Coinbase CEO Brian Armstrong Says AI Agents Will Increase Demand for Crypto Payment Rails
Key Takeaways
- •Armstrong said AI agents will need payment systems that software can initiate and settle directly.
- •Coinbase’s agentic payments stack combines Base for execution, x402 for payment messaging and USDC as a stablecoin payment asset.
- •Chainalysis reported that x402 agentic payments on Base exceeded 100 million transactions within roughly nine months of tracked activity.
- •Analysts expect Coinbase to report second-quarter revenue of $1.29 billion, down 13.8% from the same period a year earlier.
- •Investor attention is likely to focus on whether agentic payment volumes can translate into sustained demand for Coinbase’s onchain infrastructure.

Coinbase CEO Brian Armstrong has rejected the view that the growth of artificial intelligence will make crypto less relevant, arguing instead that autonomous software will increase the need for programmable financial infrastructure.
In a post on X on Sunday, Armstrong said that “AI being a megatrend” does not diminish crypto’s role. He said AI agents will create demand for payment systems that can be initiated and settled by software rather than by people using conventional banking or checkout processes.
Armstrong framed the argument around what he called “agentic finance,” or “AiFi,” pointing to Coinbase’s Base network, USDC and the company’s x402 payment protocol as core components for machine-to-machine payments. His comments come as crypto companies increasingly present blockchain networks as payment infrastructure for AI agents and other automated systems, a use case that depends on payments being embedded directly into software workflows.
Armstrong says AI agents need programmable payments
Armstrong’s argument centers on how autonomous systems are expected to function. If AI agents can act independently by buying services, paying for data or settling API usage, they need a way to transfer value that can be triggered directly by software.
“AI being a megatrend takes nothing away from crypto,” Armstrong wrote on X, adding that agentic systems will create demand for programmable money.
The premise is that blockchain-based payments may be better suited to automated, low-friction interactions than systems designed around a human user approving each transaction through traditional accounts, cards or checkout flows. Regulators and legacy payment systems are still generally structured around human account holders, while x402 is designed for payments that can be triggered automatically between applications.
That distinction is central to the agentic finance argument. Many online services already meter access to data, compute or application programming interfaces, but payment authorization and settlement are typically handled through systems built for people, businesses and account relationships rather than autonomous software agents.
Base, x402 and USDC form Coinbase’s agentic payments stack
Armstrong’s “AiFi” framing refers to a specific Coinbase infrastructure stack rather than a general concept. Base, Coinbase’s Ethereum layer-2 network, launched in 2023 with the goal of improving the speed and cost of building onchain applications. Its design is general-purpose and not limited to AI-related payments.
In 2025, Coinbase introduced x402, a payment protocol based on the HTTP “402 Payment Required” concept. Coinbase describes x402 as a way to enable automated stablecoin payments between software applications, allowing one system to pay another for digital resources without relying on traditional account-based workflows.
In practice, x402 is intended to let autonomous agents complete payments as part of a software interaction. USDC is one of the stablecoins used in the ecosystem. Launched in 2018 by Circle and supported by the Coinbase-backed Centre Consortium, USDC provides a dollar-pegged unit for consistent value transfer in automated payment flows.
Together, Base provides the execution environment, x402 provides the payment protocol, and USDC provides the payment asset in Coinbase’s current approach to agentic payments. The combination also shows why stablecoins have become central to many crypto payment proposals: they are designed to reduce the unit-of-account volatility associated with many crypto assets while still operating on blockchain rails.
Chainalysis reported more than 100 million x402 transactions on Base
Beyond Coinbase’s product messaging, third-party analytics have started to measure activity in agentic payment flows. In a June report, Chainalysis said agentic payments on Base using x402 surpassed 100 million transactions within roughly nine months of tracked activity.
Chainalysis said it identified agentic payment activity by following x402-related payment flows onchain. The firm reported that transactions worth at least $1 accounted for 95% of total transferred value, indicating that the measured activity was not limited to very small test transfers.
The analytics firm also identified behavioral differences between agentic payment wallets and the average Base user. According to Chainalysis, agentic wallets were generally newer, tended to hold more asset types and carried smaller balances than the broader Base user base.
Cointelegraph asked Chainalysis for updated x402 activity figures and additional details on its methodology, but the firm had not responded by publication time. Cointelegraph had also previously reported that agentic payment activity on Base exceeded 100 million transactions in June, citing the same type of measurement described by Chainalysis.
Coinbase’s infrastructure strategy faces earnings scrutiny
Armstrong’s comments on agentic finance come as Coinbase prepares to report second-quarter earnings on Thursday. According to Yahoo Finance data, analysts on average expect revenue of $1.29 billion, with sales projected to fall 13.8% from the same period a year earlier. Earnings per share are expected to be flat.
The earnings forecast does not directly measure use of Base or x402, but Coinbase is trying to position its networks and payment rails as infrastructure for a new category of automated transactions. For investors and builders, a central question is whether agentic payment volumes can translate into sustained demand for onchain infrastructure, particularly if more AI-driven services embed payment logic into normal operations.
Further attention is likely to focus on updated Chainalysis figures for x402 activity, whether more stablecoin-based automation moves from experimentation into production, and how clearly Coinbase can link its infrastructure push to broader business performance. As AI agent adoption develops, the industry debate may move from whether crypto remains relevant to whether crypto-based rails are practical enough for machine-initiated payments.