'Tokenmaxxing is not what we are optimizing for': Microsoft tells engineer to calm down on AI usage

Microsoft is introducing new guidelines to curb what it calls "tokenmaxxing" among its engineers, according to an email seen by 404 Media. The company aims to optimize the use of AI tokens, which are used by its GitHub Copilot platform, to focus on the return on investment (ROI) rather than simply maximizing the number of tokens consumed.

In the email, Jay Parikh, an executive vice president at Microsoft, emphasized that "tokenmaxxing is not what we are optimizing for." Instead, the company wants to ensure that engineers are maximizing outcomes that benefit customers and the business. To achieve this, Microsoft will update its internal guidance and manage token spend with the same discipline as any other critical resource.

As of July 2026, employees will have an "AI token budget target," and they can track their individual AI spending. Microsoft is also making the cheaper OpenAI GPT-5.6 model the default model for internal use. Parikh noted that while there is no specific target spend value being shared, the data shows that engineers' token usage ranges from hundreds to thousands of dollars per month.

Microsoft's decision to implement these restrictions may come as a surprise given its recent record financial results. However, it is part of the company's ongoing efforts to focus internal AI usage. In May 2026, Microsoft reportedly canceled most of its Claude Code license, prompting engineers to use GitHub Copilot CLI. Tokenmaxxing has also been an issue at other tech giants, such as Uber and Amazon, which have faced budget exhaustion due to excessive AI token consumption.