Microsoft is building Fairwater datacenters with unprecedented scale—each housing hundreds of thousands of GPUs and over 2 GW of total capacity—enabling AI training capacity to grow tenfold every 18–24 months. To manage rising COGS from compute-heavy AI, Microsoft leverages hybrid business models across subscriptions, consumption, and devices. While Copilot has reached 26 million subscribers, the broader AI coding market is expanding rapidly, allowing room for competition. Microsoft sees long-term value not just in models but in the infrastructure layer supporting autonomous agents across M365 and Windows 365. It’s investing in custom AI chips, partnering with NVIDIA and OpenAI (hosted exclusively on Azure), and renting capacity via NeoClouds to meet immediate demand. With CAPEX surging, Microsoft treats compute as R&D, balancing capital intensity with software-driven efficiency. Globally, it navigates data sovereignty demands by aligning with national policies while maintaining trust in U.S. tech leadership.