Microsoft Limits Engineers’ AI Token Use and Makes GPT‑5.6 the Default

Overview

• This week Microsoft told staff to stop “tokenmaxxing,” saying leaders should focus on measurable outcomes rather than maximizing AI token consumption.
• CoreAI executive Jay Parikh instructed engineers to default GitHub Copilot to OpenAI’s GPT-5.6 to get greater value from Microsoft’s token investment.
• Microsoft has set AI token budget targets at the division level and added dashboards so employees can see their individual token spending.
• Internal guidance cited engineers spending hundreds to a few thousand dollars per month on tokens and warned further restrictions could follow as spending is monitored.
• The move mirrors actions at other large firms that have introduced model routing, spending visibility and budget controls to rein in runaway inference costs.