At Meta, high consumption of AI computing resources is emerging as a new status symbol among employees.
On April 6, according to a report by The Information, an internal leaderboard called "Claudeonomics" has surfaced within Meta Platforms. This employee-built system, hosted on the company intranet, tracks AI token usage across more than 85,000 staff members and ranks the top 250 "super users." High-ranking employees can earn titles such as "Session Immortal" or "Token Legend."
The name "Claudeonomics" references the flagship product Claude from AI startup Anthropic. A copy of the leaderboard obtained by media shows that over the past 30 days, total recorded token usage exceeded 60 trillion. The top individual user consumed an average of 281 billion tokens—a volume that, depending on the model used, could represent a cost running into millions of dollars.
Based on Anthropic's latest public pricing, the average cost for input and output tokens using its Claude Opus 4.6 model is approximately $15 per million tokens. By this estimate, 60 trillion tokens would correspond to roughly $900 million in expenses. However, it remains unclear which specific models Meta uses and at what negotiated prices.
**Token Spending as a New Productivity Metric** This trend reflects the growing "tokenmaxxing" culture in Silicon Valley, where token consumption is being used as a benchmark for productivity and as a competitive measure of how "AI-native" an employee is.
Tech executives are endorsing this approach.
NVIDIA CEO Jensen Huang stated last month that he would be "deeply concerned" if a $500,000-per-year engineer spent less than $250,000 annually on AI tokens.
In February, Meta's CTO Andrew Bosworth remarked at a tech conference that, according to a Forbes report, a top engineer spending an amount equivalent to their salary on AI tokens could see productivity increases of up to tenfold. Bosworth commented, "It's a no-brainer. Keep doing it. There's no upper limit."
Andrej Karpathy, a former top AI scientist at Tesla and OpenAI, now leading an AI education startup, echoed this sentiment in a recent podcast: "It's all about the tokens. What's your token throughput? How much token throughput can you mobilize?"
**How the Leaderboard Functions** Employees can track their personal usage, compare with colleagues, and earn gamified rewards—ranging from bronze, silver, gold, platinum, and emerald badges to achievement titles like "Model Connoisseur" and "Cache Wizard."
According to two current employees, some staff members keep AI agents running for hours on research tasks to maximize token consumption and improve their rankings.
Meta also maintains a separate token usage dashboard for software engineers, while employees in other roles can view their own usage data. Notably, a source familiar with the matter indicated that neither Mark Zuckerberg nor Andrew Bosworth ranked among the top 250 super users.
In terms of tools, Meta employees have access to models from Anthropic, OpenAI, and Alphabet, in addition to internally developed tools such as Meta's version of OpenClaw (internally called MyClaw) and Manus, which Meta recently acquired.
A Meta spokesperson stated, "It's well known that this is a company priority. We are focused on leveraging AI to help employees with their daily work."
**Skepticism: Does Consumption Equal Productivity?** The competition is not without its critics.
Bloomberg commentator Joe Weisenthal questioned on platform X: "What exactly is the point of using total token consumption as a measure of productivity?"
He later added sarcastically: "This really has a 'backyard steel furnaces vibe,'" suggesting that the obsession with numerical metrics, while ignoring actual quality, resembles a reckless waste of resources.
This criticism highlights a fundamental issue: token consumption is an input metric, not an output metric. Just as counting printed pages doesn't measure work efficiency, burning more tokens doesn't necessarily yield better results. The practice of leaving AI agents idle for hours to boost rankings illustrates how the metric can be manipulated.
In response, well-known tech analyst Noah Brier offered a different perspective: "I don't think it makes sense, but when you're trying to steer a massive organization like Meta, sometimes you have to purposely overcorrect."
Weisenthal followed up by asking: "Even so, what exactly are they trying to correct—employee work habits, or the company's revenue model?"
From a market perspective, however, this trend sends a clear signal: enterprise-level AI consumption is expanding at a pace far exceeding expectations. For Meta alone, estimated monthly AI computing expenses could approach the $9 billion range, indicating sustained demand growth for cloud computing and AI infrastructure providers.
Comments