Coinbase CEO: Optimizing AI Costs Through Better Default Models and Routing Mechanisms, Achieving Token Usage Growth Alongside Declining Expenditure

Odaily Odaily reports that Coinbase CEO Brian Armstrong posted on X, stating that internally at Coinbase, the team is exploring how to keep AI spending stable while token usage grows exponentially, not through usage limits or constant cost reminders, but through better system design.

He noted that Coinbase is optimizing AI usage costs primarily from five aspects. First is "Better Default Models," allowing engineers to freely choose models, but the system defaults towards lower-cost open-source weight models (such as GLM 5.2, Kimi 2.7), while still encouraging the selection of suitable models for different tasks. Second is "Smart Routing," using an automated system to allocate the optimal model based on task type, cache hit rate, and model pricing, rather than manual selection.

Third is "Cache Optimization," significantly reducing repetitive computation costs by improving the cache hit rate. Fourth is "Streamlined Context," reducing ineffective token consumption during task switching and limiting unnecessary tool calls. Fifth is "Visual Transparency," giving engineers clear visibility into their own usage, without restricting usage through limitations.

Brian Armstrong stated that the goal is not to suppress AI usage, but to build infrastructure capable of supporting exponential growth. The results show that while token usage continues to grow, overall AI expenditure has nearly halved.