Because the tools agents use to pay were built around one dollar stablecoin first, and most new agents and sellers have followed the default. A crypto trading firm found 98.6% of on-chain AI agent payments between May 2025 and April 2026 settled in that single coin. The issuer’s own figure for the second quarter of 2026 was 99.3% of payment volume on the main agent payment standard.
How concentrated is agent spending, really?
Very: every measure we found puts one dollar stablecoin at around 99% of agent payments. A stablecoin is a crypto token designed to track a currency, here the US dollar. Two figures are worth separating because they measure different things.
The first comes from a crypto trading firm’s May 2026 report. It counted about 176 million on-chain agent transactions, worth about $73 million, from May 2025 to April 2026. It found 98.6% settled in one stablecoin.
The second comes from the issuer of that stablecoin, in its results for the second quarter of 2026. It said 99.3% of payment volume on x402 settled in its coin. x402 is the open standard most agents use to pay for web services. That figure covers only x402 and is the issuer’s own count, so treat it as a claim rather than an independent audit.
The difference between the two matters. A share of transactions counts every payment equally, so a million one-cent payments outweigh one large one. A share of volume weights payments by their dollar value. When both methods land near 99%, the concentration is not an artefact of how it was measured.
Either way the pattern is the same. Agents could pay in other tokens, and the standard allows it. In practice they use one.
For scale, the same issuer reported its coin’s circulation at $73.3 billion at the end of the second quarter of 2026. Agent payments of about $73 million over a year are a rounding error against that. Agents depend on the coin far more than the coin depends on agents.
Why did one stablecoin become the default?
No rule requires it. The coin got there first in the tools developers actually use, and defaults compound. Four forces did most of the work.
- It was the default in the standard’s own tooling. The x402 documentation names this coin as the most common payment asset. It also says sellers can accept any standard token on compatible networks. So the dominance is a default, not a technical lock.
- The issuer built the plumbing. The issuer runs its own payment-checking services for x402 and pays the blockchain fees to settle payments. It also offers a batched service that can carry payments as small as one millionth of a dollar. Builders take the path that already works.
- The issuer built a marketplace around it. By the end of June 2026 its agent toolkit, launched in May 2026, listed more than 900 paid services, according to coverage of its quarterly results. Every service priced in one coin pulls more agents toward that coin.
- Network effects. Sellers accept what buyers hold, and buyers hold what sellers accept. This is our inference, not a measured finding. But it is the usual reason one payment method wins and stays.
The issuer is also among the 22 organisations that backed x402’s move to an independent foundation in April 2026. That gives it a voice in how the standard develops. We would not read that as control: the foundation’s stated aim is to keep the protocol vendor-neutral.
Is agent spending also concentrated on one blockchain?
Mostly on two, and which one leads depends on who is counting and how. A blockchain is the shared ledger where the payment is recorded; agents can pay on several.
In April 2026, the foundation behind one major blockchain said its network drove nearly 65% of x402 transaction volume that year. That is a self-reported figure from an interested party. In July 2026, a joint report from a card network and a crypto analytics firm reached a different picture. After stripping out suspicious activity, it found most x402 activity on a large layer-2 network. A layer-2 is a network built on top of a main blockchain to make transactions cheaper.
The same July report adjusted x402 to about $19 million across about 134 million transactions. It also found roughly 4,000 wallets drove about 90% of that adjusted spending. So the market is concentrated in three ways at once: one coin, two networks and a few thousand active payers.
| Layer | How concentrated | Who said it |
|---|---|---|
| Which coin | About 99% in one dollar stablecoin | A crypto trading firm (May 2026); the issuer (Q2 2026) |
| Which network | Mostly two networks; the lead is disputed | One network’s foundation (April 2026); a card-network report (July 2026) |
| Who is paying | About 4,000 wallets drive about 90% of adjusted spending | A card-network report (July 2026) |
What could go wrong?
When nearly all agent payments rely on one coin, any problem with that coin becomes everyone’s problem at once. Commentary on the trading firm’s report called this a hidden systemic risk, meaning one that could spread across the whole system.
- Issuer trouble. That commentary named three scenarios: a regulatory action against the issuer, a loss of the dollar peg, or a long outage. Any of them could leave agents with no practical way to settle payments. Reliance on a single issuer concentrates operational, regulatory and reserve risk.
- No ready fallback. Most agents are configured for one coin. Switching takes code changes, new seller agreements and liquidity in another token. That cannot happen in the middle of a crisis.
- Thin, top-heavy activity. If a few thousand wallets account for most real spending, the market is smaller than headlines suggest. A handful of large users leaving could change the picture fast.
- Builder lock-in. Every seller, payment-checking service and agent framework built around one coin makes it harder for the next builder to choose another. Concentration tends to reinforce itself unless someone deliberately designs for a second option.
- Soft numbers. Counts differ by source, method and what gets filtered out. Even write-ups of the same report disagree on the average payment size. Treat every share on this page, including ours, as a snapshot.
None of this means the coin is unsafe today. It means the agent economy has a single point of failure it has not yet planned around.
The trust takeaway
One coin won agent payments by being first and easiest, not because anyone decided it should. That is efficient while it works. It is fragile if it stops.
If you run or rely on an agent that pays, ask:
- Which stablecoin does the agent hold, and who issues it? Know exactly what is in the wallet.
- Can it pay in a second asset? An agent that can only settle in one coin has no plan B.
- What happens if that coin loses its peg or is paused? Find out whether the agent stops, keeps paying, or tries to swap.
- Who is checking the payments? If the issuer also runs the payment-checking service, one outage hits both.
- How much sits in the wallet at once? Concentration risk is smaller when the balance is small.