New reporting on Microsoft’s internal AI tooling suggests companies are recalibrating AI adoption as usage grows faster than budgets and economics. The report states that Microsoft began canceling most direct Claude Code licenses and steering engineers toward GitHub Copilot CLI, despite earlier encouragement for widespread experimentation. Separately, Uber’s CTO told The Information that the company burned through its entire 2026 AI coding tools budget in four months after incentives encouraged adoption through internal leaderboards tracking tool usage. Together, the accounts illustrate an “AI paradox” in which cheaper model tokens can still produce larger total compute bills due to volume and institutional scaling. The implications for higher education are indirect but notable: as universities and labs adopt AI-assisted tooling, they may need to plan for governance over usage—particularly when AI output generation is tied to time, compute, and procurement models rather than purely software access. For institutional leaders, the developments reinforce that AI adoption is not only a pedagogy and ethics question, but also an operating-cost and compliance design issue.