After 138 million on-chain payments, what exactly are AI agents buying?

A status code forgotten by the HTTP protocol for thirty years is becoming the cash register for the machine economy.

The North America Crypto Adoption Report released by Chainalysis on October 14 revealed a set of data: As of September 2026, cumulative AI Agent payment transactions completed via the x402 protocol on Coinbase's Base chain exceeded 138 million. Including other chains such as Solana, the total transaction volume of the x402 ecosystem has surpassed 165 million, with payment amounts exceeding $50 million.

In these transactions, 87% were single payments ranging from one-tenth of a cent to ten cents. The most common payment range was between one and ten cents.

To put it another way: approximately five AI Agent on-chain payments occur every second, each costing less than a lollipop.

The Resurrection of a Status Code

The story begins with HTTP 402.

In 1997, the HTTP/1.1 standard defined the "402 Payment Required" status code, envisioning a future where web services could use it to require visitors to pay before accessing content. However, no payment infrastructure at the time could support micro-payments at the webpage level, so this status code was marked as "reserved for future use," shelved for over two decades.

In May 2025, Coinbase revived it. The x402 protocol is designed very simply: when an AI Agent sends a request to an API, if the server requires payment, it returns a 402 status code along with the price, receiving address, and supported tokens. The Agent's wallet automatically signs, settles on-chain, and gains access rights. The entire process requires no account registration, no API key, and no human approval.

An HTTP request, a USDC transfer, a data exchange. Three steps complete a machine-to-machine commercial transaction.

This mechanism works because stablecoins have driven settlement costs low enough. Swiping Visa for a $0.03 transaction might incur fees higher than the transaction itself. But settling with USDC on Base costs less than a cent. When individual transactions can be as small as one-hundredth of a cent, the entire "pay-per-use" business model becomes viable.

What Are AI Agents Buying?

Chainalysis's report broke down the consumption scenarios behind these 138 million transactions, bringing several categories to light:

Model inference is the largest single category. AI Agents call large language model APIs, paying per token. A single Q&A might cost $0.02, while a code generation task could cost $0.08. US wallets accounted for 88.1% of model inference payments, aligning with the market positions of leading US model providers (OpenAI, Anthropic, Google).

Data and search APIs followed closely behind. Agents pay for real-time data: price feeds, on-chain analytics, news summaries, and academic paper retrieval. Payments in this category are more evenly distributed geographically, with US wallets accounting for 23.2%.

Code audit and security analysis emerged as an unexpected highlight. Wallets originating from China accounted for roughly 90% of the transaction volume in this category, primarily driven by AgentLISA, an automated smart contract audit service. Agents submit contract code and pay per audit with USDC to receive reports.

Financial intelligence and market data are also growing rapidly. Chinese wallets represented 46.7% in this category, while US wallets made up a relatively small share. This likely reflects a reality: Chinese AI application developers are leveraging overseas data services to bridge information gaps.

An implicit geographical divide has emerged: US agent consumption leans toward the "production side" (inference, generation), while Chinese agent consumption leans toward the "analysis side" (audit, intelligence). Chainalysis attributes China's extremely low share in the inference category (0.1%) to "AI nationalism": domestic open-source models like DeepSeek and Tongyi Qianwen are available for free, eliminating the need to call overseas commercial APIs via on-chain payments.

From Speculative Tool to Infrastructure: A Shift in Behavioral Data

More noteworthy than transaction volume is the shift in user behavior.

A significant portion of x402's early growth came from PING, a Meme coin that required x402 payment to mint. This wave of speculation brought a surge of wallets but also led to a subsequent exodus of traffic. However, after the tide receded, the remaining users exhibited entirely different behavioral patterns.

Chainalysis tracked the "stickiness metrics" for wallets:

The percentage of wallets that recharged within 30 days of their first deposit rose from 70.6% in Q3 2025 to 89.3% in Q2 2026. Nearly nine out of ten wallets returned to continue using the system within a month of depositing funds.

Even more striking was the change in "time to first spend." At the end of 2025, the median time from wallet creation to the first x402 payment was 73 days. By Q3 2026, that number dropped to 18 minutes.

The proportion of wallets that began spending within an hour of depositing funds skyrocketed from 4% to 67%.

Taken together, these three datasets tell the same story: x402 is transitioning from a "new toy for developers to tinker with" to the "default payment layer in agent workflows." Wallets are no longer exploratory tools driven by human curiosity, but programmatically deployed infrastructure components. They spend immediately upon funding and recharge automatically, operating smoothly like utility bill auto-drafts.

The structure of payment amounts is also shifting. In early 2025, transactions over $1 accounted for only 49% of the total payment value; by early 2026, this figure jumped to 95%. Transactions between $0.10 and $1 plummeted from representing 46% to just 4%. Agents are no longer just making "exploratory micro-calls"; they are paying for heavier, more expensive services: renting GPUs by the hour, purchasing datasets in bulk, and subscribing to continuously updated financial data feeds.

Tech Giants Are Betting on This Track

x402 deserves serious attention not just because of rising on-chain transaction volumes, but because of the list of entities backing it.

In May 2026, AWS launched Bedrock AgentCore Payments, enabling natively supported x402 payments for AI Agents running on Amazon Cloud. Developers can select Coinbase or Stripe wallets, fund them with USDC, and allow Agents to autonomously pay for API calls, MCP server access, and inter-agent collaboration.

Coinbase and AWS jointly founded the x402 Foundation, with Cloudflare as a founding member, building native x402 support directly into Workers. Google integrated x402 into its own Agent Payments Protocol (AP2). Visa announced support for x402 through its Trusted Agent Protocol (TAP).

OpenAI and AWS co-released a technical guide for x402 payment flows. The Solana Foundation built a similar agent payment solution for Google Cloud services.

This list reads less like an "ecosystem partner roster for a crypto project" and more like a consensus checklist for technological industry infrastructure. When AWS, Google, Visa, and Stripe simultaneously appear on the supporter list of a single protocol, the conversation shifts from "can cryptocurrency achieve mainstream adoption" to "what should the payment standard be for the machine economy."

The statement from Circle in the report is telling: stablecoins are laying the rails for "machine-to-machine settlement," processing transactions too small for credit card networks to even consider. Will Su from BlackRock noted that AI Agents may create a completely new type of payment demand: continuous, high-frequency, low-value payment streams around the clock, which virtually do not exist in human economic activity.

A Still Tiny but Non-Negligible Figure

A word of caution is necessary: a total payment amount of $50 million spread across 165 million transactions averages just $0.3 per transaction. In the context of the crypto market, which routinely sees daily trading volumes of tens of billions of dollars, this is merely a figure at the level of statistical noise.

There are approximately 69,000 active AI Agents on Agentic.Market. Most of these agents are still in a phase where they can function but remain far from "large-scale commercial deployment."

But there is a notable structural pattern here: x402's growth trajectory closely mirrors early DeFi. In 2019, DeFi's TVL was under $1 billion and nobody paid much attention. Three years later, peak locked value exceeded $180 billion. What matters is not the current absolute numbers, but the shape of the growth curve and the underlying structural drivers.

The number of AI Agents is growing exponentially, and each one needs to autonomously procure and pay for external services. Traditional payment systems (credit cards, bank transfers, PayPal) were never designed for "autonomous machine payments"; they require human identity verification, accounts, and minimum transaction amounts. Stablecoins + x402 fill a gap that existing financial infrastructure completely fails to cover.

As you read this article, thousands of AI Agents are likely already quietly completing another round of micropayments on Base. They don't sleep, eat, or take weekends off. They simply follow code instructions, purchasing the world's data and computing power penny by penny on-chain.

This could be the first large-scale use case crypto has found that does not rely on speculation.

Join the official Coincamps community:

X: https://x.com/coincamps

Telegram: https://t.me/coin_camps