Measured 2026-09-11. 3,144 payments, 51.58 hours, one facilitator's entire settlement flow on Solana mainnet.
The x402 ecosystem has a headline it likes: Solana overtook Base. On transaction count, that appears to be true, and it is the kind of number that gets quoted in every ecosystem roundup.
We counted the rail itself rather than the directory, and found something the headline cannot show: a large share of Solana's x402 transaction count is a loop that moved twelve cents and cost roughly twenty times that in network fees.
A facilitator pays the Solana network fee for every settlement it processes, so every payment it touches appears in one account's history. We enumerated that account — PayAI's Solana fee payer — and decoded each transaction's token-balance deltas into a payer, a merchant and an amount. This measures the rail, not a directory: it sees merchants nobody has indexed. The window is 51.58 hours, ending 2026-09-11. 3,144 USDC payments settled, totalling $68.23.
These are floors, not totals. 34 of 3,195 signatures (1.06%) could not be fetched, and the counters say so. The arithmetic closes exactly: 3,195 − 2 unavailable − 32 errors = 3,161 fetched = 3,144 payments + 1 failed + 16 non-payment.
One merchant received 2,472 of the 3,144 payments — 78.6% of all transactions on the rail. Every one was for $0.00005. They came from a single payer. The entire flow moved $0.12.
That is not a market. It is a loop, and it is most of the count.
Strip it out and the rail looks like this: 672 payments, $68.11, 29 merchants, 79 distinct payers over 51.58 hours. Every headline figure about Solana's x402 volume should be read against both numbers, because the difference between them is a factor of 4.7 in the average transaction.
Every one of those 3,144 payments cost exactly 10,001 lamports in network fees, paid by the facilitator.
| Fees across all 3,144 payments | 0.031443144 SOL |
| Fees on the 2,472-payment loop alone | 0.024722472 SOL |
| Value that loop transferred | $0.1236 |
| Those fees, at the rate this wallet actually got on the day | $2.47 |
| Fee-to-value ratio | 19.98× |
Each payment moved five hundred-thousandths of a dollar and cost about a tenth of a cent to settle. The fees were twenty times the value transferred.
This does not depend on the price of SOL. At a twentieth of the rate we used, the fees would still equal the value moved. Anyone who wants to check that can do it from the two numbers above.
We are not accusing anyone of anything. The loop may be a test harness, a monitoring probe, a deliberate stress run, or an agent with a bug. We cannot tell intent from on-chain shape, and we are not going to guess. What we can say is what it cost and what it moved.
Two things follow, and both are uncomfortable for the way this market is described.
First, transaction count on a subsidised rail is not a demand signal. When settlement is free to the buyer — because the facilitator pays the fee — the marginal cost of generating transactions is zero, and count stops measuring anything. The same census found Solana's value through this facilitator at about $31.7/day, against $312/day across the Base directory we measured the day before. Solana wins the count and loses the value.
Second, facilitators are quietly paying for activity they may not want. A rail whose fee bill on one loop exceeds the value that loop moved is a rail with a cost centre, not a revenue line. That is a useful thing for anyone operating or funding one to know, and as far as we can tell nobody has published it, because everyone counts transactions and nobody counts who paid for them.
different — and on the second one we measured, the numbers genuinely do differ: no sub-cent payments at all, but a closed 18-address loop accounting for 45% of settlements and 80% of value, never touching an outside address.
day ($99.89/SOL). The ratio is what is robust; the dollar figure moves with the price.
amounts, fees and account owners all matched — and we are stating that as three rows, not as a proof.
Every figure above is re-derivable from a public RPC and a facilitator's published fee payer. We ran the computation four times, independently, by four different means, and got the same number to five significant figures each time. The method is published; the row-level data is not, because publishing a list of specific merchants' takings is a different thing from publishing what the rail costs, and we are not going to do the first one to make a point about the second.
← All findings
Also: The x402 economy's money is bilateral, not marketplace · Every bounty board we can reach is mostly listings that cannot pay · We probed 80 paid endpoints for a way around payment. We found none.