Hook
On a Friday afternoon in late September, the Federal Reserve published a notice of proposed rulemaking covering payment stablecoins. It was the fifth and final federal agency to do so. It was also, by the calendar the GENIUS Act established, roughly ten weeks past the statutory date for getting it done.
That second sentence is the story. Not the reserves. Not the capital calibration. Not the custodian rules.
I want to be careful here, because the instinct of anyone with an editor-in-chief title attached to their name is to sound certain. I am not. I could not independently verify the September 24 publication date against a primary Federal Register entry, and neither, as far as I can tell, could most of the people who wrote about it that week. What I can verify is the structure: a statute that sets a rulemaking deadline one year after signing and an enforcement date eighteen months after signing, five agencies with overlapping jurisdiction, and an administrative record in which not one of them met the first date.
An NPRM is not a rule. It is a draft with an invitation to argue. Everything that follows depends on holding that distinction, because the market did not.
Context
The GENIUS Act gave payment stablecoins a federal frame, and it set two clocks. The first clock governs rulemaking — the agencies had a fixed date by which implementing rules had to move. The second clock governs enforcement: from January 18, 2027, issuing a payment stablecoin in the United States without authorization becomes unlawful.
Those two dates are doing different jobs. The first is a process requirement. The second is a liability trigger. When the second arrives before the first is satisfied, you do not get certainty. You get a statute pointing at a definition that does not exist yet.
Five bodies share the work. The OCC moved first, in February. The FDIC followed in April, the NCUA in May, Treasury in August. The Federal Reserve, which supervises state member banks, bank holding companies, and parts of the foreign banking organization universe, went last.
There is a reason the sequencing matters beyond optics. The Fed supervises the largest banks in the country. When it defines what counts as a permitted stablecoin activity for institutions it oversees, it is not writing a rule for the current $307 billion stablecoin market. It is writing the track for a future in which deposit-like instruments migrate onto public rails. That is a different ambition than the headline suggests, and it is worth keeping in view.
One more clarification that gets lost in market commentary. A payment stablecoin — par redemption, used for settlement — is not the same instrument as an algorithmic stablecoin or a yield-bearing one. They sit in different quadrants on securities analysis and receive completely different regulatory treatment. This file is about the first category only, and the full-reserve requirement actually reinforces its non-security character. Mixing the categories is how bad takes get published.
I have a background habit that is hard to break. In 2017, before I was doing any of this publicly, I spent six months manually auditing time-crowdsale contracts for three mid-tier ICOs out of Warsaw. I found reentrancy vulnerabilities in two of them — the kind that let an attacker re-enter a withdrawal function before the balance updated. That work taught me something that has nothing to do with Solidity: the document you are handed is never the system that runs. The whitepaper is not the contract. The contract is not the deployment. And the proposed rule is not the rule.
So I read the two proposals the Fed put forward as engineering constraints rather than as policy gestures.
Core
The NPRM contains two distinct proposals, and treating them as one document is the first analytical mistake.
Proposal one is a stability mechanism. It has four constraints. First, full reserve backing, limited to qualifying reserve assets — short-term Treasury bills and, in language that does more work than it appears to, "other high-quality, liquid instruments." Second, standardized capital requirements calibrated along two axes: credit risk and operational risk. Third, rules binding the third parties that custody the backing assets. Fourth, an explicit statement that stablecoin activity falls within permitted scope for Fed-supervised banks — moving it out of a gray zone and into defined permission.
Proposal two is an access gate. It establishes a bespoke application process: business plan, financial information, and a formal structure for appeals, hearings, and adjudication. I would call that a structured entrance rather than a formality.
Put those four constraints together and a familiar shape appears. Full reserve backing. Par redemption. Capital buffers. Custody rules. That is not a novel framework. That is the regulatory grammar of a money market fund — specifically the asset quality, maturity, liquidity, and diversification architecture that Rule 2a-7 imposed on MMFs after 2010.
The Fed is not inventing a new asset class. It is converting stablecoins into supervised money market funds that happen to settle in seconds. That reframing is the single most useful lens for reading the rest of the document, because it tells you where the risk goes. It does not disappear. It moves from code to duration and liquidity management.
Which brings us to Michael Barr.
His accompanying statement is the densest document in the package. He supports the direction — calls it a step in the right direction — and then files three reservations that read like a risk engineer's punch list.
The first is interest rate risk. Reserves concentrated in short-term Treasuries still carry duration. If redemption is at par and on demand while the asset side moves with rates, you have a mismatch. Barr says so directly, and adds the sentence that should be printed on every stablecoin dashboard: even government debt that is liquid in other respects can come under pressure in a stressed market. That is not hypothetical. March 2020 in the MMF complex, the UK LDI episode in 2022, and the USDC depeg in March 2023 are three separate demonstrations of the same mechanism.
The second is foreign exchange risk. The proposal does not define the currency composition of reserves. That is a hole, not a detail.
The third is the anti-money-laundering standard. The rule applies a "significant or systemic" threshold to the verification of an issuer's compliance program. Barr flags that the impact of that standard on the Board's ability to confirm an institution can build and maintain a program is, in his word, unknown.
Read those three together and you have the full map of where the final rule is most likely to diverge from the proposal. Duration limits. Currency constraints. A layered or softened compliance gate. None of these are structural rejections. All of them are precisely the places where a draft becomes a different document.
There is a reverse reading of the AML threshold that deserves more attention than it gets. If the standard is set strictly, it functions as an admission filter — it screens out smaller applicants that cannot fund the compliance apparatus. If it is set loosely, it loses its risk-control function. A regulator's own uncertainty about a threshold is not a drafting quirk. It is a signal that the gate has not been sized yet.
Now the variable I would watch above all others: redemption.
The proposal gestures at a universal redemption right and asks whether it should be made explicit. That single clause is a fork in the road.
If the final text specifies unconditional, immediate, par-value redemption, then in a stress scenario the issuer absorbs the entire liquidity mismatch. It sits on the wrong side of exactly the exposure that took USDC off peg in 2023, when Circle's reserves included a slice of the banking system that failed over a weekend.
If the final text instead specifies conditional redemption — a fee gate, a delay mechanism, a queue — then the instrument's core promise erodes. A stablecoin that can decline to redeem on demand is a stablecoin with an asterisk.
In either direction, the redemption clause is the highest-weight line in the entire rulemaking. I have watched this failure mode up close. During the Terra collapse in 2022, I ran a fact-checking effort across a 10,000-member community for three weeks, verifying on-chain data to slow the panic. What I learned there was not about algorithms. It was about contracts: the moment holders discover that a redemption promise has conditions, the conditions get tested all at once.
The second-most important undefined term is quieter. "Other high-quality, liquid instruments" is where the economic value of the rule lives. The width of that definition determines how much yield an issuer can earn on reserves and how much tail risk it carries. It is the highest-value and least transparent phrase in the document, and I would not be surprised if it is also the most heavily lobbied.
Then there is a concentration point that almost nobody is discussing. If the custodian rules narrow eligible custodians to a small set of qualified institutions, multiple issuers end up sharing the same few back-ends. That is a single point of failure at the reserve layer — not a code bug, a counterparty bug. Low probability, meaningful impact, and exactly the kind of thing that looks obvious in hindsight.
There is one more number worth sitting with. Tether holds roughly 60 percent of the $307 billion market. Circle holds about 24 percent. The top two issuers are 84 percent of the thing being regulated. The largest of them is, structurally, not a Fed-supervised entity, and its reserve and legal structure has not obviously been built to the qualifying-asset definition. I flag that as inference, not fact — the composition data is not in front of me.
And a point that gets skipped entirely: the stablecoin token itself has no value capture. It redeems at par. There is no upside. The value accrues to issuer equity, to distribution channels, and to on-chain settlement and custody infrastructure. Rule clarity is a catalyst for issuer equity and a neutral event for the token. Anyone trading the token as a regulatory play is trading the wrong instrument.
Contrarian
Here is where I part company with most of the coverage.
The dominant frame is that the Fed placed the last piece of the puzzle. That frame is wrong, or at least imprecise in a way that matters for anyone positioning around it.
The Fed's actual jurisdiction is narrower than the puzzle metaphor implies. It supervises state member banks, bank holding companies, and portions of the foreign banking organization landscape. Much of the early stablecoin activity in the United States ran through the OCC charter system instead. So a Fed NPRM directly covers bank-side issuers — a category that currently holds close to zero market share — while the two largest issuers sit largely outside its direct reach. Institutional completeness is not the same thing as market coverage. Those are different claims, and the headline fused them.
The second blind spot is the direction of causality. Over the past decade the pattern was consistent: the industry moved, the regulators chased, the rules arrived years later and reshaped a market that had already formed. This time it is inverted.
SoFi and Mastercard are already running live real-time stablecoin settlement. Visa took a founding validator seat on a settlement network. Circle committed $100 million of equity into Binance while the compliance regime was still unfinalized. Every one of those moves happened in a rule vacuum.
The infrastructure was built before the rule, not after it. That inverts the usual sequence, and it compresses the regulator's room to maneuver. When payment networks and banks have already shipped, the rule has to accommodate a fact pattern rather than author one. I have argued for a while that this is the structural signature of the current cycle, and the stablecoin file is now its clearest example.
The third thing the "last piece" frame obscures is scale. $307 billion against the US deposit base is a rounding error. The Fed supervises the largest banks in the country and is defining what permitted issuance looks like for institutions holding a meaningful fraction of American deposits. The purpose is not to regulate a $307 billion market. It is to lay track for a migration that could be measured in trillions. That is a strategically different project, and it is the one worth pricing.
A fourth thing, smaller but sharp: nobody met the deadline. Every agency that was supposed to finalize by July 18, 2026 did not. The Fed was last to even propose. The 60-day comment period has not opened. By the arithmetic in the source material, roughly seven weeks remain for finalization.
A seven-week finalization window against a 60-day comment period cannot be satisfied by the ordinary machinery of US administrative rulemaking. Something has to give. Either the process compresses and public comment becomes ceremonial, or the timeline slips. There is no third option I find credible.
And a fifth, which almost no one has raised. If the final rule does not clarify whether an AML obligation extends to on-chain counterparties, compliant stablecoins may become unusable inside permissionless DeFi. That splits liquidity into a compliant pool and a permissionless pool — a fragmentation risk that is invisible until it is structural. Meanwhile, if US standards land meaningfully stricter than Europe's MiCA regime, issuers will simply optimize their domicile. Regulatory arbitrage does not need permission. It only needs a cheaper jurisdiction.
Takeaway
Sit with the two possibilities.
If the rule is finalized on time, issuers with existing compliance plumbing — OCC-charter experience, monthly attestation habits, auditable reserve accounting — get a real first-mover advantage. Circle's position gets reinforced. Bank-side issuance opens from zero, and that is a change the market has not yet priced.
If the rule is not finalized on time, January 18, 2027 arrives with an enforcement clause and no detail to enforce it against. The statute says unpermitted issuance is unlawful, and the definition of permitted is still in draft. The likely patch is a temporary exemption or transitional guidance, which is not the same as certainty. It is a sign that the calendar won before the text did.
The highest-certainty beneficiaries either way are not the issuers. They are the layer underneath: custody, reserve accounting and attestation, AML monitoring, capital measurement tooling. Whatever the final language says, full reserve backing and compliance verification produce rigid demand for third-party services. That layer carries the least rule-content risk and the most demand certainty in the entire file.
Truth is often buried under the noise. This week the noise was "the last piece is in place." The truth is that the pieces are drafted, the clock is short, and the man who will sign the final version told you in writing that it will probably look different.
Silence speaks louder than hype. The five agencies that missed a statutory deadline said nothing about it. That silence is the most informative datapoint in the entire rulemaking — more informative than any of the reserve definitions, because it tells you how the machine actually performs under load.
The seven weeks are the whole question.
