North America: Inside the Growth of Agentic Commerce

What it claims

Chainalysis reads on-chain x402 data as evidence that agentic payments are moving from experimentation to provisioned, recurring budgets. x402 is a payment protocol Coinbase published in May 2025; Base is the most-used chain, with over 138M x402 transactions by the end of the period. Most payments are micropayments: over 87% fall between $0.0001 and $0.10, and the most common size is one to ten cents. Circle describes the expected split as "card rails for consumer-prompted agent payments, stablecoin rails for machine-to-machine settlements that are too small for card economics." USDC is named as a leading agent currency.

Evidence

• Window. The x402 analysis covers the 16 months ending September 2026, longer than the report's standard 12-month period.
• Re-funding. Wallets are grouped by the quarter they were first funded. The share re-funded within 30 days rose from 70.6% (Q3 2025 cohort) to 89.3% (Q2 2026, the latest cohort whose window has closed). Only wallets that sent on two or more separate days are counted, so one-off experimenters are excluded and the figure overstates retention across all wallets.
• Time to first spend. The median gap between first funding and first x402 payment fell from 73 days in late 2025 to 18 minutes in Q3 2026. The share of wallets spending within an hour of funding rose from 4% to 67% over the same span.
• What agents buy. Early volume was fads such as the memecoin PING. Later spend went to code review, financial intelligence, data and search APIs, gaming, token issuance and gambling, agent frameworks and tooling, and model inference, which is one of the largest categories.
• Geography (share of geolocated agent-to-service transactions):

The page offers "AI nationalism" and free Chinese open-weight models (DeepSeek, Qwen) as one possible reason for the inference gap, against metered US models (ChatGPT, Claude). This is a hypothesis, not something the data tests.

How to use it

The report gives no protocol or integration details. Its actionable content is a compliance model for anyone deploying agents that pay:

• Accountability rests with whoever deployed and funded the agent, not with the agent.
• Individual payments are too fast, small and frequent to review, so there are three checkpoints instead: 1. the funding event, which identifies the sender and records a reviewable human budget decision; 2. the receiving service, a known counterparty; 3. the behavioural pattern, since regular machine spending makes anomalies easier to detect than in human payment flows.
• The authors warn that frameworks relying on per-transaction review "will either block agentic commerce entirely or miss it completely."

Teams building agent payments should expect to be screened at funding, counterparty and pattern level, not per call.

Limits

• The source is a vendor blog that quotes Coinbase, Circle and BlackRock. Coinbase incubated x402.
• Experts quoted still call the space an "experimentation phase."
• Country attribution covers only transactions that can be geolocated with Chainalysis's standard method, which the page does not explain.
• The data is limited to x402. The report does not cover other agent payment rails or card-based agent payments.
• The non-AI sections (stablecoin exports, tokenized Treasuries at $15.42B of a $36.07B RWA market, Canadian remittances) are outside this catalog's scope.