How big is the x402 market? What the careful numbers say, October 2026

How big is the x402 market? What the careful numbers say, October 2026

The careful measurements put genuine x402 commerce at roughly $15 million to $26 million in total over the protocol's first eleven to sixteen months, with a small and uncertain share paid by AI agents; headline figures of "over 100 million payments" count every settlement, including self-payment, wash trading and tests. Daily volume peaked in late 2025 and fell about 93% from the start of 2026 to mid-August. The infrastructure works. The economy it was built for is still small.

This article gathers the figures that circulate, says what each one counts, and gives a five-question test for any x402 statistic. It draws on the Group's book *The x402 Economy* and its economics research notes (see science.561.group). The external figures are current to 27 September 2026 unless a later date is stated.

Why is the x402 market so hard to count?

Every x402 payment settles as a transfer on a public blockchain, and anyone can download every transfer and add them up. That is rare in payments: nobody outside Visa can count Visa's transactions. But a blockchain records transfers, not intentions. It shows that one address sent another 0.01 USDC through a facilitator. It does not show whether the two addresses belong to different people, whether anything was delivered, whether the payer was a person, a script or an AI agent, or whether the payment was a purchase, a test or an attempt to make a protocol look busy.

Four kinds of activity inflate the raw count:

1. Self-payment. A seller pays its own endpoint, to test it or to make it look popular.
2. Wash trading. Addresses pay each other in circles to create the appearance of volume, often to promote a token or a project.
3. Developer tests. Builders pay small amounts to check that their code works.
4. Token launches. In October 2025 a token called PING was created entirely through x402 payments: each payment of 1 USDC minted a share of the token. Chainalysis reported that x402 transactions on Base rose more than 10,000% in a single week. Those were real payments, but they bought a token, not a service.

What are the published x402 volume figures?

Taken together, the careful figures agree more than the headlines suggest.

Filtered commerce is a few tens of millions. Visa and Artemis found $15 million of plausible commerce in x402's first eleven months. TRM found $25.6 million in its first sixteen. Removing wash trading and tests cut the dollar total by about 89% in the first study and by about half in the second.

Transaction counts shrink less than dollars. In the Visa and Artemis study, filtering removed 39% of transactions but 89% of value. The genuine payments are many and small. The inflating ones are fewer and larger.

The agent share is small and uncertain. TRM's range, 0.6% to 7.5% of filtered value, is wide because, in its words, a conventional script "can execute exactly the same sequence" as an agent. Nothing on the chain marks a payment as agentic.

Activity peaked early and has fallen. Daily settlement volume approached $800,000 on many days in late 2025 and sometimes passed $1 million. By mid-August 2026 the seven-day average was about $41,800. Much of the 2025 peak was token speculation and testing that did not continue.

Is the "last 30 days" counter on x402.org accurate?

Not as a 30-day figure. For much of 2026 the x402 project's website showed a panel labeled "Last 30 Days" with four numbers: 75.41 million transactions, $24.24 million in volume, 94,060 buyers and 22,000 sellers. Those numbers were widely quoted as the market's current monthly size. In September 2026 the writer Daniel McGlynn documented that the panel had shown the same four numbers since March. The page gave no timestamp, no method and no breakdown by chain, and reports that cited it passed the stale figures along.

We made the same mistake. An internal 561 Group note written in late September 2026 quoted the frozen numbers as the thirty days to 13 September. The note has been corrected. The mistake is typical: a number that looks official, carries a date label and matches earlier reporting is easy to repeat and hard to question.

What does an independent tracker say?

The Group's economics note of 18 October 2026 records a narrower, independent view: x402stats.io reported approximately $685,000 of filtered organic USDC volume from 94 qualifying sellers over a trailing 30 days, and about $1.1 million unfiltered by its indexing. Its monthly organic windows ran roughly $428,000 (July), $503,000 (August), $1.2 million (September) and $793,000 (the October 1 report), while the count of qualified sellers went 173, 124, 78 and 91.

That tracker's method is much stricter than the others (a minimum $100 per wallet, three buyers, an average payment of $0.001, known facilitators excluded), so it is not a comparable denominator with the Visa or TRM totals. We cite it for what it shows about shape, which is volatility and concentration, not steady growth. We have not independently reproduced it.

What does Coinbase's Bazaar catalog tell us, and what doesn't it?

Coinbase runs a directory of x402 services called the Bazaar, which lists an endpoint after its first payment settles through Coinbase's facilitator. The Group read the whole catalog on 27 August 2026. It held 14,841 resources, 14,781 of them active. Over the previous thirty days the catalog reported 292,674 calls and about $30,600 in revenue; summing unique payers resource by resource gave 42,316, a figure that counts an agent once per resource it called and so overstates distinct buyers.

Revenue was highly concentrated:

• one gift-card seller earned about $21,000 of the $30,600;
• the top ten resources together earned 78%;
• the other 14,771 active resources shared about $6,800, less than fifty cents each for the month;
• the median price was one cent, and the middle half of prices ran from $0.003 to $0.05.

The most popular services by distinct payers were thin lookups on the Ethereum blockchain: an address's balance, the owner of a name, the current block number, one transaction's details. Each drew about 1,000 payers and about 1,000 calls, a ratio of calls per payer near one, the pattern of buyers sampling a service once and leaving.

Why can't you divide Bazaar calls by on-chain transfers?

The same day, a sampled window of Base blocks suggested about 3.3 million USDC AuthorizationUsed events a month, so dividing 292,674 by 3.28 million gives about 9%. The Group's first reading of that quotient said the Bazaar undercounts x402 settlement by about eleven-fold. That conclusion was withdrawn on 17 October 2026. The two numbers measure different things. One is a facilitator-reported call count; the other is an extrapolation of a general-purpose chain event, EIP-3009 authorizations, which many applications besides x402 use. Neither was joined to the other by transaction identity, resource, facilitator, chain and time window. Bazaar calls may not be settled authorizations; multi-network calls, retries and timing windows differ; and not every x402 scheme settles through Base USDC EIP-3009.

The right reading: the Bazaar catalog measures the catalog (endpoints that opted in and settled through one facilitator), the chain sample measures a shared settlement primitive, and neither is "the market". Anyone who gives you a percentage coverage figure from dividing them has not established it. Establishing coverage would need event-level reconciliation of exact resource identities against independently attributed x402 authorizations.

How should I read an x402 statistic?

Ask five questions of any figure:

1. Raw or filtered? Does it include self-payment, wash trading, tests and token mints? Was anything removed, and how?
2. Which chains? Base, Solana and others each carry part of the traffic. A Base-only figure is not a total.
3. Which period, and when was it measured? A "last 30 days" figure should say which thirty days.
4. Transactions or dollars? Counts favor many tiny payments; dollar totals favor a few large ones.
5. Payments or commerce? A payment is a transfer. Commerce is a transfer in exchange for something delivered. The chain shows the first; only filtering and judgment estimate the second.

A sixth question applies to catalogs: do they rank sellers by raw payment counts, which are easy to fake, or by distinct payers who return over time, which are hard to fake?

So how big is the x402 market?

• The infrastructure works. Hundreds of millions of settlements moved without the protocol breaking.
• Genuine commerce is tens of millions of dollars in total, across all sellers and more than a year. That is small beside the protocol's backers or the card networks' daily volume.
• Demand has fallen from a speculative peak, most of which was not commerce, and later windows are volatile.
• Agents are a minority of payers. Nobody can yet say exactly how small a minority.
• Revenue is concentrated. In the Bazaar, a handful of sellers earned nearly everything and the typical listing earned cents.

CoinDesk summarized the position in March 2026: the infrastructure is arriving before the economy it is meant to serve. We run a small x402 seller ourselves and we agree; see What happens when you sell an API on x402? For what would have to happen for the market to grow, the book's closing chapter lists seven conditions, each with something observable to watch: services buyers return to, buyers trusted with budgets, honest measurement, fixed weak points, recourse, settled law, and crawlers that pay.

What are the sources?

Visa and Artemis, "Agentic Payments from the Ground Up" (July 2026); TRM Labs via PYMNTS, 15 September 2026; Chainalysis, "Inside x402: 100M Agentic Payments on Base" (3 June 2026); Sam Reynolds, CoinDesk, 11 March 2026; CCN via Yahoo Finance, 13 August 2026; Daniel McGlynn, the counter post (6 September 2026); and the Group's own measurement of the Bazaar catalog (27 August 2026, as corrected 17 October 2026). The Visa and Artemis report was read through secondary reports, not directly.

Related: What is x402?, Who built x402 and who governs it, x402 security.