Microsoft Tells Engineers 'Tokenmaxxing Is Not What We Are Optimizing For
5 hours ago
- Microsoft is imposing new limits on how much its engineers can spend on AI tools at work, stating that maximizing AI use internally is not the company's goal.
- The trend of reducing excessive AI usage, termed 'tokenmaxxing', is emerging across major companies as AI costs rise without proportional productivity gains.
- Microsoft is making the cheaper OpenAI GPT-5.6 the default model for internal use and plans to set 'AI token budget targets' for divisions by July 2026.
- The policy shift is not due to financial constraints—Microsoft's revenue and profits are strong—but aligns with cost-control measures at companies like Amazon, Adobe, and Citi.
- An anonymous Microsoft employee expressed skepticism, noting that the company's own difficulty affording its AI products raises questions about their value to customers.