape is awake. They do not tell you a compute contract is listed.
What happened
CFTC Press Release 9286-26, dated August 19, 2026, says the Commission issued a request for comment “to better inform its understanding and oversight of derivatives markets in compute.” Comments will be accepted for 60 days after publication in the Federal Register. A Federal Register API search for RIN 3038-AF77 at 2026-08-20T00:12Z returned no published document. The PDF itself still carries the placeholder “INSERT DATE 60 DAYS AFTER DATE OF PUBLICATION IN THE FEDERAL REGISTER,” and a header that the text is “subject to Office of the Federal Register technical corrections.”
The RFC defines compute, in the Commission’s words, as processing power “primarily used by the large language models (LLMs) at the center of the artificial intelligence (AI) economy.” It asks DCMs how they would address Commodity Exchange Act (CEA) core principles when the underlier is that scarce capacity. The document also invites comment on swap execution facilities listing contracts that settle to compute prices, and on physically settled compute swaps. That is a wide net. It is still a questionnaire.
Chairman Michael S. Selig’s quote in 9286-26 frames compute as “the commodity that will power the intelligence economy” and calls the RFC “the first step toward establishing clear rules of the road.” PYMNTS, writing the same day from the press release, repeats those lines and the 60-day Federal Register clock. I am using that outlet as a second-domain check, not as the lead. The lead is the RFC PDF.
The RFC is labeled a significant regulatory action under Executive Order 12866 and says the Office of Management and Budget reviewed it. Secretary Christopher Kirkpatrick issued the text on August 19, 2026. None of that lists a contract.
Context
Crypto already has a messy compute tape. Hashprice is one proxy for Bitcoin mining. GPU rental rates are another, mostly private. Bitcoin miners have been leasing power and halls into AI jobs; we walked through that mix on Riot’s Q2 AI lease. Those cash deals do not automatically become a CFTC-grade underlier.
The RFC’s own questions make that gap explicit. Section II asks what share of compute trades at publicly disclosed prices versus undisclosed bilateral agreements, and whether it would be “appropriate to permit trading in a derivative contract settling to a price computed from data that the Commission may not be able to observe, verify, or surveil.” Core Principle 3 still says a DCM may list only contracts not readily susceptible to manipulation. Appendix C to part 38, which the RFC cites, wants a cash settlement price that is reliable, acceptable, publicly available, and timely, drawn from a cash market that is liquid enough and not itself easy to squeeze.
That is the market-structure problem I care about. Narrative is cheap; liquidity you can audit is not. On Coinbase Exchange BTC-USD, as of 2026-08-20T00:12:05Z, the last trade was $69,622.86, the 24-hour volume was about 16,894 BTC, and depth within 1% of mid was roughly $16.9 million on the bid and $16.7 million on the ask (Coinbase Exchange stats, plus the level-2 book at the same minute). That is a public, tick-by-tick cash market. Compute’s posted GPU rates and private reservations are not built like that yet. The CFTC is asking, out loud, whether they ever will be.
A footnote in the RFC cites Federico M. Bandi’s July 1, 2026 paper “(Early) AI Compute Asset Pricing” for a 2025-Q4 installed compute stock implying a gross compute service flow of about $430 billion to $1.3 trillion per year, or about 1.4% to 4.0% of U.S. GDP. That is the Commission’s citation, not a number I independently rebuilt. Size does not solve the reference-price problem. Oil is large too. Oil also has observable cargoes and published assessments. Compute, as the RFC describes it, still leans on provider-administered posted rates and bilateral books.
Perpetual compute futures get their own question set. The Commission asks whether a perpetual would hedge risks that dated futures cannot, and what extra safeguards it would need. Crypto traders know the product shape. That does not mean a DCM can copy a crypto perp onto an opaque GPU index and call Core Principle 4 satisfied. Surveillance, the RFC notes, may require information-sharing with each venue and each capacity provider whose trades or posted rates enter the settlement print.
If you are mapping this onto Bitcoin’s history, keep the analogy tight. Crypto perps grew on 24/7 venues with public marks. Compute futures, if they list, have to survive a DCM rulebook that was written for grains, metals, and energy, then stretched toward event contracts. We already watched the CFTC tighten paper around prediction-market incentives in Letter 26-23. Compute is the next underlier test, not a shortcut around that paper.
Our read
I read RIN 3038-AF77 as a sequencing document. The Commission wants a comment file before it blesses, stays, or quietly tolerates a compute listing. Selig’s “gold standard” line is a political frame. The operative text is still Core Principles 3, 4, and 5: manipulation resistance, surveillance of the cash-settlement process, and a deliverable-supply method for position limits.
My stance: treat any “compute futures are coming in October” chatter as marketing until a named DCM files under CEA section 5c(c) and the CFTC either lets the self-certification stand or issues an approval. The RFC even asks whether listing a futures contract on a published compute index would change how providers post rates, structure bilateral reservations, or allocate capacity. That is a warning about circularity. The people who can move the cash print may be the same people who supply the chips.
Falsifiable claim: by 2026-10-31 23:59 UTC, the Federal Register will show RIN 3038-AF77 as a published Request for Comment with a comment-due date, or CFTC.gov will post a withdrawal or replacement of the August 19 RFC. If neither a Federal Register notice nor a withdrawal exists by that deadline while Press Release 9286-26 remains live, this “comment clock starts after OFR” read is wrong.
What to watch
Watch the Federal Register, not social screenshots. The 60-day clock does not start on the press date. It starts on publication. Until that notice is up, Regulations.gov will not have a stable comment-by date for 3038-AF77.
Watch DCM filings, especially CME, Cboe, or any crypto-adjacent DCM that tries a 40.6 self-certification for a compute future or a perpetual. A listing that cash-settles to an unpublished bilateral index, without answering the RFC’s Core Principle 3 questions, is the failure mode.
Watch miner and hosting IR for named offtake that could become deliverable supply. A lease is not a futures contract. It can still be the cash market a DCM would have to measure. Pair those prints with a simple market snapshot habit: timestamp the coin tape, then ask whether the underlier you are hedging even has a public mark.
A $69k BTC print does not price a GPU hour. If the two markets ever share a hedge, it will be because someone published a boring, auditable index, not because an RFC used the word “commodity.”