Meta Caps Internal AI Token Spending After Costs Approach Billions in 2026

Meta’s workers apparently turned company AI tools into a full-on office sport, and the internet is absolutely feasting on the fallout. According to reporting from The Information, employees used a jaw-dropping 73.7 trillion tokens in about a month — basically tiny chunks of AI processing that add up to very real money — and internal costs are now racing toward billions of dollars in 2026. The biggest gasp from commenters wasn’t just the price tag. It was the reveal that Meta had a token leaderboard, nicknamed “Claudeonomics,” which ranked teams by how much AI they used. As one stunned commenter put it, “Who could possibly have predicted” that people would start gaming that system.

That sarcasm set the tone. Community reaction swung between mockery and dread: mockery at the idea that workers were “tokenmaxxing” like teenagers grinding points in a video game, and dread that companies will now take the absolute wrong lesson and clamp down on useful AI tools for everyone. Some commenters wanted the real breakdown — was this engineers writing code, or people feeding giant PDF files, slide decks, and image requests into chatbots all day? Others cut straight to the management lesson: measure impact, not activity. In other words, don’t reward people for burning through digital words just to look busy. The whole drama has become a perfect tech-era morality play: if you build a scoreboard, don’t act shocked when everyone starts playing to win.