AI-agent payments on Coinbase’s Base expanded by several orders of magnitude between the Q2 2025 reference point and Q3 2026. Chainalysis says the activity is overwhelmingly made up of tiny payments—most below 10 cents—because stablecoin settlement can handle amounts that card networks cannot process economically. The scale is notable, but Chainalysis still describes agentic commerce as experimental rather than mature.
What Chainalysis measured on Base
Chainalysis tracks cumulative transactions using Coinbase’s x402 protocol, with Base as the principal network in the report. The series moved from roughly 1,500 transactions in mid-2025 to more than 138 million by the end of the Q3 2026 period. “Mid-2025” is the report’s reference point for the Q2 2025 starting figure; Q3 2026 ended in September 2026.
As an Amazon Associate I earn from qualifying purchases.
| Reference point | Cumulative x402 transactions | What the figure represents |
|---|---|---|
| Mid-2025 (Q2 2025 reference) | Approximately 1,500 | Early activity recorded on Base |
| End of Q3 2026 | More than 138 million | Cumulative activity reported by Chainalysis |
The increase shows that machine-payment experiments can scale quickly once developers have a common payment method and inexpensive settlement network. It does not, by itself, show how many businesses are earning money, how many people are using the services, or whether the transactions are profitable.
What Coinbase’s x402 protocol does
Payments made by software rather than a person
x402 is a Coinbase-developed machine-payment protocol that lets software agents pay for web resources directly in their requests. An agent can obtain a digital service—such as data, an API response or model inference—and attach payment as part of that machine-to-machine interaction, instead of requiring a person to approve every small charge.
Why Base matters
Base provides the network environment used for the activity measured in the report. The protocol and the network address different parts of the problem: x402 supplies a payment convention that software can use, while Base supplies blockchain settlement for the resulting transfers.
Why 87% of payments were below 10 cents
In the 16 months ending September 2026, more than 87% of payments fell between one-hundredth of a cent and 10 cents. The most common range was 1 to 10 cents. These are not average transaction values; they are the distribution of individual payments observed during that period.
| Payment range | Evidence reported by Chainalysis |
|---|---|
| One-hundredth of a cent to 10 cents | More than 87% of payments |
| 1 to 10 cents | Most common range |
Card networks were designed around consumer purchases where fixed processing costs can be absorbed into a larger ticket. A charge worth a fraction of a cent, or even a few cents, is difficult to handle profitably once authorization, network and merchant costs are included. Chainalysis reports that stablecoin settlement can cost less than a cent, making pay-as-you-go access to APIs, data and inference economically practical at much smaller amounts.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Circle summarized the emerging division of labor this way: “card rails for consumer-prompted agent payments, stablecoin rails for machine-to-machine settlements that are too small for card economics.” That distinction describes payment suitability, not a guarantee that every stablecoin transaction will cost less than a cent in every circumstance.
Are AI agents really using crypto for micropayments?
The x402 figures are evidence of crypto-based machine payments occurring at scale on Base. They are not evidence that most AI agents, consumers or online merchants have adopted the model. The transactions could include repeated automated calls from a relatively small set of software projects, and the reported count does not identify the economic value generated by those calls.
Signs of repeat use—and why adoption is still experimental
Chainalysis found a higher rate of repeat wallet funding among newer cohorts. The share of wallets receiving a second funding within 30 days rose from 70.6% for wallets first funded in Q3 2025 to 89.3% for wallets first funded in Q2 2026. The latter cohort’s 30-day observation window had closed by the time of the analysis.
| Wallet cohort | Wallets receiving a second funding within 30 days | Qualification |
|---|---|---|
| First funded in Q3 2025 | 70.6% | Measured over the cohort’s 30-day window |
| First funded in Q2 2026 | 89.3% | 30-day window had closed |
Repeat funding suggests that more wallets were used again rather than abandoned after one transaction. It is a persistence signal, not a direct measure of active users, revenue, consumer satisfaction or sustainable margins. For that reason, Chainalysis characterizes agentic commerce as being in an experimentation phase even while reporting signs of maturation.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →What the surge means for each participant
For developers building agents
Very small transaction sizes make it possible to let an agent buy data or compute only when needed. Developers still need controls around which services an agent may call, how much it may spend, and how its wallet is funded. A high transaction count can otherwise mask runaway automation or low-value requests.
For API and data providers
x402 creates a way to charge per request instead of relying only on subscriptions or large prepaid packages. Providers must still operate reliable endpoints and define prices that cover infrastructure, fraud prevention and support; the reported transaction totals do not establish those business economics.
For consumers and personal-finance readers
These payments are generally machine-to-machine rather than ordinary card purchases. If an agent uses a funded wallet, the practical questions are authorization, spending limits, transaction records and the ability to stop further requests. Stablecoin settlement may make tiny charges feasible, but it does not remove volatility, custody, regulatory or dispute considerations.
How to interpret the 138-million figure
The headline demonstrates rapid growth in a specific protocol-and-network combination over a defined period. It is strongest as evidence that developers are testing and repeatedly using an inexpensive rail for automated digital purchases. It should not be read as proof that agentic commerce has reached broad consumer adoption, that all 138 million transactions were paid by independent agents, or that the model is already profitable.
A complete comparison with cards or other blockchain payment systems would require consistent data on settlement cost, latency, programmability, wallet controls, geographic and regulatory reach, endpoint reliability and repeat usage. The Chainalysis figures establish Base’s x402 activity, but they do not supply an apples-to-apples comparison across every competing payment rail.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
