November 16, 2026
The post on the GENIUS Act in August ended on the point that the law existed in statute, but the operational details were still being written through rulemaking. Part of that writing is now done. On October 21, 2026 the OCC issued its final rule implementing the Act (Bulletin 2026-50). It is the first federal stablecoin regulator to finalize its rules, and it covers national banks and their subsidiaries, federal savings associations, federal branches, foreign issuers, and the nonbank issuers that the OCC supervises directly. Most of it lands in a new part of the federal regulations, 12 CFR 15.
The effective date is January 18, 2027. That matches the timing described in August: the Act takes effect on the earlier of 18 months after signing or 120 days after the final rules. A rule issued in late October puts the 120-day mark in February, so the 18-month date comes first.
The interesting part is not that the rule exists, but where it moved compared to the proposal published in March. The OCC received a few hundred comments, and some of the changes are substantial. The numbers below come from the law firm summaries published since then, mainly Gibson Dunn’s, not from my own reading of the full Federal Register text.
The yield presumption is gone, and nothing replaced it
In August I described the proposal’s rebuttable presumption: if an affiliate or a third party paid holders a return on a stablecoin, the OCC would presume that it was the yield the law prohibits, and the issuer would have to prove otherwise. That was the provision aimed at arrangements like platform “rewards” on USDC balances.
The final rule drops the presumption. It keeps the statutory ban on paying interest or yield to holders solely for holding the token, and says payments for bona fide services would generally not be prohibited. What it does not do is add a safe harbor. There is nothing explicit for white-label arrangements, small payments, or rewards programs that already exist. Each arrangement will be assessed on its facts.
So the question I expected to be litigated is still open, just in a different shape. Before, the burden sat with the issuer. Now it sits with the regulator, case by case, without a written line anyone can point to in advance.
Reserves: lower cash floor, explicit concentration limit
The 1:1 requirement stays: reserve assets must equal or exceed outstanding tokens at all times. Two details are worth noting.
The share of reserves that must sit in immediately available form (demand deposits or balances at a Federal Reserve Bank) went from 10% in the proposal to 5% in the final rule. The rest can be in the other eligible categories, including Treasuries with 93 days or less to maturity.
Exposure to a single counterparty is capped at 40% of reserves, with exclusions for US government exposure and some custody and repo arrangements. This addresses the specific risk that showed up in March 2023, when part of USDC’s reserves were stuck at Silicon Valley Bank: a fully backed token whose backing sits mostly at one institution is only as good as that institution.
If reserves fall short, the issuer has to stop issuing new tokens. If the shortfall lasts 15 consecutive business days, it has to liquidate and redeem without fees, unless the OCC grants an extension.
Redemption and runs
Redemption has to happen within two business days of a request, for any amount of at least one token. If redemption requests exceed 10% of a token’s outstanding supply in 24 hours, the issuer has to notify the OCC within 24 hours. The rule also adds run risk explicitly as a factor when the OCC evaluates applications.
This is where the regulation treats a stablecoin less like a payment app and more like what it structurally is: a short-term liability that everyone can ask to redeem at the same time.
Capital is separate from reserves
Reserves back the tokens. Capital is what keeps the issuer running. The rule asks for minimum capital of $5 million, plus an operational backstop equal to 12 months of total expenses, of which up to half can be funded with long-term debt. If the issuer misses the capital requirement at two consecutive quarter-ends, it has to stop issuing and wind down.
$5 million is a low number for a company that might have billions outstanding, but the logic is different from a bank’s. The reserves are not supposed to be lent out, so capital is not there to absorb credit losses on them. It is there to pay for an orderly shutdown.
The state path has a ceiling
Nonbank issuers with up to $10 billion outstanding can choose state supervision instead of federal, if the state regime is “substantially similar”. Above $10 billion they have to transition to federal supervision within 360 days, a deadline the OCC cannot extend, unless they get a waiver to stay with the state.
As of the rule’s publication, no state regime has been certified as substantially similar, and the initial certifications are due by January 18, 2028. In practice, for the first year, the “three paths” I described in August are closer to two.
Foreign issuers are waiting on Treasury
A foreign issuer can serve US users only if Treasury finds its home regime comparable to the US one. Treasury has not made any such determination yet. This is directly relevant to Tether, whose USDT is the largest stablecoin and is issued outside the US. Until those determinations exist, the rule describes a process that no one can complete.
What is still missing
The OCC rule is one piece. It explicitly leaves out anti-money laundering and sanctions requirements, which will come in a separate rulemaking. The Federal Reserve, the FDIC and the state regulators each have their own parts to write. And the application forms are still out for comment until November 23, so even issuers who want to apply cannot do it yet.
The August post closed by saying the legal answer to “is this token backed by something real” still had unresolved edges. Some of those edges are now settled with specific numbers: 5%, 40%, two days, $10 billion. The one I expected to matter most, yield paid through someone other than the issuer, was deliberately left without a line.