The headline reads like a triumph. Circle generated $1.4 billion in revenue in the first half of 2026. Then you do the subtraction โ $819 million to distribution and transaction costs, $473 million to operations, compliance, and technology โ and roughly $108 million remains. Social media did the math within hours: "Circle keeps only 7.7% of revenue." The narrative of the broken stablecoin model was born before the earnings call ended.
The arithmetic is correct. The conclusion is not.
I have spent years stress-testing balance sheets that looked clean until they didn't. During the Terra collapse in 2022, I built a solvency checklist precisely because the industry kept confusing large numbers with solvent numbers. The same confusion is happening here, running in the opposite direction. The $108 million figure is a residual of two disclosed cost lines โ it is not net income. It excludes stock-based compensation, tax, interest expense, amortization, and, most critically, the actual average yield on reserve assets. Presenting it as "profit" is a category error dressed up as analysis, and it is the kind of error that gets capital allocated to the wrong layer of a stack.
Where code meets chaos, truth emerges. So let's audit the structure, not the headline.
Circle is the issuer of USDC, a fiat-backed stablecoin and one of the two dominant dollar rails in crypto. Its model is deceptively simple: take dollars, hold them in cash and short-dated high-quality instruments, earn the risk-free rate on the float, and redeem one-for-one on demand. This is not a lending protocol or an algorithmic experiment. There is no yield promise to holders and no reflexive death spiral baked into the design.
That distinction matters, and it is where most commentary fails. The first instinct when anyone sees "58.5% of revenue leaves the building" is to reach for the Ponzi template โ the familiar hunt for unlock schedules, insider allocations, and reflexive emissions. None of it applies. USDC is not a governance token. There is no cap table of insiders waiting to dump. The correct frame is not token economics. It is the economics of a regulated interest-rate intermediary โ a spread bank wearing a blockchain interface.
Adopt that frame and the disclosure stops looking like a scandal and starts looking like a structure. The $819 million is not theft; it is the price of distribution. The $473 million is not bloat; it is the cost of operating under multiple regulatory regimes. The thin residual is not a collapse; it is the signature of a business that captures the spread between two external variables โ the policy rate and the channel's bargaining power โ both of which it controls weakly.
My 2020 work on liquidity as a service taught me to read protocols as layered dependencies rather than isolated products. Circle sits at one specific layer: the settlement seam between fiat and on-chain value. Its health is determined not by its code but by the machinery above and below it.
Let me decompose what the disclosure actually reveals, line by line.
Revenue of $1.4 billion is almost entirely reserve interest โ the yield Circle earns on the dollars backing USDC. This is the single most important structural fact, and it is routinely buried under talk of "payment rails." Circle does not earn meaningfully from transfer fees, API calls, or payment services at scale. It earns from the float. That makes it, functionally, a money-market fund with a minting button.
The reserve composition is the tell. The disclosure describes reserves as "cash and highly liquid financial instruments." Read that phrase carefully: it is not "all cash." It means short-term Treasuries, money-market instruments, and deposits whose yield tracks the Federal Reserve's policy rate almost mechanically. This is the technical root of Circle's rate sensitivity. The revenue line is a leveraged bet on the policy rate, and the leverage runs the wrong way in a cutting cycle.
Here is the mechanism most analysts miss. Suppose the policy rate falls by 150 basis points while USDC circulation holds flat or even grows. Reserve yield drops proportionally. Revenue falls even as the balance sheet expands. The revenue curve is negatively convex to its own growth โ you can scale the float and shrink the income. That is not a hypothesis; it is arithmetic, and the disclosure flags it directly by naming rate sensitivity as a primary driver.
Now the cost side, which is where the real story lives.
$819 million in distribution and transaction costs is 58.5% of revenue. This is the number that should dominate every conversation about stablecoin economics, and almost no one is discussing it correctly. The wording matters: the disclosure uses "payouts," not "expenses." That single word implies revenue-sharing, not a flat vendor fee. If it is a share โ as the Coinbase arrangement has historically been structured โ then Circle's marginal economics deteriorate as circulation grows. The more successful USDC becomes, the more it pays for that success. This is the critical information gap: fixed fee, volume-based split, or hybrid? The answer determines whether scale is an asset or a liability. The disclosure does not resolve it, and I will not pretend it does. But an analyst who ignores this fork is not analyzing; they are narrating.
Set that against the operating line: $473 million for operations, compliance, and technology โ 33.8% of revenue, semi-fixed, and stubbornly incompressible. Licensing across jurisdictions, audits, legal, treasury management: these do not shrink when volume dips. Compliance is simultaneously a moat and a tax. It blocks new entrants, and it cannot be cut when margins compress. In a competitive downturn, that is the worst combination a cost base can have.
Now the structural paradox that reframes the entire sector.
Circle occupies the position of critical infrastructure. USDC is the settlement seam โ the pipe through which fiat becomes on-chain value. And yet the entity that owns the pipe captures a minority of the economics flowing through it. The channel โ Coinbase, exchanges, wallets โ captures 58.5%. The nominal infrastructure provider is, in practice, a channel vassal. This is not unique to crypto. In card networks, the acquirer and issuer capture more than the network. In mobile payments, the app store takes its cut before the payment rail sees a cent. Circle has rebuilt a familiar rent-extraction geometry on new rails, and it is paying the same toll every infrastructure provider before it has paid.
The Tether contrast is instructive precisely because it is not a moral comparison. Tether pays no distribution fee because it does not buy distribution; it hands USDT to exchanges as a liquidity tool and lets network effects do the work. Circle chose the opposite path โ transparency and compliance purchased through licensed channels that must be paid. Both models function. Their margin profiles are not in the same universe. Circle's 58.5% is the visible price of legitimacy, and the market is only now pricing that price.
What does each party actually capture? Holders capture convenience and an on/off ramp, no yield. Circle captures the reserve interest on the float. The channel captures more than half the revenue. The value-capture hierarchy is inverted relative to the value-delivery hierarchy, and that inversion is the single most under-discussed fact in the stablecoin trade.
There is a second-order risk hiding in plain sight. USDC is deployed across multiple chains, and multi-chain deployment means multiple contract sets plus third-party bridges. The stablecoin contract itself is a simple token; the systemic exposure lives in the bridges that move it. That is a technical risk the disclosure never touches, and one that compounds the channel dependency because bridge liquidity is often provided by the same exchanges that extract the distribution fee.
And then the deepest structural risk: reverse integration. Coinbase is both the largest distribution partner and a potential issuer of its own dollar. A channel that owns the customer relationship can replace its supplier at will. Stablecoins are near-perfectly homogeneous โ USDC, USDT, and the regulated newcomers are interchangeable at the peg. Switching cost is effectively zero. There is no positive-feedback network effect here, which is exactly the moat logic that L1s and DeFi protocols rely on and that stablecoins simply do not have.
The consensus forming around this disclosure is that Circle's model is broken. I think that reading is lazy and, more importantly, mispriced.
The model is not broken. It is a spread bank, and spread banks are profitable in the right rate regime and structurally thin in the wrong one. That is not a scandal; it is a business description. The contrarian move is to stop asking "is Circle profitable?" and start asking "where in this stack does value actually accrue, and is the market pricing the right layer?" Because if the profit center is the channel, then the beneficiaries of the stablecoin boom are not the issuers โ they are the distribution layers, and the market still values them as if issuance were the prize.
Here is the blind spot almost everyone shares: the analysis treats stablecoins as a token story when it is a cash-flow story. The market prices USDC by circulation โ market cap, TVL, dominance โ because that is the metric that has always been visible. But circulation is a vanity number if the issuer keeps 7.7% of a float it does not fully own the economics of. The real metric is net revenue per dollar of float, and nobody is publishing it. That silence is the information gain.
And there is an angle the disclosure ignores entirely: the path from spread bank to payment infrastructure. If Circle can shift revenue toward transaction fees, API services, and cross-chain settlement โ non-spread income โ it re-rates from a rate-sensitive financial stock to a payments multiple. That transition is the only lever that lifts the valuation ceiling. The disclosure does not discuss it. That omission is itself a signal worth tracking. Watch whether the next two disclosures begin to segment non-interest revenue. If they do, the story is changing. If they don't, the spread-bank ceiling stands.
One more contrarian note on compliance. The $473 million that looks like a pure drag may be a latent tailwind. If stablecoin-specific legislation forces one-to-one cash or short-Treasury backing plus periodic attestation, Circle is already close to compliant while opaque competitors are not. Regulation made explicit becomes a competitive advantage. The cost that compresses margins today could become the moat that widens them tomorrow โ provided the rate cycle does not gut the revenue line first. The scissor is real: falling rates and rising compliance costs arriving together is the worst case, and the disclosure does not rule it out.
Auditing the narrative, not just the numbers, means recognizing that this disclosure is not a price event for USDC โ the peg does not care โ but a repricing event for the entire stablecoin thesis. The next narrative is not "issuance." It is "who captures the channel." Watch the distribution terms, watch the rate path, and watch whether Coinbase ever decides to issue its own dollar. The architecture of trust, rebuilt line by line, will tell you which layer deserves the multiple โ and it will not be the one with the minting button. The question is not whether Circle survives the rate cycle. It is whether the market can finally tell the difference between the size of the float and the size of the profit.