According to Android Headlines, Microsoft has begun tightening the management of Token usage. On one hand, it has set up monitoring for Token usage, and on the other hand, it is gradually switching internal default workloads to GPT-5.6 Sol to reduce AI costs. The trigger was a leaked internal table: in recent months, some employees have been calling AI extensively to boost data rankings, with one employee generating approximately $28,000 (about 189,000 RMB) in Token costs within 28 days, causing concern among senior management.

Significant departmental differences, CoreAI is the most expensive

According to the data, Microsoft has over 223,000 global employees, with about 350 US employees voluntarily submitting monthly AI usage data. The median monthly cost is around $300 (about 2,024 RMB). There are significant differences between departments: CoreAI has the highest median cost, around $975, with an individual reaching as high as $16,000; Microsoft AI is about $490, Experiences and Devices is about $250, and Azure is about $241. The employee who set the record of $28,000 comes from the Customer and Partner Solutions department.

To control costs, Microsoft is strengthening monitoring of the value of Token usage, identifying employees with unusually high AI spending, and setting stricter budget limits. At the same time, the company is gradually shifting internal default workloads to OpenAI's GPT-5.6 Sol, and no longer encourages employees to use third-party AI tools such as Anthropic Claude for programming tasks—besides reducing costs, this also consolidates internal usage into its own ecosystem.