Seven Cents on the Dollar: Circle's Margin Is a Function, Not a Flaw

One dollar of revenue. Seven cents retained. The headline writes itself, and that is precisely the problem. The proposition is delivered as a structural indictment. A business that 'can only keep seven cents' is presumed to be broken at the model level. It is not. It is a fee schedule. The gap between a fee schedule and a broken model is the difference between a company that scales into profitability and one that is structurally unprofitable. In a single quarter's numbers, the two are indistinguishable. Over a full rate cycle, they are opposites. Before any verdict on Circle Internet Group and the USDC float, the caliber of 'seven cents' must be fixed. This is where most commentary fails. It reads the numerator and ignores the denominator. Code executes exactly as written, not as intended — and financial statements do the same. The number means what its definition says it means, not what the headline wants it to mean. Circle is the issuer of USDC, the second-largest dollar-denominated stablecoin by circulating supply — roughly $60 billion against Tether's ~$150 billion, order of magnitude, to be verified against filings. It listed on the NYSE in mid-2025. Its revenue line is not a payments business. More than 95% of it is interest earned on the reserve assets backing USDC — short-duration U.S. Treasuries and repo, managed through BlackRock, custodied at banks, audited by a major firm. The float is the product. This single fact reframes everything downstream. Circle does not monetize transactions. It monetizes the spread between what it earns on reserves and what it pays to acquire and retain the float. Every other line item — API fees, CCTP transfers, Programmable Wallet usage — is rounding error against the interest line. The industry context matters here. In 2021, I flagged Terra USD's algorithmic stability mechanism as mathematically unsound. That model used a seigniorage token to defend a peg, and the defense required continuous new demand to pay earlier holders. When demand stopped, the mechanism inverted and $40 billion evaporated. Circle is the inverse of Terra. Its stability mechanism is trivially sound — one dollar of reserve per token — which is why it is boring, and being boring is why it survived 2022. But soundness of the peg says nothing about the soundness of the margin. Those are two different questions, and the market routinely conflates them. The peg is a liability-management problem. The margin is a revenue-architecture problem. Circle solved the first and is fighting the second. The first task is definitional. 'Seven cents' is almost certainly a net margin figure — net income divided by total revenue. If it were gross retention — revenue minus distribution costs, divided by revenue — the number would sit closer to 40–45%. If it were free cash flow over revenue, it would be lower still and depend on capitalization and stock compensation treatment. The headline's phrasing, 'can only keep,' implies a structural defect. The data supports a cost structure. These are not the same thing, and the difference decides the investment case. A net margin of 7–9% on a float business is not alarming in isolation. It is what a distribution-heavy financial intermediary looks like before operating leverage kicks in. The question is not whether the margin is small. The question is whether it is fixed. It is not. That is the entire argument. More than 95% of Circle's revenue is reserve interest. This makes it a float business with a net interest margin, not a payments company. The cash flow source is U.S. Treasury interest — real external yield, not new-entrant capital paying old participants. This must be stated plainly because the comparison gets abused: Circle is not a Ponzi. It is an interest-rate arbitrage plus a channel revenue-share model. The distinction is not semantic. A Ponzi requires continuous inflow to service outflow. Circle's inflow comes from the Treasury, which does not care whether new USDC is minted. But the same fact cuts the other way. A business whose revenue is 95% reserve interest is a business whose revenue is a function of the short rate. That is a macro variable, not a product variable. Circle does not sell a product that customers need more of in a downturn. It holds a pile of Treasuries and collects the coupon. The model reduces to a single line: net profit approximately equals average USDC float, multiplied by the short rate, multiplied by one minus the distribution split, minus operating expenses. Three variables. Circle controls one, and only partially — the float depends on adoption, which depends on network effects it does not own. It does not control the rate. It does not control the split, which is contractual and concentrated in a single counterparty. This is the model the seven cents actually describes. The number is the output of a three-variable function, not a fixed property of the business. Change any input and the output moves. That is why the headline's framing — a fixed seven cents — is a category error. It presents a variable as a constant and then moralizes about the constant. The structural cause of the seven cents is the Coinbase agreement. The terms, as publicly reported and subject to verification against Circle's S-1 and 10-K: USDC held on Coinbase earns reserve interest that goes 100% to Coinbase. USDC held off-platform splits reserve interest roughly 50/50 between Circle and Coinbase. That single contract is the largest single line in the cost structure. Coinbase is Circle's largest implicit shareholder. It extracts USDC economics without carrying reserve management, custody, or compliance cost. This is not a partnership in any meaningful sense. It is an asymmetric dependency. Circle needs Coinbase's distribution more than Coinbase needs Circle's issuance, and the contract prices that asymmetry. In my 2017 audit of the 0x protocol, the lesson was the same in a different form: the party that controls the interface controls the economics. The protocol layer does the work; the distribution layer takes the margin. Circle is the protocol layer here, and it is being taxed by its own channel. This is why the seven cents exists. Not because Circle is badly run. Because Circle is structurally positioned at the wrong end of its own value chain. USDC's technology is not a moat. Mint and redeem is standardized engineering — any licensed financial institution can replicate the logic in six to twelve months. The real barriers are licenses across jurisdictions, distribution contracts with exchanges and payment processors, and reserve management at scale. None of these are cryptographic. All of them are institutional. This is the premise for understanding why the margin is compressed. A protocol with a technical moat — Ethereum, for instance — retains most of the fee revenue because no one can fork the network effect. A commodity issuer with no technical moat must cede value to whoever controls demand. Circle cedes 50–60 cents on the dollar because it has nothing the channel cannot replace. If Coinbase wanted to issue its own dollar token tomorrow, the engineering would take a quarter. The license would take longer. The distribution is already there. That is the bargaining position, and it is weak. Utility is the vacuum where hype goes to die. USDC's utility is real — it is the most regulated on-chain dollar, and that matters to institutions. But utility without a moat is a commodity, and commodities earn the cost of capital, not a technology multiple. The real risk is not competition. It is the rate cycle. Run the arithmetic. If the short rate falls from 4% to 1%, reserve revenue falls roughly 75% ceteris paribus. Part of the distribution cost is rigid — contractual, not proportional. Operating expenses do not fall at the same speed as revenue; compliance and headcount are sticky. The net margin compresses toward zero, and in a severe enough cut, below it. This is the scenario the seven-cent headline does not discuss. It treats the margin as a fixed defect when it is a rate-sensitive output. Across a cycle, the same business will print three cents in one quarter and twenty in another. The seven cents is a thermometer of the rate environment, not a diagnosis of the model. Chaos reveals itself only when the noise stops — and the noise here is the bull-market narrative that stablecoin total addressable market grows forever. The rate is the variable that decides whether the margin survives the noise. The only true moat is regulatory, and it is stronger than the market credits. The GENIUS Act-type framework does three things: it establishes stablecoins as legal payment instruments, it requires issuers to hold licenses and 1:1 high-quality reserves, and — critically — it prohibits paying interest to stablecoin holders. That third clause is the invisible pillar of Circle's business. If interest were permitted, Circle would have to share reserve yield with holders, and the seven cents goes negative. The prohibition is what allows Circle to keep the float yield and split it only with the channel. This means the regulatory moat is a double-edged structure. Tighter rules help Circle because they push small issuers out. Looser rules hurt Circle because Tether, banks, and technology companies flood in. The moat's strength is a function of legislative strictness, and that is not a variable Circle controls. Any move to permit interest-bearing stablecoins is an existential risk, not a competitive annoyance. It should be tracked as a binary event, not a sentiment. The competitive threat that deserves real weight is not USDT. It is the tokenized money market fund — a yield-bearing on-chain dollar instrument. If that structure scales, USDC's zero-yield design becomes a visible disadvantage, and the float migrates to whoever pays. The no-interest rule currently blocks this. The market has underpriced that regulatory dividend, and it has equally underpriced the risk of its reversal. Both belong in the same column. Which brings the strategic question into focus. Why is Circle building Arc, a self-operated layer-one chain? Why Programmable Wallets, why CCTP? These are not technical vanity projects. They are de-channelization moves. If Circle can distribute USDC directly to its own wallet and settle on its own chain, it recaptures the 50–60 cents currently paid to the channel. Success takes the margin from seven to thirty or forty. Failure leaves Circle as a permanent issuer-for-the-channel, a utility that funds someone else's economics. History repeats, but the code changes the syntax. The 2017 ICO wave learned that the protocol that captures the interface captures the value. Circle is now trying to build its own interface after a decade of renting one. Whether Arc succeeds is the single most important variable in the equity story, and it is almost entirely absent from the seven-cent discussion. What the bulls got right, and the bears miss. The regulatory moat is real, and it is widening. The no-interest clause is not a bug to be fixed; it is the structural feature that makes the float business viable. Circle's compliance burden is a cost today and a barrier tomorrow. In the segment that matters for the next cycle — regulated on-chain dollars for institutions — Circle is the default pipe, and defaults are hard to dislodge. The float business is also not a Ponzi, and the comparison to Terra is lazy. Terra's yield came from its own token printing; Circle's comes from the U.S. Treasury. One is endogenous and collapses on itself. The other is exogenous and survives the issuer. That distinction is the whole ballgame, and the bears who wave 'stablecoin equals Ponzi' are waving a flag they have not read. Where the bulls are wrong is the valuation framework. They price CRCL as a payments network — a technology multiple on a total-addressable-market story. It is not a payments network. It is a rate-sensitive financial intermediary with a regulatory franchise. Price it as the latter and the seven cents stops being an embarrassment and starts being a forecast. The bulls are right about the position and wrong about the multiple. The market is currently paying for the position at the multiple, which is the definition of a mispricing waiting to resolve. Watch the rate cycle, not the quarterly narrative. The seven cents will swing from three to twenty across a cycle; the only question that matters is what happens at the bottom. If Arc lands, Circle becomes a settlement-layer player and the margin recovers. If it does not, seven cents is a ceiling, not a floor. Either way, the number to track is not the margin. It is the split — because whoever controls the channel controls the economics, and Circle is still renting its own distribution.