Two business days.
That is the redemption window the Federal Reserve has proposed for payment stablecoins issued by banks and their subsidiaries. T+2. Two working days from request to settlement.
Now open any mainstream stablecoin dashboard. USDC settles in hours. USDT, with all its opacity, clears the bulk of redemptions inside a single day. The largest issuers already operate at T+0 to T+1. The regulator has authored a floor that sits beneath the private sector's existing ceiling.
That is not an oversight. It is a design decision. And it says more about the intent of this proposal than anything in the coverage that followed the announcement. I spent three weeks in 2022 stress-testing a stablecoin peg mechanism for a liquidation cascade that nobody wanted to model. The lesson from that exercise stayed with me: the binding constraint is never the headline rule. It is the distance between the rule and real-world practice. That distance is where capital either survives or evaporates.
This proposal has a very specific distance. I want to map it.
The Framework Behind the Rule
The proposal lands under the GENIUS Act, which imposes a 120-day approval clock on stablecoin issuers. The Federal Reserve's contribution is narrower than the legislation but structurally heavier. It governs how banks and their subsidiaries may issue payment stablecoins, what those tokens may be backed by, and how holders may exit. A 60-day public comment period is now open. The text is not final. Treat every number below as provisional.
Michael Barr, the Fed governor who has become the public face of the effort, framed it around two pillars. First, universal redemption rights — the holder's ability to redeem at par, elevated from a commercial promise into a regulatory obligation. Second, anti-money-laundering thresholds, which he argued should trigger only at a "material or systemic" level. That second point is a quiet signal. A regulator arguing for a higher enforcement bar is not a regulator looking to choke the asset class.
The reserve list is the part that deserves real scrutiny. Eligible assets, as proposed: US cash, Federal Reserve balances, certain bank deposits, US Treasuries with remaining maturity of 93 days or less, eligible repurchase agreements, and eligible investment funds. Tokenized forms of these assets may be recognized.
Read that list slowly. It is a money-market-fund construction fused to central bank reserves. No commercial paper. No corporate credit. No crypto collateral. The framework does not describe a novel financial instrument. It describes a regulated one. The proposal's central act is not invention — it is the conversion of an industry habit into a legal rule.
I trust the code, not the community. And in this case, the code is a compliance schedule.
What the 93-Day Constraint Actually Does
Start with the maturity ceiling. Ninety-three days. Not 90, not 120. The choice of 93 days maps almost exactly to one calendar quarter plus a buffer, which is the maturity band that money-market funds reserve for "cash equivalent" treatment under Rule 2a-7. This is not a crypto-native parameter. It is imported wholesale from traditional asset management.
The economic consequence is precise. A stablecoin issuer backed by ≤93-day Treasuries holds an asset with near-zero duration risk and near-zero credit risk. It also holds an asset with a fixed and low yield. The entire profitability of the issuance model collapses into a spread between short-term Treasury yields and operating costs.
During my audit work on the 2020 DeFi Summer arbitrage, I built a Python script to monitor Uniswap v2 pools. I found a persistent 0.3% mispricing caused by oracle latency in thinner pools. I executed 142 micro-transactions over three weeks and netted $4,500. The point of that exercise was never the profit. It was the discovery that value in DeFi accrues to whoever understands the constraint that others ignore. Here, the constraint is duration. Once the remaining maturity is capped at 93 days, an issuer loses the ability to earn a term premium. Duration arbitrage — the classic shadow-bank maneuver of funding short and lending long — becomes structurally impossible.
That is the anti-run design in a single number. A stablecoin issuer that cannot take duration risk cannot be squeezed by a rate shock. The 2023 regional banking failures were, in part, duration mismatches dressed in deposit liabilities. This proposal forecloses that failure mode by law.
The trade-off is that the model becomes a utility, not a yield engine. Yield is often the interest paid on risk you didn't correctly price. The Fed has priced that risk at zero and removed the yield with it.
The Capital Tiers and a Missing Band
The capital requirement is where the design gets interesting and where the text gets thin. As proposed: a 2% capital charge on the first $20 billion of issuance, dropping to 1% above $500 billion.
Notice what is absent. The $20 billion to $500 billion band is undefined in the current draft. That gap is not a rounding error. It is where the two largest regulated issuers in the world currently sit. A framework that specifies the floor and the ceiling but leaves the middle unstated is a framework that has not finished its own arithmetic.
The tiering logic, however, is legible. A flat percentage charge would be regressive to small issuers and trivial to large ones. A declining percentage charge rewards scale: the bigger you are, the less capital you must hold per dollar of token. This is an explicit scale-economy subsidy. It accelerates consolidation. In a market where the top two issuers already command the overwhelming majority of supply, a rule that makes size cheaper will make the market more concentrated, not less.
Silence is the most expensive asset in a bubble. Here the silence is a five-hundred-billion-dollar range. Until the final text names it, no issuer can model its own capital cost, and no analyst can model the industry's.
Tokenized Reserves: The Quiet Infrastructure Signal
Buried in the eligible asset list is a line that most coverage skipped: tokenized forms of eligible reserves may be recognized. Read that again. A tokenized Treasury — a BUIDL-class or Ondo-class instrument — may qualify as a reserve asset for a regulated bank-issued stablecoin.
This is the only genuinely forward-leaning technical element in the proposal, and its implications are structural. It creates a compliant rail for on-chain government debt to sit inside the reserve backing of a payment stablecoin. The stablecoin reserves and the tokenized-Treasury sector stop being parallel markets. They become vertically integrated.
The dependency chain looks like this: the Federal Reserve and the GENIUS Act set the rules, the bank subsidiary holds the reserves, and the application layer — payments, DeFi, real-world assets — consumes the settlement token. Tokenized reserves insert themselves as middleware between the issuer and the Treasury market. That is a durable position in the stack.
I led an on-chain verification project in 2026 that cross-referenced satellite imagery against title transfers to reduce RWA fraud. The hardest problem was never the cryptography. It was establishing that the off-chain asset and the on-chain claim were the same object. Tokenized reserves face the same reconciliation problem at a smaller scale — and solving it inside a regulated reserve structure is a proof of concept that the entire RWA sector will watch.
Redemption Rights: From Promise to Obligation
Barr's emphasis on universal redemption rights is the load-bearing wall of the proposal. Today, redemption at par is a commercial commitment. It appears in terms of service and holds until the issuer decides it does not. The proposal would convert it into a regulatory duty, with a reserve shortfall triggering forced liquidation and full redemption.
The distinction matters more than it appears. A commercial promise is enforceable in court after the fact. A regulatory duty is enforceable in real time, with the regulator holding a liquidation trigger. That is a meaningful shift in who bears the residual risk.
The T+2 floor, however, cuts the other way. By writing two business days as the legal maximum, the framework sets a bar that the largest issuers already clear and that smaller or newer issuers will treat as a target. The regulation standardizes at the bottom, not the top. Established players will simply keep their faster internal SLAs. New bank entrants will build to the legal line. The efficient frontier of redemption speed does not move. Only the compliance floor appears.
The AML threshold is the second soft signal. Barr's preference for a "material or systemic" trigger suggests the Fed wants an enforcement posture calibrated to genuine risk rather than broad surveillance. That is friendlier to the industry than many expected, and it may attract internal opposition before the text is finalized.
The Contrarian Read: Banks Are Late, and Concentration Wins
The dominant narrative is that this proposal throws the door open for banks to enter stablecoins and reshapes the competitive landscape in their favor. The narrative is half right, and the half that is wrong is the part markets will trade on.
Banks are entering on a slow clock. GENIUS Act approval runs 120 days. The comment period runs 60. Any bank subsidiary must clear dual federal and state supervision. Meanwhile, the incumbents — Circle, Tether, and the issuers not directly bound by this bank-specific rule — keep operating under their existing regulatory paths. The temporal gap between "a bank may now issue" and "a bank is issuing at scale" is measured in quarters, not weeks. Markets that price the announcement as an adoption event are pricing correlation as causation.
Here is the causal chain that actually holds: the proposal lowers regulatory uncertainty, which raises institutional willingness to hold and settle in stablecoins, which increases reserve demand for short-term Treasuries, which marginally suppresses front-end yields at the margin. That chain is real but slow and small. The chain that markets are pricing — banks dislodge incumbents within a quarter — is neither real nor small, and it is wrong.
The second-order effect is the one worth watching. A capital charge that declines with scale, combined with a reserve regime that favors cheap duration, produces a market where the largest issuers become structurally cheaper to run. If the final text raises the capital charge, it could discourage mid-sized bank entrants entirely and reinforce the very concentration the framework was meant to diversify. Anti-run design and pro-competition design are not the same objective. This proposal achieves the first and only gestures at the second.
We have seen this movie before. In 2021, I analyzed wallet clustering for a profile-picture NFT project and found that roughly 60% of its "community" was wash-trading bots controlled by three wallets. The marketing said community. The data said three keys. The gap between the story and the chain is where retail money goes to die. The story here says bank-led competition. The mechanics say scale-led consolidation. Hold both in your head and trust the mechanics.
What to Watch Next
The proposal is not a technology document. It is a positioning document. It fixes the definition of a stablecoin as a full-reserve money-market instrument, locks that definition into federal rule, and builds a compliant on-ramp for banks while leaving the incumbents' fate to be decided by the GENIUS Act's division of labor.
The signals that will determine whether the medium-term reshuffling becomes reality are narrow and specific. Watch the Federal Register for whether the $20 billion to $500 billion capital band is defined — that single number decides whether mid-market issuers can exist. Watch for the first complete bank subsidiary application; its submission date will tell you whether the 120-day clock is real or ceremonial. Watch the final AML language, because "material or systemic" versus a lower bar changes the compliance cost curve for everyone. And watch whether the tokenized-reserve provision survives the comment period, because that clause is the one thing in this proposal that builds rather than constrains.
Two business days is the floor. The ceiling is still being written in a comment file nobody is reading yet. That is where the real rule lives.