I just read an interesting statement from Stripe's leadership about the future of blockchain and AI agents. Patrick and John Collison raise a question in their annual letter that few discuss openly: for AI agents to truly become full participants in online transactions, we need infrastructure capable of processing about a billion operations per second.

Why has this even come onto the agenda? Because right now, blockchain is clearly not ready. Remember the surge in activity with meme coins? That’s a clear example of how quickly the system begins to choke. Fees skyrocketed, delays became noticeable. And that was just a local spike of interest from regular users.

Currently, the leaders in speed are Internet Computer Protocol and Solana, which deliver approximately 1196 and 1140 TPS respectively. It sounds impressive, but when you think about the scales Collison envisions for mass adoption by AI agents, it’s just a drop in the ocean. When AI agents start not just filling out web forms but independently making complex decisions and anticipating user needs, transaction volumes will be entirely different.

This is an interesting point — the industry is shifting from loud promises to real applications. And for this transitional period, serious technical infrastructure is needed. The question is, who will be the first to create an infrastructure capable of supporting such a level of TPS? It seems this will become one of the key races in the crypto ecosystem.