en.Wedoany.com Reported - Meta plans to limit the number of AI tokens employees can use, according to an internal memo sent to approximately 6,000 staff. This shift comes shortly after the company encouraged the same group to integrate AI into their work, driven by cost pressures: as reported by The Information, employees consumed 73.7 trillion tokens in 30 days, and the company is expected to spend billions of dollars on AI in 2026.
To control spending, Meta is building a centralized dashboard called AI Gateway, which tracks consumption and spending by team in real time, sets budgets, and sends automatic alerts for abnormal spikes. Strict token quotas will take effect later, and the company plans to structurally allocate budgets in 2027. Meanwhile, engineers are being steered away from external tools like Anthropic's Claude toward the internal coding assistant MetaCode (formerly Devmate), to reduce spending on external APIs and refine its own products.
Meta's AI consumption overheating is largely self-inflicted. In November last year, the company announced that AI-driven impact would become a basic expectation for 2026 performance reviews, making AI usage a potential prerequisite. Employees launched an internal leaderboard called Claudeonomics (named after Claude), ranking the top 250 most active token users and awarding titles such as "Token Legend." Competition for leaderboard positions led to the 73.7 trillion token monthly consumption, with some leaving AI agents idle for hours just to climb the rankings. In April this year, Meta's Chief Technology Officer Andrew Bosworth tried to cool the enthusiasm, writing that no one should use AI for the sake of using AI, and that token consumption alone measures nothing, but words alone were insufficient—a dashboard was needed.
The main issue in this story lies in the economics of the entire industry. Meta is known as a major customer of Anthropic, and if token budgets are cut, it will hit AI labs' own revenue. Signs have already emerged: the LLM Token Spending Index, which reflects the market's willingness to pay for AI, has fallen for seven consecutive trading days, the longest losing streak since January. According to The Wall Street Journal, OpenAI is considering significant price cuts to retain enterprise customers. In the past, when companies competed over who could provide employees with more AI access, the question was "Can we use it?" Now, the question has become "How efficient is it?"
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
