Meta will limit employee AI usage: It will build its own gateway for monitoring, with internal costs potentially reaching billions by 2026.

According to Beating, Meta Platforms plans to control soaring AI costs within the company by limiting employee token usage. A leaked internal memo indicates that Meta is building a central gateway called AI Gateway to monitor employee AI usage and spending in real time, set budgets, and cap token consumption. Meta anticipates internal AI usage alone will generate billions of dollars in costs by 2026. This capping measure contrasts sharply with Meta's previous aggressive promotion of AI. In November 2025, Meta notified employees that demonstrating "AI-driven impact" would be a core performance indicator for 2026, linking performance bonuses to AI usage rates. This excessive promotion sparked a "tokenmaxxing" frenzy among employees, even leading to the emergence of an internal leaderboard called "Claudeonomics" to publicly display usage rankings. Before the leaderboard was shut down, the total amount of tokens consumed by employees within 30 days surged to 73.7 trillion. Meta's Chief Technology Officer, Andrew Bosworth, subsequently issued a warning, emphasizing that simply increasing token consumption does not equate to output, and employees should use AI tools in scenarios that genuinely improve efficiency. To further reduce spending, Meta began shifting its internal AI development focus towards self-developed tools. Leaked memos revealed that Meta is pushing employees to gradually phase out third-party programming tools such as Anthropic's Claude, and instead use its self-developed programming assistant, MetaCode (formerly known as Devmate). Meta's newly established Applied AI Engineering department has been tasked with fully improving MetaCode, generating high-intensity reinforcement learning training data by creating programming challenges. While Meta still allows employees access to external models, it will implement stricter budget and quota approval mechanisms in its self-built gateways in the future. Meta is not the only company facing financial pressure due to overloaded large model usage. In early 2026, companies like Uber and ServiceNow exhausted their entire year's Anthropic quotas within just a few months. ServiceNow has implemented daily usage monitoring for its employees, and some venture capital firms have also begun to set daily spending limits on internal AI Token usage to prevent unrestrained expansion of computing power costs.