Unstable

The record combines public category work and public market materials. TRM Labs (2025), McKinsey (2025), and Artemis (2024) provide the volume decomposition logic. Circle’s S-1 and CCTP materials, Visa’s public disclosures, Stripe’s Bridge announcement, and public funding reporting provide the valuation and market-positioning layer. The $390 billion figure used here is a residual classification estimate, not a claim of exact measurement. Its force lies in the order of magnitude it establishes once identifiable non-payment categories are removed from the aggregate.

That method supports a narrower claim than a full stablecoin sector history or a complete valuation model. It identifies a classification gap between aggregate volume and payment function, then asks what market work that gap performs.

Stablecoin volume aggregates distinct functions: DeFi routing, AMM rebalancing, exchange settlement, bridge mechanics, treasury movement, flash loans, and a residual that plausibly qualifies as payment activity. The headline aggregate ($33 trillion per Artemis/Bloomberg, approximately $35 trillion per TRM Labs, depending on chain coverage and classification methodology) combines activities whose economic character, counterparty structure, and institutional meaning diverge so sharply that no single comparison can describe them without distortion. Classification is structurally prior to valuation because treating the aggregate as a unified category confuses the object before measurement begins.

Once the aggregate circulates as a single number, it becomes the basis for market-sizing narratives and the valuations those narratives authorize. Circle’s S-1 (Circle) references USDC’s cumulative $25 trillion in on-chain movement and projects a $1.9 trillion stablecoin market by 2030. Stripe’s Bridge acquisition (Stripe) framing referenced stablecoin volume growth as a directional signal. Venture funding into the stablecoin middleware layer exceeded $1.5 billion in 2025, including Tempo’s $500 million Series A at a $5 billion valuation (Fortune Crypto 2025). The mechanism linking imprecision to mispricing operates through institutional incentive: aggregate statistics flatten heterogeneous activity into a label that reads as "payment volume," and the actors best positioned to correct the label are those whose valuations depend on preserving it.

Goodhart’s original formulation (Goodhart 1975) holds that a measure ceases to be reliable once it becomes a target, but stablecoin volume presents a sharper case than ordinary Goodhart dynamics. DeFi routing, AMM rebalancing, flash loans, liquidation cascades, exchange wallet management, and bridge burns and mints all generate transaction volume as a normal byproduct of operation. Deliberate wash trading further inflates the aggregate (Cong et al. 2023), but the endogeneity operates even without intentional manipulation: the infrastructure produces the metric in the course of performing its designed function. Each protocol interaction routed through a stablecoin adds to the aggregate regardless of whether the interaction bears any relationship to payment activity. Volume is endogenous to the system, and this endogeneity creates circular justification: volume validates the infrastructure while the infrastructure generates the volume.

Three independent sources have separated stablecoin transaction categories with varying granularity. McKinsey’s 2025 payments analysis estimated that consumer and B2B payments represented a low-single-digit percentage of total stablecoin volume after filtering DeFi, exchange, and bridge activity (McKinsey 2025). Artemis’s stablecoin category work provided chain-level volume decomposition showing DeFi and exchange settlement as dominant categories across Ethereum, Tron, and Solana (Artemis 2024). TRM Labs' own usage report (TRM Labs 2025) places the aggregate closer to $35 trillion, while Artemis data reported by Bloomberg yields the $33 trillion figure. The difference is methodological and does not affect the classification problem the article addresses.

For purposes of this analysis, payment activity denotes transfers whose primary economic function is the settlement of an obligation between a payer and a payee for goods, services, or remittance, as distinct from transfers whose primary function is protocol operation (DeFi routing), market infrastructure (exchange settlement), or cross-chain mechanics (bridge burns and mints). This is a functional definition, and boundary cases exist: B2B settlement and OTC transactions serve both payment and infrastructure functions depending on context. The $390 billion figure used here is a classification residual: the amount remaining after subtracting identifiable non-payment categories from the $33 trillion aggregate. The subtracted categories are DeFi routing (AMM swaps, liquidations, flash loans, identifiable through contract-level interaction patterns), exchange settlement (CEX deposit and withdrawal flows, identifiable through known exchange wallet clusters), bridge mechanics (burn-and-mint transfers, identifiable through bridge contract addresses), and treasury management (large-denomination wallet-to-wallet movements with no downstream retail signature). What survives this filtering is the residual plausibly attributable to remittance, payroll, merchant settlement, and person-to-person transfers.

The approximate per-category volumes, derived from Artemis’s chain-level decomposition and cross-referenced against TRM Labs' usage categories and McKinsey’s filtering methodology, decompose the $33T aggregate as follows. These are order-of-magnitude estimates rounded to the nearest trillion (or half-trillion where finer resolution is supported), and different boundary choices within each category would shift individual figures by several hundred billion in either direction without altering the residual’s order of magnitude.

Each category estimate carries independent uncertainty, and the directional biases tend to inflate the residual. The DeFi routing figure ($14T) derives from Artemis’s contract-interaction classification, which identifies transactions involving known DeFi protocol addresses; transactions routed through aggregators or multi-hop paths that obscure the originating protocol may be underclassified, pushing the DeFi estimate down and the residual up. The exchange settlement figure ($15T) relies on known exchange wallet clusters maintained by Artemis and TRM Labs; unidentified exchange wallets would produce the same directional bias. Bridge mechanics ($2.5T) is the most precisely bounded category because bridge contracts are architecturally identifiable. Treasury and speculative flows ($1.1T) is the least precise because large-denomination wallet movements serve multiple functions (OTC settlement, treasury rebalancing, fund management) whose classification requires assumptions about intent.

The four non-payment categories sum to approximately $32.6T, leaving the $390B residual after rounding. Because each component estimate carries its own uncertainty band, the residual should be read as an order-of-magnitude figure. A reader reconstructing this arithmetic with different boundary assumptions will arrive at a residual in the $200B to $800B range. A reviewer applying maximally conservative classification boundaries (attributing ambiguous multi-hop DeFi transactions to payments, including all unidentified exchange wallets in the residual, treating OTC settlement as payment activity) could push the residual above $1T. Even under those assumptions, classified payment activity remains below 3% of the aggregate, and the gap between the aggregate and the payment residual spans two orders of magnitude that no reasonable boundary-drawing exercise closes.

The residual methodology carries two structural biases that readers should hold in view. First, every unclassified transfer defaults to the payment category, which means the $390 billion figure functions as an upper bound rather than a point estimate. Second, the residual inherits the summed uncertainty of all four subtracted categories; if each carries +/- 20% error, the residual’s absolute uncertainty can approach or exceed the residual itself. The figure should therefore be read as an order-of-magnitude claim (hundreds of billions, low single-digit percentage of the aggregate) whose analytical force derives from the two-order-of-magnitude gap between the residual and the headline aggregate, not from the precision of $390 billion as a number. Different analysts will draw boundaries differently (some classify B2B settlement as payment activity, others place it under treasury management), and the residual absorbs every classification ambiguity as additional payment volume. The analytical conclusion does not depend on this figure’s precision. Tripling the payment residual to $1.17 trillion would place stablecoin payments at roughly 3.5% of the aggregate, still two orders of magnitude below the comparison the $33 trillion figure is routinely asked to support.

The Visa comparison persists because it performs pricing work. Circle’s S-1 materials (Circle) positioned USDC inside a "money-movement" market measured in the tens of trillions. Stripe’s public Bridge acquisition announcement (Stripe) adopted adjacent framing. Venture capital into stablecoin middleware followed the narrative the $33 trillion figure authorized, with over $1.5 billion deployed in 2025 including Tempo’s $500 million Series A at a $5 billion valuation (Fortune Crypto 2025). The comparison attaches an aggregate associated with observed growth to a familiar benchmark associated with payment dominance, and once that connection holds long enough, the market stops interrogating whether the two numbers are commensurable.

The infrastructure beneath the comparison is substantive enough to lend it surface plausibility. Circle’s CCTP V2 (Circle) addresses liquidity fragmentation in cross-chain USDC movement. Stripe’s Bridge acquisition signals institutional commitment to cross-border settlement. Visa’s own stablecoin-settlement disclosures (Visa) exist at pilot scale. These developments are real, though none of them close the gap between aggregate movement and classified payment activity. B2B settlement, payroll, cross-border remittance, and merchant use in jurisdictions with weak dollar access constitute the operative story, and letting total movement stand in for that story without disaggregation obscures the applications that warrant sustained analysis.

Once the Visa analogy is stripped away, the stablecoin business assumes a different shape. Circle becomes a regulated reserve-income business with a token issuance layer, distribution agreements, and a meaningful but far smaller payment surface, closer in structure to a money-market fund administrator with cross-border settlement capability than to a card-network operator. The intermediaries built on USDC and USDT become software companies competing inside a still-maturing settlement stack.

The reframing changes what should count as evidence. If the valuation rests on payment replacement, payment evidence should underwrite the multiple. If the evidence consists mostly of reserve income, bridge throughput, and institutional settlement pilots, the valuation story should describe those things on their own terms. Mispricing follows from treating the aggregate as though it already described the end state the infrastructure is still building toward.

An efficient-market objection merits direct engagement. If the volume-payment distinction is analytically obvious, sophisticated market participants should already discount the aggregate and price stablecoin infrastructure on its actual payment base, which would mean current valuations reflect rational expectations about future payment adoption. Under this reading, investors are pricing trajectory, and aggregate volume serves as a legible proxy for infrastructure growth potential. Institutional investors with access to detailed on-chain analytics can and do distinguish volume categories; Artemis, Chainalysis, and internal trading-desk research perform exactly this disaggregation.

Two barriers prevent the disaggregated picture from correcting the aggregate framing, and they operate through distinct mechanisms. The first is structural: in token markets where short-selling is constrained (Makarov and Schoar 2020) and institutional lock-ups restrict the temporal window for correction, mispricing persists because the arbitrage mechanism that efficient-market logic relies on is partially disabled. The second is positional: the institutional actors who benefit from the aggregate framing (issuers, exchanges, venture funds) occupy positions where the framing shapes capital flows through S-1 narratives, acquisition multiples, and venture allocation frameworks. These actors influence pricing through narrative positioning, and their valuations face testing against revenue outcomes only after public listing. The short-selling constraint is a market-structure limitation applicable to any mispriced crypto asset. The positional incentive is specific to the classification problem: the actors best positioned to correct the label depend on maintaining it.

Three independently testable conditions would overturn the classification-gap thesis.

First, the classification gap closes. If classified payment volume (retail and commercial transactions, excluding DeFi routing, bridge mechanics, exchange settlement, and treasury management) exceeds 25% of total stablecoin volume for four consecutive quarters as measured by at least two independent analytics providers (Artemis, TRM Labs, Chainalysis), the aggregate comparison becomes defensible and the classification critique loses force. The current estimate places the ratio near 1.2%. A tenfold increase sustained over a year would indicate a functional shift in the composition of on-chain activity that cannot be attributed to boundary-drawing artifacts.

Second, the pricing mechanism decouples. If stablecoin infrastructure providers' equity and token valuations converge on revenue-based multiples (10-20x net revenue, comparable to payment processors such as Adyen, Marqeta, or Fiserv), the category error identified here no longer drives the valuation distortion the analysis attributes to it. Circle’s post-IPO trading multiple provides the natural test case.

Third, the functional composition shifts. If the top three stablecoin issuers by market capitalization derive more than 40% of gross revenue from transaction fees on classified payment activity (as opposed to reserve income, interest on treasury holdings, or distribution agreements) for two consecutive fiscal years, the gap between transfer infrastructure and payment network has closed commercially regardless of volume composition.

If the classification gap and its valuation consequences are structural, several outcomes should follow.

Stablecoin infrastructure is processing large amounts of value, but the aggregate usually cited beside Visa does not describe payment function on comparable terms. It describes heterogeneous transfer activity inside which payment activity is only one component.

That distinction changes the valuation question. The relevant issue is not whether stablecoin infrastructure matters. It is how much of the observed volume belongs to payment function, how much belongs to routing and settlement mechanics, and what pricing follows once those functions are separated.