Meta Faces Crisis Over Surging Internal AI Costs

Meta's newly formed Applied AI unit is facing a serious internal cost crisis and employee discontent. According to an internal memo, costs from internal AI usage are expected to reach billions of dollars by 2026, prompting the company to introduce an AI Gateway dashboard and establish strict Token Budget plans.

This crisis stems from a phenomenon called 'Tokenmaxxing,' which arose after AI usage was incorporated into performance metrics. As employees artificially inflated their consumption via an internal leaderboard called 'Claudeonomics,' 73.7 trillion tokens were consumed in just about 30 days.

In addition, engineers assigned to the 6,500-person Applied AI unit are strongly pushing back against simple data-generation work, referring to it as the 'Gulag,' and organizational culture tensions are intensifying, including instances of verbal abuse during internal meetings.