Coinbase Cut Its AI Spend in Half Without Throttling Engineers - Here's the Playbook

Coinbase's CEO outlined 5 strategies to keep AI spend low at his company without limiting token usage

Coinbase cut AI spending to half its peak while maintaining record token usage through cheaper default models (Chinese-developed GLM 5.2, Kimi 2.7), intelligent routing by task difficulty, and caching. Signals industry shift from uncontrolled tokenmaxxing to smart model orchestration—CTOs must invest in LLM gateways and per-task routing to scale AI spend sustainably.