Microsoft Weighs AI Token Budgets to Rein In Copilot Costs

As generative AI spreads through corporate workflows, the tokens consumed by model inference have become a significant recurring expense. Microsoft sells Microsoft 365 Copilot to customers while deploying AI tools broadly across its own workforce, making cost discipline an important test of whether heavy AI investment can produce measurable business value. The company is increasingly focused on the impact generated by each token, rather than treating higher usage alone as evidence of successful adoption.

Microsoft has recently updated internal guidance and is considering “AI token budget targets” for individual departments, according to the report. Employees have been told that maximizing token consumption is not the objective. The company has also made lower-cost OpenAI GPT models the default choice for internal use. Microsoft has not disclosed department-level limits, a formal rollout date or a monetary budget, but the measures are intended to curb rising operating and tool costs while preserving benefits for customers and the business.