Microsoft has spent three years urging AI adoption but is now telling its own engineers to use it more carefully. In an internal email to his CoreAI organisation, executive vice president Jay Parikh wrote that "tokenmaxxing is not what we are optimizing for," backed by new policy. OpenAI's cheaper GPT-5.6 is now the default model for internal use, every Microsoft division has an AI token budget target, and employees can view a dashboard showing their AI usage costs. The memo, first reported by 404 Media, says internal Copilot guidelines note many engineers spend from a few hundred dollars to a few thousand dollars a month on tokens; no hard cap has been shared, but further restrictions may follow. Parikh framed the change as discipline, not distress, stating, "We are not optimizing for fewer tokens. We are optimizing for more impact per token." Microsoft's latest earnings beat Wall Street on revenue, operating income and net income. Per-token prices have dropped roughly 98 percent since late 2022, yet enterprise AI bills have tripled. In June, at a New York Times' Hard Fork podcast taping, Satya Nadella called himself a tokenmaxxer, saying, "It's addictive," and advised matching the model to the task. Microsoft cancelled most Claude Code licences in its Experiences and Devices group in May. Amazon, Adobe, Atlassian and Citi have all introduced AI throttling or spend visibility, with Meta shutting down "Claude