Microsoft is tightening its internal rules on AI token use among engineers, asking employees to focus on useful results rather than consuming large amounts of computing resources, as the company moves to control rising AI costs.
The change was outlined in an internal memo from Jay Parikh, Microsoft’s executive vice president. He told employees that using as many AI tokens as possible should not be treated as a measure of success.
Parikh said engineers should pay attention to how they use tokens on platforms such as GitHub. He stressed that the company wants employees to focus on results that benefit customers and the business.
The term “tokenmaxxing” refers to using very large amounts of AI tokens, which are units used by AI systems to process and generate information, more token use can require more computing power and increase costs.
Microsoft is now updating its internal guidance and plans to manage AI token spending more carefully. The company wants employees to treat AI resources like other important business resources that need to be monitored.
As part of the changes, OpenAI’s GPT 5.6 model will become the default choice for Microsoft engineers for routine coding work, the cheaper model will replace more expensive options for everyday tasks.
Microsoft has not publicly announced specific spending limits for individual engineers. However, the internal guidance reportedly shows that some employees currently use AI services costing hundreds or even several thousand dollars each month.
Microsoft’s move reflects a wider challenge for technology companies as businesses increase their use of powerful AI tools while facing higher computing expenses.
