Inside Meta, burning AI compute has become a status symbol. And the numbers are staggering.

An internal leaderboard called "Claudeonomics" has emerged across the company, tracking AI token consumption for all 85,000 employees and ranking the top 250 power users. The name comes from Anthropic's Claude, which appears to be the model Meta's workforce is devouring. Top performers earn titles like "Session Immortal" and "Token Legend." This is not a joke. This is corporate culture in 2026.

The Information reported on April 6 that the leaderboard recorded over 60 trillion tokens consumed in the past 30 days. The single heaviest user burned through 281 billion tokens. At Anthropic's public pricing of roughly $15 per million tokens for Claude Opus 4.6, that 60 trillion figure translates to approximately $900 million in a single month.

Let that land for a second. One company. One month. $900 million in AI token consumption.

Now, Meta almost certainly negotiates volume pricing that makes the real cost a fraction of that headline number. But even at 10 cents on the dollar, you are still looking at $90 million a month just on AI inference. And the consumption is accelerating, not plateauing.

The Real Question Nobody Is Asking

Does burning more tokens actually make you more productive? Or has Meta accidentally created a gamification system that rewards AI consumption regardless of output quality? When your internal leaderboard celebrates volume over value, you are incentivizing employees to use AI for everything, whether it helps or not.

This also reveals something fascinating about the current state of the AI industry: the biggest AI companies are not just building models. They are their own best customers. Meta is simultaneously developing Llama while burning enormous volumes of Anthropic's Claude. Google employees use ChatGPT. OpenAI engineers reportedly prefer Claude for certain tasks. The model loyalty that these companies sell to the public does not exist inside their own walls.

The Claudeonomics leaderboard is a perfect snapshot of where enterprise AI adoption actually stands: chaotic, competitive, expensive, and completely unmonitored in terms of ROI. If Meta, one of the most sophisticated technology companies on Earth, cannot measure whether its AI spending is productive, what hope does everyone else have?

First reported by The Information.