What to Know
- AI agents have made more than 138 million transactions through Coinbase’s x402 protocol on the Base blockchain.
- The payments were used to buy data, computing power and access to AI models, highlighting the rise of autonomous onchain commerce.
- U.S.-linked wallets accounted for 88.1% of geographically identified payments for AI model inference.
- Chinese-linked wallets accounted for just 0.1% of identified AI model inference payments, but roughly 90% of identified payments for security and code-analysis services.
- More than 87% of analyzed x402 payments were worth between one-hundredth of a cent and 10 cents.
- In July, x402 processed roughly 75 million payments worth just $24 million over 30 days.
- Payment networks and infrastructure companies including Visa, Mastercard and Stripe have backed the x402 protocol as interest grows in agentic transactions.
- Wallet behavior has changed sharply, with the average time from funding to first x402 payment falling from 73 days in late 2025 to 18 minutes in Q3 2026.
AI Agents Move From Conversation to Commerce
AI agents are beginning to act less like passive software tools and more like participants in digital markets. Through Coinbase’s x402 protocol, autonomous programs have now made more than 138 million crypto transactions, using blockchain-based payments to purchase data, computing power and access to AI models. The activity points to a fast-emerging category of onchain commerce in which software can pay for digital resources without a human approving every transaction.
The shift matters because the internet economy has traditionally been built around human users, bank cards, subscriptions and platform accounts. AI agents do not fit that model neatly. They may need to request a fraction of a service, buy a single unit of data, query a model for a narrow task or pay for a brief security check. These actions can be too small, too frequent or too automated for legacy payment rails. Crypto networks, particularly when paired with stablecoin-style settlement and programmable wallets, are being tested as a way to make these machine-to-machine transactions practical.
The x402 protocol has become a focal point for that experiment. Incubated by Coinbase and running through Base, it allows AI agents to make automated payments for online services. The protocol’s growth has attracted attention from major payment and technology firms, with Visa, Mastercard, Stripe and other payment giants backing the framework. Coinbase has also begun allowing business customers to accept payments from AI agents, further signaling that agentic commerce is moving from concept to early infrastructure.
U.S.-Linked Wallets Dominate Paid AI Model Access
The most striking geographic split appears in payments for AI model inference, the process of running a model to produce an output. U.S.-linked wallets accounted for 88.1% of geographically identified payments in that category, while Chinese-linked wallets accounted for just 0.1%. That gap suggests the United States is currently the dominant identified market for paid AI model access through x402, at least within the transactions that can be geographically attributed.
One possible explanation is that Chinese developers may rely more heavily on open-weight models such as DeepSeek and Qwen, which can reduce the need to purchase metered inference services from third-party providers. That interpretation remains cautious rather than definitive. The data measures identifiable activity through x402, not the entire global market for AI payments, and it does not prove that any region’s developers are avoiding paid AI services more broadly.
Still, the contrast is important for crypto market participants because it shows that early agentic payment flows are not evenly distributed. The United States appears to be leading in onchain payments for AI model usage, while China-linked activity is concentrated elsewhere. For investors and builders tracking crypto adoption, that could matter as much as raw transaction growth. A market dominated by tiny automated payments may behave differently depending on whether the demand comes from model inference, security tooling, data access or developer infrastructure.
China-Linked Wallets Lead Security and Code-Analysis Payments
Chinese-linked wallets showed a different pattern, accounting for roughly 90% of geographically identified payments for security and code-analysis services. Much of that activity was directed toward AgentLisa, an AI-powered smart contract auditing platform. That suggests China-linked usage within x402 is not absent, but rather concentrated in a separate category from paid AI model inference.
Security and code-analysis services are natural candidates for agentic crypto payments. Smart contract auditing, vulnerability screening and automated code checks can be broken into repeatable digital tasks. An AI agent could submit code, pay for a scan, receive a response and continue iterating without waiting for a conventional billing flow. In crypto markets, where software vulnerabilities can directly affect funds, faster and more granular access to security tooling could become a meaningful use case.
At the same time, the category mix shows why broad conclusions should be avoided. A low share of AI model inference payments from one region does not necessarily mean weak use of AI agents or weak use of onchain payments overall. It may mean different developer habits, different model preferences, different service availability or different commercial priorities. FXCOINZ views the x402 data as an early snapshot of behavior rather than a final map of where agentic commerce will settle.
Micropayments Remain the Core Use Case
The economics of the transactions are just as important as the count. More than 87% of analyzed x402 payments were worth between one-hundredth of a cent and 10 cents. That range is far below what traditional card networks can usually process economically, especially when fixed fees, fraud checks and chargeback systems are taken into account. Crypto rails are attractive in this context because they can support highly granular payments that would be impractical under conventional card infrastructure.
For AI agents, micropayments are not a side feature. They may be central to how autonomous software consumes the internet. An agent might pay a tiny amount for a single data point, a brief compute burst, a short model response or a narrow verification task. If millions of agents perform such actions continuously, the number of payments can rise quickly even when the total value remains modest.
That dynamic is already visible in the x402 figures. In July, the protocol processed roughly 75 million payments worth just $24 million over 30 days. The number of transactions was large, but the value attached to them remained small. This supports the view that agentic commerce is currently more about payment frequency and automation than large-volume settlement.
Strong Growth, But Demand Questions Remain
The transaction figures offer evidence that AI agents are using crypto payment systems, but they do not settle the question of genuine commercial demand. High transaction counts can reflect real usage, experimentation, subsidized activity, testing patterns or behavior designed to inflate network metrics. Earlier research also raised questions about artificial activity on the network, making it important to examine wallet behavior rather than transaction totals alone.
Some signs point toward more sustained activity. Among wallets first funded in the second quarter of 2026, 89.3% received additional funding within 30 days. That compares with 70.6% for wallets first funded in Q3 2025. The analysis excluded wallets that made payments on only one day, which helps filter out some short-lived or one-off activity. A higher rate of additional funding may indicate that more wallets are being used repeatedly rather than abandoned after initial testing.
The speed of activation has also changed dramatically. The average time between a wallet receiving funds and making its first x402 payment fell from 73 days in late 2025 to just 18 minutes in Q3 2026. That suggests newer wallets are being created or funded with more immediate use in mind. In other words, wallets may increasingly be part of a working agent payment flow rather than dormant infrastructure waiting for future deployment.
Why Stablecoin-Style Payments Fit AI Agents
Stablecoins are widely viewed by market participants as well suited to early agentic payments because they combine crypto-native settlement with a value reference that is easier for businesses and software systems to price. AI agents buying compute, data or model access need predictable units of account. Volatile assets can complicate pricing for small services, while stablecoin payments can make a one-time query or micro-service easier to meter.
Unlike traditional card payments, stablecoin-based transfers can support transactions worth fractions of a cent. That capability is especially relevant when the payment is not for a monthly subscription or a large purchase, but for a tiny digital action. If AI agents become a meaningful share of internet users, the ability to settle millions of small, automated payments could become a competitive advantage for crypto networks.
Industry figures have suggested that AI agents could become major blockchain users in the years ahead. Yat Siu, co-founder and chairman of Animoca Brands, has estimated that up to 100 billion AI agents may transact on the internet over the next few years and that they will use crypto. Such projections remain forecasts, not established outcomes, but they capture the scale of the opportunity that crypto infrastructure providers are trying to address.
What This Means for Crypto Markets
For crypto markets, the rise of x402 activity reinforces a broader narrative: blockchain demand may increasingly come from software, not only people. If every AI agent can function as a new blockchain user, the addressable market for payments, wallets, stablecoins and developer infrastructure could expand substantially. That does not mean the market has already reached maturity. The current payments are mostly tiny, and total value remains modest relative to the transaction count.
The key question is whether agentic payments become embedded in real commercial workflows. If businesses, developers and AI platforms rely on x402-style payments for compute, data access, inference and auditing, the activity could become durable. If the activity is mostly testing or promotional usage, headline transaction counts may overstate the economic significance. FXCOINZ expects market participants to watch repeat funding rates, transaction value, service diversity and wallet retention closely as better indicators of adoption.
For now, the data shows that autonomous software is starting to develop spending habits onchain. The United States leads identified paid AI model access, China-linked wallets dominate identified security and code-analysis payments, and micropayments remain the core economic pattern. The early shape of agentic commerce is therefore not a single global trend but a set of specialized behaviors forming around digital services that are easy for software to buy, consume and repeat.
Frequently Asked Questions (FAQs)
What is x402?
x402 is a crypto payments protocol incubated by Coinbase that enables AI agents to pay automatically for digital services such as data, computing power and AI model access.
How many crypto payments have AI agents made through x402?
AI agents have made more than 138 million transactions through the x402 protocol on Coinbase’s Base blockchain.
What are AI agents buying with crypto payments?
They are using crypto to buy data, computing power, access to AI models and services such as security and code analysis.
Which region leads identified payments for AI model inference?
U.S.-linked wallets lead identified AI model inference payments, accounting for 88.1% of geographically identified activity in that category.
How much of identified AI model inference activity came from Chinese-linked wallets?
Chinese-linked wallets accounted for 0.1% of geographically identified payments for AI model inference, though they were much more active in security and code-analysis payments.
Why are most x402 payments so small?
Many AI agent transactions involve narrow digital tasks, and more than 87% of analyzed x402 payments were worth between one-hundredth of a cent and 10 cents.
Does high transaction count prove strong commercial demand?
No. Transaction counts show usage, but they do not prove how much activity reflects durable commercial demand, experimentation or other behavior.
Why are stablecoin-style payments important for AI agents?
Stablecoin-style payments can support tiny, automated transactions with more predictable pricing, making them useful for agents buying small units of digital services.
What should crypto market participants watch next?
Key signals include repeat wallet funding, transaction value, service diversity, retention and whether AI agents continue using crypto payments in real commercial workflows.