The Fed priced a stablecoin dollar. It gets cheaper above $50B.

The Federal Reserve has put a price on issuing a stablecoin dollar, and the price falls as you get bigger. On 24 September the Board asked for comment on two proposals under the GENIUS Act. The long one, Docket R-1899, runs 392 pages and carries the formula. An issuer the Fed supervises would hold capital equal to 2% of the first $20 billion of coins outstanding, 1.5% of the next $30 billion, and 1% of everything above $50 billion.

Read that as a cost per dollar issued. The first billion costs 2 cents of capital for each dollar of coin. The hundredth billion costs 1 cent. At exactly $50 billion outstanding an issuer owes $850 million, a figure the Fed prints in its own table. Run the same arithmetic at the size of USDC, $75.3 billion on 26 September 2026, and the issuance charge comes to about $1.10 billion, or 1.47% of coins outstanding. A $5 billion issuer pays the full 2%.

The Fed has a reason for the slope and states it. It also prints the alternative it considered, a flat 1% or 2%, and says what that version would fix: it "would ensure that increases in stablecoin issuance receive equal marginal operational risk capital requirements regardless of the scale of the firm." That sentence describes, by subtraction, what the proposal in front of you does.

What the two notices actually contain

Docket R-1899 is the substance. Reserves have to match the par value of outstanding coins one for one, and the eligible list is short: cash, balances at a Reserve Bank, eligible deposits, Treasury bills with 93 days or less left to run, and certain repo and reverse repo. Redemption has to clear within two business days. A newly approved issuer holds at least $5 million of capital through a three-year de novo period, indexed to nominal GDP.

Docket R-1900 is the paperwork. A Board-supervised bank that wants to issue files an application, the Board has 30 days to say whether the file is substantially complete, and 120 days from the submission date to decide. Both clocks come from the GENIUS Act rather than from the Fed, and so does the penalty for missing the second one. An application the Board fails to rule on inside 120 days is "deemed approved."

Neither notice is a rule. Comments are due 60 days after publication in the Federal Register, and as of 26 September the documents still carried a bracketed placeholder where that date belongs.

The curve, priced per dollar

Those are calculations, not quoted figures. They come from applying the Fed's three tiers to five issuer sizes, and they leave out a second component of the operational-risk charge: 25% of the three-year average of annual revenue from non-reserve assets. Revenue earned on the reserves themselves is excluded, which for most issuers is nearly all the revenue there is.

The average matters less than the margin. An issuer below $20 billion posts 2 cents of capital to put one more dollar of coin into the world. An issuer past $50 billion posts 1 cent. Same peg, same redemption promise, same two-day window, half the marginal cost. USDT and USDC together hold 84.9% of the $305.2 billion outstanding as of 26 September. A schedule that charges the challenger double at the margin does not read as neutral on that number.

The 2% that argues reserves out of banks

Section 247.17(a) stacks credit-risk charges on top of the issuance charge. There are two, both set at 2%. One applies to uninsured eligible deposit claims held as reserves. The other applies to the undercollateralized slice of a reverse repurchase agreement, measured after haircuts. Treasury bills inside the 93-day window draw neither.

Follow the incentive. Hold $1 billion of reserves as uninsured bank deposits and you carry $20 million of capital for it. Hold the same billion in short bills and that line is zero. The Fed is open about the case history: the notice recounts that in March 2023 Circle held roughly $3.3 billion of USDC reserves as uninsured deposit claims against Silicon Valley Bank, and the coin broke its peg. The charge is calibrated against the 20% risk weight banks already carry on the same exposure.

The consequence runs past stablecoins. Every regulated issuer that grows moves its reserves toward the front end of the Treasury curve and away from bank deposits. The bank lobby spent two years arguing that stablecoins would drain deposits. This rule, written by the banks' own supervisor, makes the cheapest reserve a Treasury bill.

The clause that reaches your exchange account

The GENIUS Act already bars an issuer from paying interest or yield to a holder for simply holding the coin. The Fed proposes to close the obvious way around it. Under proposed section 247.10(c)(4)(i), an issuer is presumed to be paying prohibited yield when it has an arrangement to pay an affiliate or a related third party, and that party in turn pays coin holders for holding. A "related third party" is defined to include any person "offering to pay interest or yield to payment stablecoin holders as a service." The issuer can rebut the presumption in writing.

That structure describes most stablecoin reward programs you can sign up for today. The issuer does not pay you; a platform pays you, out of economics the issuer shares with it. The Fed says it is following the approach the OCC proposed in March, so the presumption is not a Fed invention. What changed is that it now sits in the rulebook governing the banks queuing up to issue.

If you hold a stablecoin for the reward rate rather than for payments, that is the paragraph to read. Yield on crypto collateral is legal in other shapes with different rules, and what Ethereum staking pays after fees is a separate calculation.

Almost nobody large is covered by this

Here is what the headlines skipped. The proposal reaches Board-supervised permitted payment stablecoin issuers, meaning subsidiaries of state member banks and a short list of related entities. Circle, which issues USDC, sits with the OCC: it took final approval for a national trust bank charter on 10 July 2026. USAT, Tether's US coin, is issued by Anchorage Digital Bank, also OCC-chartered. Offshore USDT is not a US-regulated issuer at all.

And the OCC has not written a formula. Its proposed rule from 2 March 2026 sets a $5 million floor for the de novo period, then asks each issuer to calculate its own minimum "based on an evaluation of the risks associated with its business model and risk profile," reviewed through examination. The text is explicit: the OCC is "not currently proposing any floors on the minimum capital requirement or frameworks for determining a minimum capital requirement for those risks in the rule text."

One statute, two supervisors, two prices. A graduated schedule at the Fed and a supervisory judgment call at the OCC. Which licence you hold decides which one you pay, which is the pattern that keeps repeating in tokenized finance: the registration is the moat, not the rail.

The case for the slope, and where it thins out

The Fed does not present the graduated schedule as a courtesy to large issuers. Its argument is that operational risk grows more slowly than issuance, and it works the example. Take unauthorized minting: a thief who mints a large block of a stablecoin has to swap it for something else before the issuer can freeze or burn it, and the bigger the block, the harder that swap becomes. On that reading a dollar of coin at a $100 billion issuer genuinely carries less operational risk than a dollar at a $2 billion one.

The reasoning holds for the risk it names. It covers theft, exploits, and process failure. It says little about the event that has actually killed stablecoins, which is a run, and runs do not get easier to survive as you scale. The proposal answers run risk elsewhere, through the reserve list and the two-day redemption window, and that is the honest response. The capital schedule still prices one risk and stays quiet about the other, while the shape of the schedule falls on competition.

The alternatives section shows the Fed knows the slope is arguable. It asks whether a flat 1% or 2% would serve better, and it is the Fed, not a critic, that wrote down what the flat version would restore.

What to watch

Federal Register publication. The 60-day comment clock has not started. When it does, the comment letters on the tier structure will show which institutions want the curve and which want it flattened.

The OCC's final rule. If the OCC adopts a formula resembling the Fed's, the same price applies to Circle and to everyone chartered alongside it. If it keeps case-by-case capital, the gap between the two charters becomes a standing arbitrage that any applicant can see.

The $10 billion line. A covered issuer that is an uninsured state-chartered depository institution and crosses $10 billion of consolidated outstanding issuance has 360 days to move into the Fed's framework or stop issuing on a net basis until it falls back under. That threshold, more than the capital table, decides how many issuers the Fed ever supervises.

Sources

• Federal Reserve Board press release, 24 September 2026
• Implementing the Federal Reserve Board's Responsibilities under the GENIUS Act, Docket No. R-1899 (392 pages)
• Application Procedures for Board-Supervised Insured Depository Institutions, Docket No. R-1900
• OCC proposed rule implementing the GENIUS Act, 91 FR 10202, 2 March 2026
• Circle receives final OCC approval to establish a national trust bank, 10 July 2026
• Stablecoin market capitalization tracker, 26 September 2026
• Unchained on the Fed's yield presumption, 25 September 2026

This is not financial advice.