The Fed Wrote Stablecoin Size Into a Price Tag
A $10 billion stablecoin issuer now needs $200 million in capital parked outside its reserves. Not yielding. Not backing a single token. Just sitting there, as the price of being allowed to operate.
I found that number the way I find most things — by rebuilding someone else's spreadsheet. The Federal Reserve's September 24 proposal on payment stablecoin issuer capital is a thin document wrapped around a thick formula, and I spent a week transcribing its rate structure into a model so I could see where the curve bends. Every conversation since has opened the same way. This is clarity. This is bullish. The gray zone is finally over. Then I show them the curve.
Here's the part nobody screenshots. The rates are marginal, not average. Crossing a threshold doesn't reprice the whole balance sheet. It reprices only the dollars above the line. Which means the largest issuer in the room pays the least per dollar of float, permanently, by construction. We didn't get a stablecoin rulebook. We got a cost function shaped like a moat.
To understand why, you have to be precise about what this document is. It isn't securities law. The GENIUS Act settled that argument — payment stablecoins are payment instruments and quasi-bank liabilities, not investment contracts, and the Howey analysis collapses quickly when there's no profit expectation and a one-to-one redemption promise attached. What we're watching now is prudential regulation, the same family of rules that governs depository institutions. The Fed is building a capital stack for dollar tokens.
Two blueprints exist, and they disagree about how to think. The OCC's March 2 proposal sets a $5 million floor — trivial in this context — then layers bespoke, firm-by-firm capital on top, plus a twelve-month fee-based liquidity pool requirement. Its comment period closed on May 1. The Fed's September 24 proposal does the opposite: it replaces judgment with a rate table. Marginal bands, an income add-on, a credit charge, and one multiplier. Its comment window runs sixty days from publication, and the final text is still open, still mutable, still worth arguing with.
That sixty days is not a formality. In every rulemaking I've tracked since 2017, the comment record is where small operators get heard precisely because large ones are too busy drafting their own submissions. If you run a payment product, a remittance corridor, or a payroll rail that settles in stablecoins, this window is your only formal voice in the architecture of your own cost of doing business. Most people reading this will not file a comment. That asymmetry is exactly how rules get written for the largest participant in the room.
Scope matters as much as substance. The Fed's framework reaches only issuing subsidiaries supervised by the Fed, plus certain state-chartered transitional issuers once they cross $10 billion in circulation. That is a narrower set than the dollar stablecoin market itself. The largest offshore tokens — the ones that clear the most volume in DeFi — sit outside this document entirely, untouched, unburdened, and structurally advantaged by a rule they never had to satisfy.
In 2017, I abandoned a scheduled fiat audit engagement for three months to build a crude proof-of-knowledge demo in ZoKrates, because Vitalik's writing had convinced me that mathematics was becoming the new social contract. I've spent the years since watching that idea collide with the least mathematical institution we have. This proposal is the collision, rendered as a rate schedule.
The formula is where the argument actually lives. Operational risk capital accrues on marginal bands: two percent on float up to $20 billion, 1.5 percent from $20 billion to $50 billion, one percent above that. Then a 25 percent add-on applied to three-year average non-reserve income. Then a credit risk charge of two percent against uninsured deposit claims and undercollateralized reverse repos. Then a loss scalar, which I'll return to, because it's the only number in the whole document that genuinely matters.
I ran the bands. At $1 billion of float, base capital is $20 million. At $10 billion, $200 million. At $20 billion, $400 million — the last marginal dollar still costing two cents. At $50 billion it's $850 million: $400 million from the first band plus $450 million from the second, landing at 1.7 percent of float. Push to $100 billion and you add $500 million at the one percent band, so $1.35 billion total, or 1.35 percent of float.
Read that sequence again. The effective capital rate falls as the issuer grows, from 2.0 percent to 1.35 percent, while every new entrant is forced to live in the most expensive band. That is not a drafting accident. Someone chose marginal brackets over a flat rate, and the consequence is that scale has been converted into a compliance cost advantage. Concentration isn't a side effect of this rule. It is the arithmetic.
Now the add-on, which has received almost no attention. Tying capital to 25 percent of non-reserve income means an issuer earning custody fees, settlement fees, or infrastructure revenue from its own float pays capital on that diversification. Stay a pure spread business and your capital base stays small. Become a payments company and you are charged for the ambition. It's an anti-diversification clause dressed as a risk buffer, and it quietly freezes the stablecoin business model at its narrowest possible version.
The reserve requirement is separate from all of this, and that separation is what most coverage misses. Reserves must be held at fair value at or above the face value of outstanding tokens — one to one, independent of capital. Liquidity isn't solvency. A one-to-one reserve proves you can pay today's redemption; capital proves you survive the week when redemptions and losses arrive together. Regulators now demand both. The industry keeps quoting only the first number, which is why the real burden of this proposal is systematically understated in nearly every headline I've read.
The credit risk charge deserves its own moment, because it's a quiet reserve policy in disguise. Penalizing uninsured deposit claims and undercollateralized reverse repos at two percent pushes issuer reserves toward Treasuries and away from bank deposits. Do that across every compliant issuer and you've nudged the structure of short-end dollar demand, without ever writing the words "reserve composition mandate." That's elegant regulation. It's also a reminder that capital rules are never only about capital.
And then the loss scalar. It's a multiplier the Fed can tune up or down against historical loss experience. Every analyst praising this framework for being formulaic instead of discretionary has missed something simple: the discretion didn't disappear. It moved. The flat bands are the visible layer, the part that fits in a press release. The scalar is the steering wheel.
I know this primitive better than I'd like. Last year I worked with a Chicago-based AI ethics lab on an ethical constraint protocol for autonomous DAO treasuries — a dynamic multiplier on spend limits that governance could retune based on incident history. We learned something painful in that work: a tunable parameter is not a rule. It's a governance surface, and whoever holds the dial holds the policy. The Fed just published the same design in banking vocabulary, and almost nobody has noted the family resemblance.
So here is where I part company with the consensus, which has settled comfortably on "regulatory clarity equals bullish."
Clarity is a cost schedule, and the cost is asymmetric. A compliant issuer now carries capital its offshore competitors don't, while both tokens trade at the same dollar, sit in the same liquidity pools, and are treated as fully interchangeable by every AMM on chain. Liquidity doesn't price the capital stack behind a token. It prices the ticker. Which means the compliance cost cannot be passed to users, because the market can't distinguish the products. It gets absorbed by the issuer, as permanent margin compression.
Identity isn't what a token prints on its whitepaper. It's the balance sheet standing behind it — and the market currently has no mechanism to see that balance sheet, let alone price it. That's the structural disadvantage nobody is modeling. Regulatory moats only function when the market can price the difference between inside and outside. Here, it can't.
There's a second blind spot, and it's larger. The Fed didn't only publish a capital formula on September 24. It also proposed procedures for bank subsidiaries to apply as stablecoin issuers. Put the two documents side by side and the strategy becomes legible: on-chain dollars are being routed into the bank holding company framework. The major beneficiary of American stablecoin regulation is not a crypto company. It's a bank that already has a capital stack, an existing compliance apparatus, and a parent that already knows how to talk to the Fed.
And one more consequence worth saying plainly. If the Fed's formulaic path proves more expensive than the OCC's bespoke one, issuers will choose charters the way they choose jurisdictions. Multi-track supervision inside a single country doesn't eliminate regulatory arbitrage. It relocates it domestically, which is arguably worse — now the competition is between American regulators, and the loser is whichever one prices risk honestly.
None of this is an argument that the proposal is wrong. Capital requirements for entities issuing the settlement layer of crypto finance are overdue, and there's genuine information gain in watching a regulator write down what it believes an operational failure costs. The bear market taught me to read documents like this for survival signals rather than upside, and the survival signal here is real: predictable rules beat ambiguity, even when predictability carries a price.
But watch the scalar. Not the bands, not the headlines, not the press releases. Every other component of this framework is fixed arithmetic. The loss scalar is the one number that decides whether operational risk capital is a rounding error or a tax on existence, and it hasn't been published yet. In the DAO work, the parameter nobody watched was always the one that governed everything.
So the question I keep returning to isn't about percentages. It's about custody of the primitive. If on-chain dollars are going to carry bank-grade capital requirements, who should issue them — an institution that already has a capital stack, or a network that would have to build one while competing against offshore tokens that never will?
What makes a dollar programmable isn't the code. It's the presence of consent — and the terms of that consent are being written right now, in a comment period most of the people who depend on stablecoins will never read.