According to The Information, Meta is currently experiencing a frenzied token consumption competition internally. An employee created a leaderboard called "Claudeonomics" on the company's internal network, tracking the AI token (Tokens) consumption of more than 85,000 employees in real time. Within just 30 days, Meta employees consumed a total of 60 trillion tokens, with the top user averaging an astonishing 281 billion tokens.
To encourage employees to deeply integrate AI into their daily work, the leaderboard also features gamified titles such as "Token Legend," "Model Connoisseur," and "Cache Master." However, this data-driven evaluation has also led to negative consequences: some employees are even letting AI agents run in the background just to "rank up," causing expensive computing resources to be wasted unnecessarily.
Despite the controversy, the concept of "token maximization" has become a core metric for measuring engineers' productivity in Silicon Valley. NVIDIA CEO Jensen Huang once stated that if an engineer earning $500,000 per year consumes tokens worth less than $250,000 annually, he would be deeply concerned about their work performance. Meta Chief Technology Officer Andrew Bosworth also supports this logic, revealing that one of the company's top engineers increased his efficiency by ten times by consuming tokens equivalent to his annual salary.
However, the scientific validity of this evaluation system still faces significant challenges. There is currently no concrete data proving a necessary link between token consumption and business outcomes. Critics argue that using token consumption to measure productivity is akin to judging a truck driver's performance based on fuel consumption—it only proves the engine is running, but not whether the cargo has been delivered.
For companies like Google and Meta, which are eager to demonstrate the rationality of their massive AI investments to investors, token consumption often serves as an alternative indicator of user growth. However, as the artificial inflation of reasoning tokens continues, this measurement method, which excludes actual revenue growth, may struggle to support Silicon Valley's long-term narrative logic.
