Jay Parikh: Microsoft limits engineers' AI spending to stop tokenmaxxing
Microsoft is tightening up how much its engineers can spend on AI tools like GitHub Copilot.
The goal? To stop "tokenmaxxing" (basically, using tons of AI without real results).
Jay Parikh, the company's executive vice president, reminded everyone to focus on what actually helps customers and the business.
Microsoft assigns AI token budgets
Starting July 2026, Microsoft divisions will get an "AI token budget target" so they can track and manage their AI usage.
Microsoft says this isn't because of money problems (they're still seeing strong profits), but it's about being smart with resources as they push to become more "AI-first."
They've also switched to OpenAI's GPT-5.6 model in-house to cut costs, joining other big names like Amazon and Atlassian in reining in unnecessary AI use.