Meta Caps Internal Token Spend After 73.7 Trillion Tokens in 30 Days

Meta distributed a memo to 6,000 employees documenting exponential growth in third-party AI usage. The memo cited an internal Claudeonomics leaderboard that ranked teams by token volume. CTO Andrew Bosworth issued a follow-up directive stating token counts alone measure no output. The company will replace the leaderboard with centralized monitoring and steer usage to MetaCode.

Consumption data shows 73.7 trillion tokens processed in roughly 30 days. Parallel cases include Uber exhausting its 2026 AI coding budget in four months and imposing $1,500 monthly caps per tool. KPMG reports only 26 percent of firms maintain visibility into AI costs. Goldman Sachs forecasts enterprise token volume reaching 120 quadrillion per month by 2030.

The controls address balance-sheet exposure while Meta commits $135 billion to AI infrastructure through 2026. Shifting workloads to MetaCode reduces external API payments and generates internal training data. Industry patterns indicate adoption phases are giving way to governance once per-token charges appear on operating statements.

Budgets take effect January 2027 with automated alerts for spending anomalies. Meta will track attribution between token volume and code commits or productivity metrics.