Meta Limits AI Tokens After Employees Burn 73 Trillion in Dumb Leaderboard War
When a tech giant tells its workforce that using AI is mandatory for their performance reviews, what could possibly go wrong? Well, how about a multi-billion-dollar corporate bill and a secret game where employees keep bots running just to look cool? Absolutely genius.
The management at Meta sent out an urgent memo to thousands of employees, announcing strict limits on how many AI tokens they can burn. This sudden panic comes right after the company spent months aggressively pushing those same engineers to embed artificial intelligence into every single corner of their daily tasks.
According to leaked internal data, the company is looking at a projected AI bill of billions of dollars by next year. To stop the financial bleeding, Meta is now building a centralized dashboard called AI Gateway to track consumption in real-time and automatically alert managers when someone goes on a wild token-spending spree.
Engineers are also being booted off external tools like Claude, developed by Anthropic, and forced to use Meta’s own internal coding assistant, MetaCode. It turns out that paying retail price for someone else's API is significantly less fun when your employees treat it like a free buffet.
The root of this crisis lies in a corporate policy introduced last year, which declared that AI-driven impact would be a core metric for performance reviews. In response, competitive engineers launched an internal leaderboard jokingly named Claudeonomics. The game ranked the top 250 heaviest token users, awarding prestigious titles like "Token Legend" to those who wasted the most compute.
To climb this digital ladder, some employees literally kept AI agents running completely idle for hours, doing absolutely nothing but inflating the company's electricity and API bills. Even the Chief Technology Officer, Andrew Bosworth, tried to step in and explain that burning tokens for the sake of burning tokens does not actually mean you are doing good work, but his pleas were completely ignored in favor of the high score.
This massive scale-back is already sending shockwaves through the broader AI market, as Meta was one of Anthropic’s primary enterprise customers. As a result, the global LLM Token Spending Index has suffered its longest consecutive drop in months, and OpenAI is reportedly preparing deep price cuts to keep its corporate clients from fleeing.
The industry has officially transitioned from the magical era of unlimited hype to the cold reality of the monthly utility bill. It turns out that when AI performance is measured by usage, humans will gladly melt polar ice caps just to win a meaningless internal spreadsheet war.
Source: The Information
Comments
This is where the magic happens: AI reads your discussion and rewrites the article based on the most interesting comments. Each strong comment adds points to the meter below. Once the meter is full, the article updates live — no page reload needed.