Coinbase CEO Brian Armstrong Expects AI Agents to Drive More Than 99.9% of Transactions Within a Decade
Key Takeaways
- •Brian Armstrong forecasts that AI agents will conduct more than 99.9% of all transactions within the next decade as autonomous software increasingly pays other machines for services.
- •Coinbase is developing financial accounts, payment tools, and marketplaces for agent-to-agent transactions, with early activity running on Base using the USDC stablecoin.
- •Crypto infrastructure supports programmable, 24/7 payments of fractions of a cent, whereas card systems designed around human cardholders and per-transaction fees cannot handle such microtransactions.
- •Armstrong is already using AI agents to trade cryptocurrencies across different portfolios, though he acknowledged the results have been mixed so far.
- •Coinbase is building tools that let users create trading strategies, publish performance records, and let others auto-copy them, with spending limits and account permissions managing risk.

Coinbase CEO Brian Armstrong expects AI agents to account for more than 99.9% of all transactions within the next decade, as autonomous software increasingly conducts payments and purchases without direct human involvement — a scenario in which machine-driven activity, rather than human spending, makes up the overwhelming majority of transactions.
According to Armstrong, AI agents will eventually outnumber humans, and as they take on more tasks independently they will increasingly need to pay one another for services. He argued that this dynamic could create demand for payment infrastructure capable of handling millions of small, real-time transactions between machines.
Coinbase is developing financial accounts, payment tools, and marketplaces designed to facilitate these agent-to-agent transactions. Much of the early activity is running on Base, the Ethereum layer-2 blockchain developed by Coinbase, using USDC, a stablecoin pegged to the US dollar.
Unlike traditional card networks, crypto infrastructure can support programmable, round-the-clock payments and very small transactions between autonomous agents, making it well suited to commerce carried out entirely between software programs. Card systems, built around human cardholders and per-transaction fees, are not designed for payments of a fraction of a cent.
Armstrong said he is already using AI agents to trade cryptocurrencies across different portfolios, although he acknowledged that the results have been mixed so far.
The company is also developing tools that would allow users to create trading strategies, publish records of their performance, and let others automatically copy them. Spending limits and account permissions would be used to manage risk, keeping human account holders in control of what an agent can spend.
The bet underlying these efforts is that AI agents could become a major new source of on-chain transaction activity, extending cryptocurrency's role beyond human-driven trading and payments. The rollout of these agent-facing tools, and the volume of agent-to-agent activity that builds on Base, are the observable indicators to track.
The remarks were reported by BitcoinKE.