Coinbase CEO Brian Armstrong says his company has become one of the most AI-enabled businesses in the world, and the results are showing up in how much product its engineers are shipping.
“Were in the age of the super builder,” Armstrong said, describing how AI has transformed what individual engineers at Coinbase can accomplish. He added that the company is now shipping twice as much code overall, with some engineers becoming what he called ten-times contributors who teach best practices across teams.
Reacting to the same, a user said that former Coinbase employees now working at other crypto companies have described Coinbase as being far ahead of everyone else when it comes to integrating AI into its engineering culture.
Cutting AI Spend While Usage Grows
Armstrong also shared how Coinbase has managed to nearly halve its AI spending even as token usage has grown exponentially. The approach comes down to three techniques rather than friction-based controls or budget alerts.
The first is smarter model routing, which automatically directs each task to the cheapest model capable of completing it well. The second is aggressive caching, which eliminates redundant outputs when the same query appears repeatedly. The third is a deliberate shift toward cheaper open-weight models for routine tasks where frontier-model performance adds nothing of value.
Coinbase is one of the most AI-enabled companies in the world, based on all the feedback I hear.
Were in the age of the super builder. pic.twitter.com/NeFLukYIer
— Brian Armstrong (@brian_armstrong) July 3, 2026
“How to keep AI spend flat while token usage grows exponentially: not with friction and spend alerts. With better defaults, routing, and caching,” Armstrong said.
Building for Scale, Not for Constraint
Armstrong was clear that the goal of reducing cost is not to cap how much AI the company uses. It is to build the infrastructure layer that allows usage to compound sustainably without triggering budget pressure later. Efficiency gains create headroom rather than ceilings.
The framing positions Coinbase not as a company managing an AI budget but as one building the foundation to outpace any competitor through AI at scale. Armstrong said he hears this assessment from the companys board to its partners.
In early June, Armstrong had also argued that access to energy and compute matters more than model quality when it comes to AIs biggest bottleneck. The new spending data on routing and caching extends that argument into the infrastructure layer, giving a more complete picture of how Coinbase is thinking about sustainable AI adoption.
