There is a two percent buried in the Federal Reserve's September 24 capital proposal for payment stablecoin issuers, and nearly everyone has read it as a flat rate. It is not. It is marginal โ which means the first billion dollars a stablecoin ever issues is the most expensive billion it will ever issue, and the hundredth billion is the cheapest.
At a one-billion-dollar float, the rule demands twenty million dollars of operational risk capital. At ten billion, two hundred million. The ratio appears constant. Cross twenty billion in circulation and the rate applied to each new dollar drops to 1.5 percent. Cross fifty billion and it drops to one. A hundred-billion-dollar issuer carries an effective rate near 1.35 percent; a one-billion-dollar issuer carries the full two. The formula does not price risk. It prices scale.
I spent part of 2017 inside MakerDAO's earliest stability fee contracts, hunting for exactly this category of mistake โ a calculation that treats a marginal rate as an average one and quietly eats the solvency of whoever trusts it. I found one. The team fixed it, anonymously, and never learned my name. That was the week I stopped assuming decentralization produces accountability on its own, and started reading capital formulas the way I read poetry: for what the meter conceals.
The GENIUS Act, signed in 2025, gave the United States its first statutory framework for payment stablecoins โ tokens redeemable one-for-one at par, used for settlement rather than speculation. What it did not give was arithmetic. That is what the September 24 Federal Reserve proposal supplies: a rule-based capital formula, opened to a sixty-day comment period, sitting alongside an OCC proposal published March 2 whose comment window closed on May 1. Two agencies, two entirely different theories of how to make a stablecoin issuer safe, and no obvious mechanism yet for reconciling them.
The Fed's scope is deliberately narrow. It covers issuing subsidiaries of insured depository institutions supervised by the Fed, plus certain state-chartered transitional issuers above a ten-billion-dollar threshold. The reserve requirement sits separately and independently: reserve assets at fair value must equal or exceed the face value of outstanding tokens. Capital is not reserve. Reserve is not capital. Hold that distinction; it does most of the work in this proposal.
The capital stack has four layers, and only the first is widely quoted.
First, the benchmark operational risk capital, tiered by float: two percent on the first twenty billion, 1.5 percent on the next thirty, one percent above fifty. Because it is marginal, the blended rate slides from 2.0 percent at ten billion to roughly 1.7 percent at fifty billion and about 1.35 percent at a hundred billion. Scale is rewarded twice โ once through fixed-cost dilution, and again through the formula itself.
Read those brackets as a competitive landscape and the picture turns uncomfortable for the middle. A five-billion-dollar issuer pays the same rate as a five-hundred-million-dollar one, full two percent on every dollar, while an incumbent at fifty billion pays an effective 1.7 percent and shrinks toward 1.35 as it grows. There is no tier here that rewards a small issuer for being careful. There is only a tier that rewards a large one for being large.
Second, a loss scalar. This is a multiplier the supervisor can adjust upward or downward against historical loss experience. Truth emerges when the ledger is transparent, and here the ledger is a formula โ which is precisely why the scalar stands out. A discretionary lever inside a rule-based regime is where the real policy lives. Everything else is furniture.
Third, a twenty-five percent add-on on three-year average non-reserve revenue. Read that clause twice. An issuer that earns custody fees, settlement fees, or transaction revenue is charged capital for the privilege of earning it. The rule nominally guards against operational failure and functionally penalizes operational diversification. The prudent issuer, under this arithmetic, is the narrow one: hold Treasuries, collect the spread, do nothing else.
Fourth, a credit risk capital charge of two percent on uninsured deposit claims and undercollateralized reverse repo positions. Reserve composition is no longer a treasury decision; it is a capital decision.
Compare the OCC's approach and the divergence sharpens. The OCC proposes a five-million-dollar floor, bespoke capital set per institution, and a twelve-month fee-funded liquidity pool. The Fed proposes a formula. One is predictable and inflexible; the other is flexible and unpredictable. An issuer can, within limits, choose which regime governs it. That is regulatory arbitrage wearing a compliance badge.
Here is where the widely quoted two percent understates the burden. Capital and reserves are independent. Run the numbers with that separation in mind. A ten-billion-dollar issuer holds ten billion in reserve assets โ Treasuries, bank deposits, reverse repos โ and simultaneously funds two hundred million in equity capital. The token float can be redeemed at par at any moment; the capital cannot be redeemed at all without triggering a supervisory event. The two pools have opposite liquidity profiles and identical labels on a balance sheet.
I have audited fifty post-mortems of failed protocols, and the recurring failure mode is never a single requirement. It is two requirements that look like one. Founders budget for the visible obligation and discover the invisible one during a stress event, when the two cannot be substituted and both come due on the same afternoon.
The dominant reading of this proposal is that regulatory clarity is bullish. I think that reading mistakes mood for mechanism. A formula that charges two percent on your first billion and one percent on your hundredth billion is not clarity โ it is a moat, drafted in federal language and ratified by the comment process. It does not disadvantage incumbents. It encodes their advantage.
I watched Europe deliver its own version of clarity. MiCA arrived with reserve requirements and compliance costs that large issuers absorbed without blinking and small ones simply could not. The pattern is not a coincidence of jurisdiction. It is arithmetic, and arithmetic travels.
And note who is not in the room. The Fed's scope covers a narrow slice of the dollar stablecoin float. Issuers outside Fed-supervised subsidiary structures โ including several of the largest and most liquid โ face none of this directly. Compliance becomes a cost borne by the regulated, and a cost advantage enjoyed by the unregulated, which is a strange way to design a safety regime.
Watch three things, not the headline rate. Watch the final calibration of the loss scalar, because that multiplier decides more than the brackets do. Watch whether the Fed and the OCC reconcile their methods, because parallel rules manufacture parallel incentives. And watch the subsidiary application pipeline, because the proposal's quiet beneficiary is the traditional bank that now holds a legal path to issuance.
Openness is not a feature; it is a philosophy. But a philosophy that cannot survive its own capital formula is merely a brochure. Humanity remains the only non-fungible asset โ and the formulas we write around it will decide, quietly, who gets to stay.