Meta has turned internal AI adoption into a company-wide competition.
They've launched a leaderboard, nicknamed 'Claudeonomics,' that tracks how much AI every employee uses. Over 85,000 employees are now competing for badges like 'Token Legend,' creating a culture where high AI usage is a status symbol. In a recent 30-day period, this gamification led to a staggering 60 trillion tokens being consumed.
To put that number in perspective, if Meta were paying the public list price for a top-tier AI model ($15 per million tokens), that usage would translate to a cost of $900 million for the month. That's an annualized run-rate of over $10 billion. But it's crucial to understand this is just an analogy; Meta's actual internal cost is far, far lower because they use their own infrastructure and models.
This didn't happen in a vacuum. There are a few key drivers. First, company leadership, including Mark Zuckerberg, has set a clear expectation that 2026 will be the year AI transforms how work gets done. With the company planning to spend up to $135 billion on capital expenditures for AI infrastructure, there's immense pressure to show that these expensive new tools are actually being used.
Second, the entire tech industry is embracing this 'use it or lose it' mentality. NVIDIA's CEO, Jensen Huang, recently normalized the idea that a top engineer should have a 'token budget' worth hundreds of thousands of dollars a year. This created a cultural tailwind, making Meta's internal competition seem not just normal, but necessary.
Third, Meta's own financial commitments created the perfect conditions. After announcing massive spending plans for AI, the company needed a way to encourage rapid adoption and learning. A leaderboard is a simple, if imperfect, way to kickstart that process and justify the expenditure.
In essence, Meta's leaderboard is a cultural tool designed to accelerate AI adoption. The initial focus is on raw usage—'how much AI you burn.' However, the conversation is already shifting. The real challenge ahead is moving from simply measuring consumption to measuring the actual value and productivity gains generated by each token.
[Glossary]
- Token: The basic building block of text for an AI. One English word is typically 1-2 tokens.
- Capex: Capital Expenditure. Money spent on physical assets like servers and data centers, which are essential for running AI.
- Goodhart's Law: A principle that says when a metric becomes a target, it stops being a good metric. For example, if employees are rewarded just for using more tokens, they might do so without creating real value.