Coinbase chief executive Brian Armstrong said artificial intelligence and crypto should not be viewed as competing technology trends, arguing instead that crypto will serve as financial infrastructure for autonomous AI agents.

Armstrong said he expects AI agents to conduct more daily transactions than humans through crypto-based systems. Coinbase is developing what it calls Agentic Finance around x402, Base, USDC, wallets, trading tools and business payments, while recent research has questioned x402 adoption metrics and identified security weaknesses in facilitator-led machine payment systems.

In a July 27 post on X, Armstrong said agents “will eventually transact far more per day than all humans combined,” presenting the claim as a forecast for payments and other online financial activity.

"If you're in crypto, pivot to AI." I used to hear versions of this, and it's the wrong way to think about the world. It's zero sum, scarcity thinking. Crypto is a general purpose technology. It's infrastructure, the same way electricity or the internet is infrastructure. It… — Brian Armstrong (@brian_armstrong) July 26, 2026

https://x.com/brian_armstrong/status/2081504081902780564

Armstrong said AI agents cannot use traditional banking services in the same way that people or companies can. Such agents may need to pay for data, software, computing power and other agents without direct human approval. He said blockchains and stablecoins can provide fast, programmable and global settlement for those transactions.

Armstrong describes crypto as AI’s financial layer

Armstrong’s post addressed the view that crypto companies should abandon blockchain work and shift toward AI. He rejected that framing, saying it reflects “zero sum, scarcity thinking.” In his view, AI provides programmable intelligence, while crypto provides programmable money.

He described the combined model as “Agentic Finance,” or “AiFi,” and said Coinbase is building products for that market. The Coinbase chief did not provide a date by which agents could exceed human transaction counts, nor did he estimate the payment value such activity might represent.

His statement focused on transaction frequency. That number could increase if software agents make small payments for every API call, data request, computing task or interaction with another service. The model differs from typical consumer payments, which generally involve fewer transactions with larger payment amounts. It also explains why Coinbase is emphasizing machine-readable payment flows rather than only consumer checkout tools.

Coinbase broadens Agentic Finance products

Coinbase has already released several products designed for autonomous software. In June, the company launched Coinbase for Agents, which connects AI systems to user accounts through a command-line interface and Model Context Protocol tools. Users can set limits while agents trade crypto, monitor markets, rebalance portfolios and carry out defined financial tasks.

On July 23, Coinbase expanded the service by adding live market data and plain-language conditional commands. It also added x402 support for Coinbase Business, enabling companies to accept USDC payments initiated by AI agents. A new developer kit allows websites and API providers to add x402 payment acceptance with a small amount of code. As crypto.news reported, the rollout covers businesses, users and developers building agent services.

Coinbase also introduced Agentic.market in April. The marketplace allows agents to find and pay for data, search, computing, inference and trading tools. Coinbase describes it as a discovery layer for machine commerce because agents can locate services and buy access without a conventional subscription or manually issued API key.

x402, Base and USDC form the core of Coinbase’s strategy

Armstrong said Coinbase pioneered the model through x402, Base and USDC. x402 adapts the HTTP 402 “Payment Required” response so a website or API can request payment during an internet interaction. A wallet signs the payment, a facilitator checks it, and the service delivers the requested resource after approval. The status code has long been reserved for payment use, and x402 is one attempt to turn that unused web primitive into a working payment flow for APIs and online services.

Coinbase uses USDC as the main payment asset in many x402 products, while Base provides low-cost blockchain settlement. The protocol also supports other networks and assets. Coinbase’s developer documentation lists support across Base, Solana, Polygon, Arbitrum and World, depending on the payment method and facilitator.

As crypto.news previously reported, Coinbase has said AI agents already use x402 to buy data, computing resources and digital services. Related coverage also reported that Coinbase Business customers can now receive USDC directly from agents. Together, the products translate Armstrong’s broader claim into a commercial strategy built around wallets, stablecoins, trading access and payment tools.

Researchers question adoption data and security

Armstrong’s claim remains a prediction. Current totals do not demonstrate that independent AI agents have created an economy larger than human commerce. A July paper examining x402 activity on Base found that transaction counts were highly concentrated. The authors also said some payments were internal or inexpensive to generate, making headline transaction totals a weak measure of adoption.

Separate July research tested 15 x402 facilitators and reported rule violations across every system examined. The researchers described risks involving unpaid services, asset theft, denial of service and gas abuse. They said affected providers, including Coinbase, received the findings and adopted fixes. The papers remain preprints and have not completed peer review.

Regulators are also examining automated finance. Bank of England Deputy Governor Sarah Breeden said in June that existing rules did not account for autonomous agents. She raised the possible use of guardrails, circuit breakers and stronger recovery systems if AI-driven trading or payments create wider problems.

Coinbase continues to position crypto as the payment layer for machine activity. Armstrong’s latest post links that strategy to a broader claim: AI agents may become economic actors that hold funds and transact at high frequency. Whether they overtake humans in transaction counts will depend on real usage, security, regulation and demand for paid machine services.