Coinbase co-founder and CEO Brian Armstrong said the balance of online transactions could shift sharply, with AI agents potentially outnumbering humans as active participants. His reasoning is simple: artificial intelligence systems cannot open traditional bank accounts, but they can control crypto wallets with far fewer barriers.
That view points to a machine-native payment model rather than a human-centered one. If software systems begin handling subscriptions, purchases, sales, and data payments on their own, blockchain networks could see a large increase in transaction volume. The source article notes that AI tools are already being used for trading bots, data processing, and automated online services. Giving those systems direct payment capability changes the scale.
Why crypto wallets fit machine-driven payments
Armstrong’s argument rests on the difference between banking rails and blockchain accounts. Traditional banks are built around identity checks, paperwork, and compliance procedures designed for human users. Machines do not fit that structure. A blockchain wallet, by contrast, only requires a cryptographic key, which allows software to hold and move value directly.
In that setup, AI agents paying for compute power, APIs, datasets, or digital services would need an infrastructure that software can access natively. The article frames crypto as a likely payment method for this machine-to-machine economy.
Coinbase’s agentic wallet push
Coinbase introduced the idea of agentic wallets in February, presenting them as wallets built for autonomous agents. The company describes this as part of its AI strategy. These wallets are intended to let software systems interact with money and blockchain services without a human stepping in for each action.
According to the source, the wallets can securely hold digital assets, send payments, trade tokens autonomously, earn yield through DeFi protocols, and connect directly to blockchain-based services. Coinbase also says the tools include safety features aimed at reducing misuse while still enabling experimentation.
Efficiency gains come with security questions
The article highlights clear advantages. Automated systems could handle subscriptions, cloud-computing bills, and micropayments instantly, which may reduce cost and speed up settlement across borders. For blockchain networks, that would mean a larger flow of high-frequency, small-value transactions generated by software rather than people.
The concerns are just as clear. If safeguards are weak, large-scale machine decision-making could introduce security failures, abusive trading bot activity, or malicious software that disrupts markets. The source does not provide specific incidents, but it does stress that automated financial systems create new categories of risk as they scale.
What an AI transaction economy could look like
The article places Armstrong’s comments inside a broader AI economy thesis. Machines can operate continuously, process information instantly, and execute tasks much faster than humans. In that scenario, AI systems could one day make billions of transactions a day, paying on their own for data access, computing resources, and digital services.
From Coinbase’s perspective, AI agents, crypto wallets, and onchain payments are moving into the same product path. Trading, payments, asset holding, and yield generation are no longer framed only as human actions. They are also being designed as functions software may carry out independently.