Hook
On September 4, 2026, two staff economists at the Federal Reserve Board published a FEDS Note on stablecoin measurement. Kristen Payne and Mary-Frances Styczynski did not announce a policy. They did not propose a rule. They published a methodology, and inside that methodology sits one sentence that should end a certain kind of conversation permanently: stablecoin reserves โ bank deposits, Treasury bills, government money market fund shares โ are already counted inside M1 and M2. Add the token's face value on top and you have not measured money. You have measured it twice.
That is the whole game. Everything else in the note is scaffolding erected around a single arithmetic constraint. The market spent the following week debating whether the Fed had "endorsed" stablecoins. The market was debating the wrong document.
I have spent thirteen years watching this industry mistake narrative for structure. The stablecoin conversation has been narrative for a decade. It is now arithmetic. Arithmetic is less flattering, and it is considerably harder to market.
Context
H.6 is the Federal Reserve's money stock release. By download volume it is the second most accessed dataset on FRED โ a detail that matters more than it appears, because it means the accountability surface is public, permanent, and revisable in the open. M1 currently stands at roughly $19.9 trillion. M2 at roughly $23.2 trillion. These are the denominators against which any stablecoin number will eventually be judged, and they are large enough that the numerator has to be argued about carefully.
Payne and Styczynski are explicit that the note does not change how H.6 is published today. It establishes a conceptual basis. Commentators read that as hedging. It is not hedging. It is correct sequencing: a statistical agency defines classification logic before it commits a reporting line, because a reporting line is expensive, permanent, and auditable by Congress.
Three institutional tracks are running in parallel. The GENIUS Act legislates stablecoin issuance and, in Section 4(a)(11), prohibits direct payment of interest to holders. Its execution date is January 18, 2027. The Office of the Comptroller of the Currency, under Comptroller Jonathan Gould, has publicly committed to finalizing implementing rules by November 2026. The New York Fed, via Athreya, has produced parallel research on deposit outflow. And the Board's own staff have now written, on the record, that they cannot currently separate tokenized deposits from ordinary deposits in the data they collect.
That last admission is the most honest sentence in the document. It is also, functionally, a procurement request. When a central bank publishes its own capability gap, it is describing the shape of the system it intends to build next.
Core
The functional test
The proposed method is functional rather than legal. A stablecoin is classified by what it does, not by what it is called. If it functions as a daily transaction medium, it belongs in M1. If it functions as a store of value, or as inventory held for crypto trading, it belongs in the non-M1 portion of M2.
This mirrors the Board's 2020 reclassification of savings deposits, when the distinction between transaction accounts and savings accounts was collapsed on the grounds that the behavioral difference had evaporated. The precedent is instructive. The Board does not classify instruments by their labels. It classifies them by their usage. Usage is observable. Labels are marketing.
There is a discretionary problem here, and I want to flag it before anyone sells a research note claiming certainty. "Daily transaction medium" is not a bright line. A stablecoin used for payroll settlement four times a month and posted as collateral the remainder of the time sits in a gray zone the note does not resolve. It creates a category and leaves the boundary unpoliced. Boundaries that go unpoliced get litigated later, and litigation is a cost that lands asymmetrically on smaller issuers.
From my 2017 work auditing more than forty ICO whitepapers โ the year I rejected an Ethereum-based project with a centralization flaw in its multisig wallet structure, and was told I was missing a thousand-x โ I learned that ambiguity in a classification system is always where value leaks. The projects that survived were not the ones with the best narrative. They were the ones whose token had one unambiguous function. Ambiguity is a subsidy paid to whoever can afford the best lawyer.
The double-counting knot
Here is the mechanics, stated plainly.
Suppose a stablecoin issuer holds $100 of reserves as a bank deposit. That $100 is already inside M2. It may also be inside M1, depending on account type. The issuer then mints $100 of tokens. If the Board adds $100 to the money stock for the tokens, total measured money rises by $100 while zero new money has been created. The $100 was measured before the token existed and is measured again after it.
This is not a rounding error. It is a category error, and the staff acknowledge it in print. The note states that the underlying assets are typically held in forms already included in existing aggregates: bank deposits, Treasury securities, government money market fund shares. Three buckets, all already inside the perimeter. The token is a wrapper around money that has already been counted.
Scale is the question everyone skips. Assume stablecoins reach $2 trillion โ aggressive, but not absurd within this decade. Assume an overlap rate of 80 percent. You are looking at $1.6 trillion of phantom money under naive counting. That is 8 percent of M1. That is enough to distort a policy discussion, enough to make a rate decision look wrong in retrospect, and enough to embarrass an institution that publishes on a fixed calendar.
The deduplication problem is not an accounting nicety. It is a precondition for the aggregate being usable at all. And the Board does not currently possess the granularity to solve it, because tokenized deposits are not separately tracked in the reporting it receives.
I built a structurally similar model in August 2020, simulating Compound's interest rate curves in Python on a laptop in Rome, and identified the liquidity crunch that appears when ETH collateralization drops below 150 percent. The lesson there and the lesson here are identical. The failure mode is never in the headline rate. It is in the composition of the balance sheet underneath it. Total value locked tells you nothing about what happens when composition turns.
The interest ban does the classifying
Section 4(a)(11) of the GENIUS Act prohibits paying interest directly on stablecoins. Most commentary filed this under consumer protection. Read it as a classification mechanism instead.
The non-M1 portion of M2 is largely defined by yield. Savings deposits, money market fund shares, small time deposits โ what they share is that holders receive compensation for parking funds. Remove the yield and the behavioral justification for placing an instrument outside M1 disappears with it.
A stablecoin that cannot pay interest is not a savings instrument. It is a transaction balance. Transaction balances belong in M1. The note says this without saying it: a stablecoin barred from paying interest is functionally a checking account, and checking accounts are M1.
Now observe what has occurred. The legislature banned the yield. The central bank then classified the resulting instrument as M1. Two institutions with different mandates, different constituencies, and different institutional incentives arrived at the same structural conclusion. That is not coordination in the conspiratorial sense. It is convergence, which is more durable, because convergence survives a change in personnel.
A stablecoin barred from paying interest is not a deposit substitute with a yield advantage. It is a checking account with a settlement advantage. Those are different products with different economics, different marketing, and different regulatory exposure. The industry has spent five years selling the first and will now operate the second.
Incrementality
The question the market keeps asking is whether stablecoins expand the money supply. It is the wrong question, and it is wrong in a way that flatters the sector. The right question is whether they change how money moves.
On current evidence, they do not create money. They repackage it. A dollar in a bank account becomes a dollar in a token redeemable for a dollar in a bank account. The aggregate is unchanged in substance. What changes is the settlement layer, the velocity, the programmability, and the counterparty.
Measured against a $19.9 trillion M1 base, the current float is a rounding artifact. Even at aggressive projections, the perturbation is small relative to the measurement error the Board already tolerates in its own aggregates. Quantity is not the story.
The signal value of this framework exceeds its quantity value by at least an order of magnitude. What the Board is communicating is not that stablecoins are large. It is that stablecoins are permanent enough to require a counting rule. That is a different sentence, and a more consequential one, because statistical permanence is a form of institutional permanence โ and institutional permanence is precisely what the previous decade could not deliver.
Volatility is the tax on unproven consensus. This note retires a piece of consensus โ the proposition that stablecoins sit outside the monetary perimeter โ and replaces it with a measurement question. Measurement questions are less volatile and more durable. They are also considerably harder to trade.
Where value actually accrues
If stablecoins cannot pay interest, and if they sit in M1 as transaction balances, the issuer's economics collapse into a single line: the spread between what reserve assets earn and what the issuer pays holders. The issuer pays holders nothing. Therefore the issuer captures the entire reserve yield.
That makes the business identical to a money market fund's, minus the pass-through. In a high-rate environment this is extraordinarily profitable. In a declining-rate environment it shrinks with no offsetting fee line, because the product's differentiator โ settlement speed โ is not something the market has historically paid for at scale. Nobody has yet shown a willingness to pay a basis point for faster finality.
I want to be precise, because this is a common modeling error. The stablecoin is not an investment product. The issuer is an investment product. The token is a settlement instrument sitting on top of a carry trade. Anyone modeling the token as the asset has misidentified the entity generating the return.
The 2024 ETF experience taught me the same lesson from the opposite direction. I built a basis trade between Bitcoin futures and spot across three venues, ran $5 million at a 2.5 percent annualized premium, and returned 4.2 percent in three months while the market went nowhere. The return did not come from being correct about Bitcoin. It came from correctly identifying which instrument carried the exposure and which instrument merely sat on top of it. Most participants confuse the wrapper with the risk, and the wrapper is always the cheaper thing to be wrong about.
Tokenized deposits are the competitor nobody is pricing
This is the part of the note that is being under-read, and I suspect deliberately so.
The staff flag that stablecoin reserves are typically held in bank deposits, Treasuries, and government money market funds. They then flag that tokenized deposits โ the on-chain representation of a bank liability โ are not separately tracked, which complicates deduplication.
Read it in reverse. The Board is describing a data gap that, once closed, produces a new subcategory. Tokenized deposits and non-bank stablecoins will sit in the same statistical neighborhood, competing for the same settlement volume, and the Board will possess the granularity to observe which one is winning.
Now consider incentives. A tokenized deposit is a bank liability. It sits on a bank balance sheet. It exists inside the existing regulatory perimeter, inside the deposit insurance framework if structured accordingly, and inside an existing supervisory relationship with an examiner who already has a phone number for the treasurer. A non-bank stablecoin has none of that.
If you are a bank, and the central bank is building measurement infrastructure specifically capable of distinguishing the two, your rational strategy is to lobby for the distinction to matter. Once it matters in the statistics, it matters in the rulemaking. Once it matters in the rulemaking, the compliance cost asymmetry does the rest. That is not a prediction about intentions. It is a description of how regulated industries behave when given a category advantage.
The Fed's deduplication problem and the banking lobby's competitive problem have the same solution, which is why I expect the solution to arrive. That is the mechanism. It is not a conspiracy. It is an alignment of administrative need and commercial interest, which is historically the most reliable generator of policy outcomes that exist.
The transmission map
The most direct channel is competition for deposits. If a stablecoin is classified as M1 and behaves as a transaction balance, it substitutes for a checking account, not for a savings account. The marginal dollar entering stablecoins comes out of demand deposits. Demand deposits fund bank lending. The New York Fed is already researching deposit outflow, which tells you the institution regards this as a live stability question rather than a hypothetical one.
On the reserve side, the mechanics favor the front end of the Treasury curve. Stablecoin reserves sit in bills and government money market funds. Supply growth transmits into structural demand for short-dated government paper. This is not speculative; it is a mechanical consequence of the reserve composition the note itself describes, and it is the single most legible second-order effect in the entire document.
On infrastructure, the note observes that Wall Street is constructing settlement rails on this base. That is the longest-duration item in the publication. Settlement infrastructure is a decade-scale capital commitment. Institutions do not build it for a product they expect to be prohibited.
Downstream, the effect on DeFi is indirect and, in my view, overstated. A stablecoin legally classified as a checking account is not automatically a composable DeFi primitive. Composability is a function of smart contract risk, oracle reliability, and sequencer liveness. The Federal Reserve measures none of those, and a favorable statistical classification does not audit a single line of code. Anyone extrapolating institutional legitimacy into DeFi total value locked is skipping an entire step of the causal chain.
Contrarian
The consensus read arrived within hours: the Fed has recognized stablecoins, therefore bullish. I want to argue that the opposite reading is at least as defensible, and that the market is pricing the wrong variable.
Classification is not endorsement. When a statistical agency decides to measure something, it is not promoting it. It is establishing the baseline from which the thing will be regulated, taxed, and โ if circumstances require โ constrained. The moment an instrument enters the monetary aggregate, it enters the policy transmission mechanism. Dollars in a bank account are countable, but they are also reserveable. Instruments inside M1 are instruments the central bank can, in principle, reach with ordinary tools.
Consider what else the note contains. It is a plan to build a survey, not a program to build an industry. Institutions distinguish carefully between having decided to act and having decided to measure, and the differences show up in language. The note explicitly declines to promise policy. Reading it as a commitment is the most common analytical error of the last two cycles. In 2022, the market read the existence of a 20 percent yield on an algorithmic stablecoin as evidence of sustainability. It was evidence of mechanism, and the mechanism was a reflexive loop with a finite half-life. I shorted LUNA through perpetual DEXs and lost 15 percent to slippage while preserving capital. The lesson was not about direction. It was about the persistent human habit of confusing the existence of a structure with the durability of one.
Third, and this is where I would actually stake capital: the framework creates an accounting category โ tokenized deposits โ that the banking system is structurally better positioned to fill than any non-bank issuer. The Board has publicly flagged the gap. Closing the gap produces a line item. The line item creates a compliance standard. The standard favors the entity that already holds deposits, already holds a charter, and already files reports to a supervisor who can revoke its existence.
Statistical recognition is not legal permission. It is the construction of a scoreboard, and every scoreboard favors whoever is already on it.
Where the bull case is genuinely strong, I will grant this: institutional legitimacy reduces tail risk. The probability of a punitive prohibition event against dollar stablecoins declines materially once the Federal Reserve has published a classification methodology for them. That is a real, unglamorous improvement in the risk-adjusted return profile of any exposure to the sector. It is worth something. It is not worth what the market will likely pay for it, because the market is pricing recognition as adoption, and recognition is the cheaper of the two.
Takeaway
Watch H.6. Not the note. Not the speeches. The release.
If a tokenized deposit subcategory appears in the money stock publication, the framework has teeth and the competitive dynamic described above is live. If the next several releases are unchanged, this was a research note and a data-collection wish list, and any sector repricing on the back of it was noise with better branding.
The confirming event is administrative, scheduled, and public. That is the rarest configuration in this industry. Use it, and stop reading the note as if it were a press conference.