Microsoft Azure tops $100B annual revenue as Copilot passes 30M seats

Microsoft closed a record fiscal year: annual revenue of $331B (+18%), Microsoft Cloud at $214B (+27%), and Azure crossing $100B for the first time, with Q4 year-over-year growth accelerating to 41% (some reports cite 43%, up from 40% in Q3). Microsoft 365 Copilot passed 30 million paid seats, and net income benefited from Microsoft's investments in both OpenAI and Anthropic.

The strategic subtext, which Nadella emphasized on the earnings call and in a follow-up LinkedIn post, is a deliberate reduction of single-vendor dependence. 'We are building a new model system, where the harness, context, memory, and action space are separate from any one model family,' he wrote — a design that moves the frontier on the cost-to-outcome curve and adds business-continuity resilience. Microsoft is designing its own models alongside its own chips, claiming efficiency gains of up to 40%, and is testing Moonshot AI's Kimi K3 on Azure, a move analysts estimate could cut inference costs by roughly $600 million.

Nadella also demoed the strategy personally, describing an 'ROIC Intelligence App' he built with a single Copilot code prompt and a /drill-me skill, keeping all artifacts inside Microsoft's enterprise environment — a pointed illustration of the Copilot superapp direction.

The competitive read: enterprises framed this as evidence of 'a real cooling of OpenAI dependence,' a recurring theme in enterprise AI discussions. The caveat is execution risk — building competitive in-house models is expensive, and testing a Chinese open-weight model on Azure carries its own procurement and geopolitical scrutiny. Still, $100B in Azure revenue with 41% growth is the clearest signal yet that enterprise AI spend is real and compounding.