Seven Cents on the Dollar: The Plumbing Behind Circle's Float

Seven cents. That is the number making the rounds: for every dollar of revenue Circle books, roughly seven cents survive as profit. The framing writes itself — a money printer that keeps a rounding error, a stablecoin empire running on fumes. I have read that framing a dozen times this month, and it is structurally the wrong reading.

Before anyone argues whether seven cents is good or bad, the denominator has to be fixed. On a net-margin basis — net income over total revenue — Circle's 2024 figures of roughly $1.68 billion in revenue against net income in the $150 million range land at 7 to 9 percent. That matches the headline almost exactly. On a gross-retention basis — revenue minus distribution costs, divided by revenue — the number is closer to 40 or 45 percent, because distribution alone consumes 55 to 60 percent of revenue. Two completely different businesses hide inside those two definitions. The headline borrows gross-style rhetoric — only keeps — to describe a net-margin fact. That is the first contradiction, and it matters more than the number itself.

Context: What Circle Actually Is

Circle is not a protocol. It is not a token issuer in the sense most of this industry uses the phrase. Circle Internet Group is a licensed financial institution that runs a fiat-backed centralized stablecoin, USDC, and it listed on the NYSE in June 2025 under the ticker CRCL. That single fact — equity, not token — reorganizes the entire analysis. There is no governance token to model, no emissions schedule to stress-test, no staking yield to decompose. Value capture happens at the equity layer or it does not happen at all.

USDC itself sits on Circle's balance sheet as a liability. Every USDC minted is a one-dollar redemption obligation; every dollar of obligation is matched by a dollar of reserve, held primarily in short-dated US Treasuries and repo, managed through BlackRock, custodied at BNY, attested by Deloitte. The float earns interest. That interest is the revenue line. Everything else — the API revenue, the CCTP transfer fees, the Programmable Wallets subscriptions — is noise at the margin. Roughly 95 percent or more of Circle's revenue is reserve interest, not payment services. Read that twice. Circle is not a payments company with a float attached. It is a float with a payments company attached.

The competitive moat is equally clarifying. Mint-and-redeem logic is standardized engineering; any licensed financial institution with a treasury desk can replicate it in six to twelve months. Cryptography is not the barrier. The barriers are three: licenses across jurisdictions, distribution contracts with the venues where dollar liquidity actually concentrates, and the operational capacity to manage tens of billions in reserve assets without breaking a peg. Of those three, only one is genuinely defensible, and it is the one that costs the most — compliance.

Code is law, but incentives are god. And the incentive structure here points in exactly one direction: an issuer with no technical monopoly must rent its distribution.

Core: Decomposing the Dollar

Start with the arithmetic. A dollar of reserve interest enters Circle and gets sliced three ways.

Roughly 50 to 60 cents flows out as distribution cost — paid to Coinbase, to exchanges, to market makers, to the venues that hold USDC and make it liquid. Another 30 to 40 cents goes to operating expense: headcount, legal, audit, licensing, engineering, marketing, the entire compliance apparatus. What remains is the seven to ten cents that becomes the headline. The seven cents is not a mystery. It is a subtraction.

The structural cause is one contract. Under the revenue-sharing arrangement between Circle and Coinbase, USDC held on Coinbase's own platforms generates reserve interest that accrues 100 percent to Coinbase. USDC held off-platform splits roughly 50/50 between the two. That clause alone explains the shape of the income statement. Coinbase is not a distribution partner in any ordinary sense. It is an unconsolidated equity holder in USDC's economics, collecting a substantial share of the float's yield while bearing none of the reserve management cost and none of the licensing burden. It is an asymmetric arrangement, and it is the single largest line item between Circle's gross yield and its net margin.

Which brings us to the classification problem. Circle's net income approximates average USDC circulation, multiplied by the short-end rate, multiplied by one minus the distribution share, minus operating expense. That is a three-variable model in which Circle controls one variable. It does not set the Fed funds rate. It does not, in any practical sense, renegotiate the Coinbase split while Coinbase remains the deepest USDC liquidity venue on earth. What Circle controls is operating expense and the slow, grinding work of building distribution it owns outright.

This is why I keep insisting that Circle is not a payments company being valued as one. It is a net-interest-margin business wearing a fintech costume. In 2020, during DeFi Summer, I ran a cross-protocol liquidity strategy across Compound, Uniswap, and Aave — half a million dollars reallocated every forty-eight hours, chasing rate differentials, finishing six months up roughly 40 percent. The returns were real. The lesson was not. What I actually learned in that period is that a yield number tells you almost nothing until you decompose where the cash comes from. When the source is genuine external cash flow, the structure survives stress. When the source is the next participant's deposit, it does not.

So let me be precise about Circle, because a lot of people are quietly implying the wrong thing. Circle is not a Ponzi. Its cash flow originates in US Treasury interest — real, external, politically sourced, paid by the largest sovereign balance sheet in the world. It does not require new entrants to pay old ones. The comparison people reach for is Terra, and the comparison is wrong. When Terra collapsed in 2022 I wrote that the failure was less about algorithmic design than about dollar-denominated leverage stacked on a reflexive asset; I shorted three exchange tokens into that thesis and it paid. Circle has none of that reflexivity. There is no death spiral available to a 1:1 fiat-backed issuer with T-bill collateral. Its risk is not insolvency-by-design. Its risk is margin erosion, and margin erosion is a slow, boring, entirely survivable kind of death.

Now the part the headline misses completely. Seven cents is not a structural constant. It is a function of the rate cycle and the scale of the float. Three consequences follow, and they point in different directions.

First, operating leverage. If USDC circulation grows faster than operating expense — and compliance cost scales sub-linearly with asset size — the net margin expands through pure denominator effect. The seven cents is not a ceiling.

Second, and far more dangerous: rate sensitivity. Reserve interest is a numerator term, and it moves with the Fed. If the short end falls from roughly 4 percent to roughly 1 percent, Circle's revenue line contracts by something close to 75 percent with no change in a single operational variable. Distribution cost, meanwhile, contains fixed and contractual components that do not compress proportionally. Run that through the model and the seven cents does not shrink — it approaches zero, and in a bad configuration it goes negative. This is the most important risk in the entire Circle story, and it appears nowhere in the headline's framing. A rate cut is more lethal to this business than any competitor.

Third, cyclicality of the metric itself. Across a full rate cycle, the retained cents per dollar plausibly ranges from 3 to 20. Presenting a variable, cycle-dependent margin as a fixed structural defect is not analysis. It is a category error dressed as a finding.

There is a second-order point here that connects directly to my own audit history. In 2017, at the height of the ICO boom, I spent two months tearing apart three ERC-20 contracts and found a reentrancy vulnerability in a high-profile gaming platform's smart contract — the kind of flaw that lets an attacker drain a balance mid-call. I forced a mainnet delay and, by my estimate, prevented around $2 million in early-investor losses. The lesson I carried out of that audit was that technical integrity precedes market value. But the definition of "technical" shifts with the asset class. For USDC, the attack surface is not reentrancy. The code is boring, and boring is correct. The attack surface is the reserve attestation cadence, the freeze list, and the single-point custody chain running through BlackRock and BNY. Same discipline, different object of inspection.

The deeper structural read is this: Circle's margin problem is a moat problem, not a cost-management problem. A protocol with genuine technical lock-in — Ethereum is the obvious case — retains the overwhelming majority of its fee revenue because there is no substitute to route around it. A commoditized issuer with near-zero switching costs must concede value to whoever controls the liquidity venue. A USDC holder can swap to USDT or DAI in a single transaction on the same chain. Economically, USDC's user retention rate does not exist. When your product is essential but not exclusive, your counterparties price you accordingly. That is why the fifty to sixty cents leaves, and it is why no amount of internal efficiency work brings it back.

Which clarifies what Arc actually is. Circle's own layer-1, its Programmable Wallets, and its CCTP cross-chain transfer protocol are routinely described as product expansion. They are better understood as de-channelization — an attempt to move USDC distribution onto infrastructure Circle controls, so the fifty to sixty cents currently paid to venues can be recovered internally. If Arc achieves real adoption, the retained cents could move from seven toward thirty or forty. If it does not, Circle remains permanently a toll collector for other people's toll roads. Every product announcement out of this company should be evaluated against that single question: does this reduce dependence on third-party distribution?

Competitive positioning sharpens the same point. Tether runs somewhere around $150 billion in circulation and roughly 60 percent share; USDC sits near $60 billion and about a quarter. Tether's retained margin is materially higher than Circle's, and the reason is not operational brilliance. It is a lighter, less licensed, less audited structure. In stablecoins, compliance is a cost line, not a revenue line. Circle pays for the licenses that make it the institutional default and then watches that cost compress its margin. The payoff is real — MiCA compliance, the deepest US licensing footprint, the trust of regulated counterparties — but it is a strategic trade, not a free lunch.

The genuine competitive threat is not Tether. It is tokenized money market funds. A tokenized T-bill product can pay yield to holders; USDC cannot, because US legislation — the GENIUS Act framework — explicitly prohibits paying interest on payment stablecoins. That prohibition looks like a restriction. It is actually the invisible pillar holding up Circle's entire business model. If stablecoin holders could be paid interest, Circle would have to share the float yield with them and the seven cents would turn negative overnight. The ban keeps the float captive. The thing protecting Circle's margin is a legislative prohibition, not a competitive advantage — and any political movement toward permitting interest-bearing stablecoins is an existential risk, not a competitive one.

And watch the banks. A stablecoin is functionally a deposit that pays its spread to the issuer instead of the depositor, and it competes directly with the deposit franchise. Banking lobbies have every incentive to push for restrictions on non-bank issuance, or for a claim on reserve interest. That vector is entirely absent from the seven-cents conversation and it is the one that could actually break the model.

Contrarian: The Number Is Not the Story

The consensus reading is that seven cents is a verdict on Circle. The contrarian reading is that seven cents is a verdict on the market's valuation framework, and almost nobody is running that analysis.

If CRCL trades on stablecoin total addressable market — the two-trillion-dollar-slash-everything-goes-onchain story — then unit economics are decorative. Nobody buying the narrative is underwriting a net interest margin. They are underwriting a position in the compliant dollar layer of a future financial system. Under that framework, a rate-driven collapse in retained margin compresses earnings without touching the multiple, because the multiple was never earnings-based to begin with. That divergence between narrative and cash flow is the actual risk, and it is far larger than the seven cents.

But there is a subtler signal. The appearance of unit-economics autopsies is itself a cycle marker. When coverage of an asset shifts from "how large can this become" to "what exactly is the margin," the expansion phase of the multiple is usually already over. Bubbles don't die of old age; they die of leverage and of arithmetic. The arithmetic arriving in public is rarely early.

And the blind spot in the seven-cents debate runs the other way: the number is being read as a weakness when part of it is a subsidy. The interest-payment prohibition is worth more to Circle than any product line it ships. Framing the seven cents as failure ignores that the seven cents exists only because a law prevents Circle's customers from claiming the rest.

Takeaway: Watch the Rate Path, Not the Chart

The plumbing here is legible, and it says something specific. Circle's retained margin is a leveraged bet on the front end of the US curve, subsidized by a legislative prohibition, and constrained by a distribution contract it cannot easily renegotiate. None of those three variables is a product decision.

So don't watch the price; watch the plumbing. Track the Fed path and the Coinbase stablecoin revenue line in tandem — they are the same number viewed from opposite ends of the pipe. Track Arc adoption as the only genuine structural fix. And ask the question the headline never reaches: if the short rate returns to one percent, and the float stops paying, what precisely is left of the business — and what is the market actually holding?