Microsoft Tries to Reel in Engineers' AI Spending Habits

Microsoft has informed its engineers that they will continue to use AI tools at work but must be mindful of costs. In an internal email, executive vice president Jay Parikh emphasized that the company is not optimizing for token usage but rather for impact per token.

The email confirms that Microsoft has made OpenAI's GPT-5.6 the default model for internal use, which is a cheaper option compared to other models used by engineers.

While there is no cap on AI usage, employees can now see their individual spending on an internal dashboard, and further restrictions may be imposed if token spend exceeds targets set for each division.

The company's move comes after Microsoft CEO Satya Nadella acknowledged that tokenmaxxing, treating AI consumption as a proxy for productivity, was a problem within the company. He had warned about it weeks ago, saying 'it's addictive' and advising engineers to use frontier models only when necessary.