Microsoft restricts AI spending amid rising costs

Microsoft has implemented stricter controls over AI token usage, following concerns over excessive spending by some employees. A leaked internal report revealed that one employee incurred $28,000 in token costs within 28 days, prompting management to take action. The company is now monitoring token usage more closely, identifying high spenders, and imposing tighter budget limits. Additionally, Microsoft is shifting internal workloads to OpenAI’s GPT-5.6 Sol to cut costs and reduce reliance on third-party tools like Anthropic’s Claude. Departmental spending varies significantly, with the CoreAI department having the highest median cost at $975, while Azure’s median cost is around $241. The employee with the highest individual cost, $16,000, belongs to the Customer and Partner Solutions department. Microsoft’s move aims to consolidate AI usage within its ecosystem while managing rising expenses. The company has over 223,000 global employees, with about 350 US employees voluntarily sharing monthly AI usage data. The median monthly cost across all departments is approximately $300. By redirecting workloads to GPT-5.6 Sol, Microsoft hopes to reduce costs and streamline internal AI operations. The changes reflect a broader strategy to control expenses and ensure responsible AI usage across the organization.