Meta Caps Internal AI Token Spending After Costs Approach Billions in 2026

Meta Platforms is taking decisive action to rein in its burgeoning internal artificial intelligence costs, which are on track to hit billions of dollars in 2026. An internal memo, circulated to approximately 6,000 employees and initially reported by The Information, highlighted an "exponential increase" in AI usage across the company. This surge in consumption has prompted Meta to implement centralized spending controls and a more structured approach to AI resource allocation. The memo detailed that Meta employees collectively consumed a staggering 73.7 trillion tokens over a 30-day period. This high usage was partly attributed to an internal leaderboard, humorously dubbed "Claudeonomics," which tracked employee token consumption. While intended to encourage AI adoption, this gamified system inadvertently incentivized "tokenmaxxing"—the practice of inflating AI usage metrics, sometimes without a direct correlation to genuine productivity gains. In response to this trend, Meta's Chief Technology Officer, Andrew Bosworth, issued a separate memo cautioning employees against using AI tools merely for the sake of it. He underscored that "All motion is not progress and token usage alone is not a measure of impact of any kind," signaling a shift in focus from raw consumption to measurable impact and efficiency. To address the escalating costs, Meta plans to roll out a new "AI Gateway" dashboard in the coming weeks. This system will provide teams with better visibility into their AI consumption and will include automated alerts for unusual spending spikes. Furthermore, the company intends to implement formal token budgets and allocations starting in early 2027. This strategic move aims to bring more financial discipline to internal AI operations, ensuring that the substantial investments in AI infrastructure translate into tangible business value. A key part of Meta's cost-control strategy involves steering employees away from third-party AI tools, such as Anthropic's Claude, and towards its own proprietary coding assistant, MetaCode. This shift serves a dual purpose: reducing external API costs and promoting the "dogfooding" of Meta's own AI products, thereby enhancing their development and internal adoption. Meta's experience mirrors a broader challenge faced by many large enterprises that have aggressively adopted AI. Companies like Uber have also encountered similar issues, with Uber reportedly exhausting its entire 2026 AI coding budget in just four months. These incidents highlight a growing industry-wide recognition that while AI adoption is crucial, effective cost governance and a clear link between AI spending and measurable output are becoming increasingly critical. The transition from open-ended AI tool access to metered usage marks a significant cultural and operational shift within Meta, positioning it to manage its massive AI investments more strategically.