Multiple outlets report that employees at large tech firms have been inflating internal AI usage metrics, a practice dubbed “tokenmaxxing.” Financial Times reporting, cited across outlets, says Amazon set targets to get more than 80% of its developers using AI weekly and exposed an internal leaderboard tracking token consumption. Fortune and The Information reported Meta ran a ranking called “Claudeonomics” that registered more than 60 trillion tokens in a 30-day window. Tom’s Hardware and allied coverage note that agentic AI workflows can consume dramatically more tokens than single LLM queries, with Tom’s Hardware citing up to 1000x higher token use depending on the number of steps. The reporting shows companies including Amazon, Meta, and Microsoft responding by restricting dashboards or nudging employees toward internal tools as the cost and noise from inflated usage grew.
