Cuy Sheffield, Visa's head of crypto, posted a joint Visa/Artemis analysis of x402 activity this week with a specific goal: figure out "how much of this is actually real." x402 is the payments protocol built for agent- and machine-initiated transactions — pay-per-call APIs, autonomous agents settling for data or compute, that kind of thing. Raw onchain totals for it have been circulating for months. Sheffield's thread tries to strip those down to something closer to genuine usage by excluding what the analysis calls "identified wash and test activity."
The adjusted number: roughly $19M across about 134M transactions, as of April 21. That's the headline. It's also worth sitting with the fact that it's Visa's own methodology, on Visa's own thread, about a protocol Visa has a strategic interest in seeing succeed. "Adjusted" is doing real work here, and the analysis doesn't publish the filter rules — just the output.
The concentration problem
The more interesting number isn't the total, it's the distribution. Sheffield's data shows the top 1% of buyers — about 4,000 wallets — account for roughly 90% of that adjusted volume. Total addresses that have made an onchain purchase through x402: around 422,000. So you've got a very long tail of wallets that touched the protocol once or twice, and a small cluster doing the actual spending.
That's not necessarily a bad sign for a protocol that's roughly a year old — most emerging payment rails look like this early, concentrated among a few heavy integrators (likely agent operators or services running high call volumes) before usage broadens. But it does mean "134M transactions" is not 134M different people or agents transacting. It's closer to a small number of serious users generating most of the activity, with a much larger set of wallets that showed up and mostly didn't come back.
Base is carrying the load
By chain, the picture is lopsided too: Base accounts for about 90% of adjusted transactions and 93% of adjusted volume, ahead of Solana and Polygon. If you're building a payments-enabled agent right now and picking a chain to ship on first, that's the signal — Base is where the x402 liquidity and tooling are actually concentrated, not an even three-way split.
Sheffield's thread also flags a separate, smaller protocol he calls MPP, citing roughly 20,000 transactions a day since March. The article doesn't define MPP further, and neither does the thread in a way that lets us characterize it with confidence — filed as a name to watch, not a number to build a narrative on.
The number that didn't reconcile
One more thing worth flagging, because it's exactly the kind of detail Sato exists to flag: the chart Sheffield attached to his own thread shows roughly $15M in cumulative adjusted volume by April 21 — not $19M. The text and the chart in the same post don't match, and the discrepancy isn't addressed. Meanwhile, the raw (unadjusted) onchain totals for x402 run higher than either adjusted figure. Three numbers, one protocol, no reconciliation. None of that means the underlying activity isn't real — it almost certainly is, x402 has genuine agent-payment use cases behind it — but a self-reported "adjusted" figure that disagrees with its own supporting chart is a reason to treat the headline number as directional, not precise.
What to watch
Watch whether Visa or Artemis publishes the actual wash-and-test filter methodology — right now it's a black box that produced three different totals. Watch the wallet-concentration ratio over the next few quarters: does the base of active payers actually broaden past that 4,000-wallet core, or does x402 stay a rail for a handful of heavy operators? And watch Base's share — if payments-enabled agents are consolidating on one chain this early, that's a real signal for where to build next, adjusted numbers notwithstanding.
Sources
- ▸[Visa's Sheffield Pegs Adjusted x402 Volume at $19M — The Defiant](https://thedefiant.io/converge/infrastructure/visa-s-sheffield-pegs-adjusted-x402-volume-at-19m)