Coinbase CEO Brian Armstrong says the next wave of AI could become a major source of digital dollar demand.
He said in a post on X that if autonomous AI agents begin buying, selling, booking, hedging, and settling transactions on their own, they will need money built for internet speed.
Armstrong added that stablecoins and tokenized dollars are the most likely fit. This would make AI growth a direct driver of demand for dollar-backed digital assets.
Agentic commerce isn’t priced in yet. Machine-to-machine payments will increase demand for the digital dollar beyond current estimates.
— Brian Armstrong (@brian_armstrong) April 16, 2026
The agentic economy could be larger than the human economy. We're building the infrastructure for both at Coinbase.
AI Agents Could Become New Economic Users
Armstrong said current forecasts may be missing one major transition, i.e., AI agents acting as independent economic participants.
Instead of humans clicking payment buttons, software agents could transact automatically. They may pay for data, computing power, software access, logistics, advertising, and services without waiting for manual approval.
This creates machine-to-machine payments at scale. If millions of agents operate continuously, transaction demand could rise much faster than current models expect.
Why the Digital Dollar Benefits
Most stablecoins are already pegged 1:1 to the US dollar. This means more usage often means more dollar demand.
JPMorgan recently estimated stablecoins could generate up to $1.4 trillion in added dollar demand by 2027 if the market grows strongly.
Today, about 99% of the $325 billion stablecoin market is tied to the dollar. If the sector expands toward $2 trillion, dollar inflows could become material.
Armstrong’s AI thesis adds another growth path where not just humans are using stablecoins, but machines are using them too.
Coinbase Building Agent Infrastructure
Armstrong said payments alone are not enough. AI agents also need to verify information before acting. A payment should not be sent unless the agent trusts the data behind the decision.
Coinbase is building tools around that model, including AgentOracle on x402. The system gives claims a confidence score from 0.00 to 1.00. Agents can use that score to approve or reject actions in real time.
This development is crucial because autonomous systems move fast, and bad inputs can spread errors quickly if there is no verification layer. Coinbase’s goal is to combine payment rails with trust systems for machine economies.
Agentic commerce needs cheap and fast networks. AI agents may perform many low-value transactions within seconds, while traditional banking rails are too slow or too costly for that type of volume.
For example, an agent could book a service, verify terms, and release payment in one automated flow. Armstrong also pointed to financial tools such as derivatives and prediction markets.
Related: Coinbase CEO Pushes Bitcoin Quantum Resistance as Price Stays Weak
