Stripe says blockchains may eventually need to handle as many as 1 billion transactions per second to support the rise of artificial intelligence agents.
In their annual letter published on X on Tuesday, CEO and co-founder Patrick Collison and co-founder John Collison reviewed Stripe’s performance in 2025 and shared their outlook for the near future.
A central theme of the letter was the growing adoption of AI agents and how their widespread use could reshape digital commerce. The Collison brothers argued that blockchain activity could surge dramatically as AI agents increasingly handle online transactions on behalf of users and businesses.
However, they cautioned that current blockchain infrastructure is not equipped to meet that level of demand, warning of a major scalability shortfall.
They pointed to an incident last year when a memecoin trading frenzy on a major blockchain caused significant network congestion, delaying payouts for one Bridge user by more than 12 hours and driving per-transaction costs up by 35 times. While such disruptions are already problematic, they said, the strain will only intensify as transaction volumes grow in an AI-driven economy.
“In our view, agents will most likely soon be responsible for most internet transactions, and we will likely need blockchains that support more than one million — or even one billion — transactions per second.”
Data from Chainspect shows that Internet Computer and Solana currently lead the industry in transaction throughput, processing around 1,196 and 1,140 transactions per second (TPS), respectively.
They are the only two networks consistently handling more than 1,000 TPS. At peak performance, Internet Computer has reached 25,621 TPS, while Solana has processed as many as 5,289 TPS.
Even so, their theoretical maximum capacities remain limited — approximately 209,708 TPS for Internet Computer and 65,000 TPS for Solana — figures that fall far short of the billion-TPS scale Stripe believes could eventually be required.
AI commerce moves beyond hype
Alongside their projections for surging transaction demand, Stripe executives Patrick Collison and John Collison described the evolving role of AI agents in online commerce.
They argued that AI-driven commerce has progressed beyond the “pure hype” stage and entered a period of active development and real-world testing.
The pair outlined five levels of AI agent capability. The first two include completing web forms and performing descriptive search — enabling agents to deliver results based on contextual descriptions rather than rigid keywords or attributes.
According to the executives, AI agents are currently “hovering on the edge” of these first two stages.
The remaining three levels involve more advanced features: persistence, where agents retain user preferences and requirements; delegation, such as independently managing tasks like grocery shopping; and anticipation, in which agents proactively solve problems or organize schedules without explicit prompts.
“As with the early internet, the future success of agentic commerce is contingent on universal interoperability,” they wrote. “Our ascent through the five levels depends on our ability to work together.”
“If all goes well, the little critters won’t be cooped up in walled gardens, but will be zooming down the wide-open protocol highways.”
