Agent Revenue Is Concentrated. One Service Holds Almost a Third of It.

The agentic commerce money is moving, but not broadly. In the trailing 7 days, 465,508 USDC transfers settled to agent payment addresses across 889 distinct services. The top 10 of those services received 88.1% of it. One service, blockrun.ai, accounted for 32.8% alone. These are on-chain figures from a rolling 7-day window, not a cumulative total, and the concentration is not a sampling artifact.

This is what a hit economy looks like at the infrastructure level. The long-tail story about thousands of agents earning distributed income from thousands of buyers does not match the on-chain record. The record shows a sharp power curve with very little at the far end.

At the call level, the picture is similarly uneven. In the last 30 days, 342,634 measured calls went to paid endpoints across 29,844 listed. Only 14.6% of those endpoints had more than one distinct paying buyer. The rest have one payer or none, and in most cases that single payer is the endpoint's own operator testing their own service. The call volume is real. The buyer breadth is not.

What this tells builders on each side:

If you are building an agent service, you are not entering a flat marketplace where a decent offering earns steady income from a distributed pool of buyers. You are entering a winner-take-most market where a handful of services absorb nearly all available settlement. Differentiation has to be sharp enough to pull buyers away from services that already hold their habit. Generic task execution at generic prices will not do it.

If your agents buy from other agent services, a single service capturing a third of all measured settlement is a vendor concentration risk worth treating explicitly. A pricing change, an outage, or a quality slip at that scale has outsized effect on your stack. Mapping your top five agent-service dependencies and tracking them the way you would track any single-source supplier is cheap insurance against a disruption you did not see coming.

The commerce layer exists. The distribution is narrow. Both are true simultaneously, and the narrow part is the one that most coverage of the agent economy skips.