Microsoft pairs strong AI growth with a more flexible OpenAI deal

Microsoft's AI revenue is now large enough to matter at company scale. In fiscal Q3 2026, Microsoft reported $81.4 billion in revenue, record capital expenditure above $37.5 billion, and said Azure is expected to grow about 40% in the current quarter. That combination means the company is no longer asking the market to fund AI on faith alone - it is showing both heavy spending and visible demand, even if the cost side is still rising fast.

The practical signal is that Microsoft remains in full build-out mode. Record spending focused on Azure and AI infrastructure tells readers that management still sees supply, not demand, as the binding constraint, and it also raises the bar for execution because investors will now expect continued monetization, not just capacity expansion.

Microsoft 365 Copilot is moving from promising add-on to established business line. Microsoft said Microsoft 365 Copilot now has 20 million paid enterprise seats, up from the 15 million paid users and more than $10 billion in annual recurring revenue pace disclosed days earlier, with weekly engagement per user matching Outlook and queries per user growing nearly 20% quarter-over-quarter.

That matters because it turns Copilot into more than a showcase for generative AI inside Office. If those usage patterns hold, Microsoft has evidence that Copilot is becoming habitual software rather than trial spending, which strengthens its pricing power and helps justify the wider AI infrastructure bill behind Azure and Microsoft 365.

Microsoft and OpenAI have loosened a defining alliance without breaking it. The amended partnership keeps Microsoft's primary cloud rights and a non-exclusive license to OpenAI's IP through 2032, while giving OpenAI broader multi-cloud flexibility and revising revenue-sharing as the companies continue AI infrastructure and silicon collaboration.

This is a meaningful change in control and dependence. For Microsoft, the immediate gain is continued privileged access without having to preserve the old exclusive structure at all costs; for the wider market, it confirms that frontier model access is becoming more multi-cloud and that Azure can no longer rely on OpenAI exclusivity as its simplest competitive story.

A revision is worth noting here. Earlier in the window, OpenAI's frontier models, Codex, and Managed Agents arrived on AWS in limited preview via Amazon Bedrock, and that looked like a competitive breach; the later partnership update reframes it as part of a more flexible structure rather than a one-off exception.

Microsoft is widening its sovereign and regulated AI position while rivals attack from multiple angles. Microsoft said Azure Local can now scale sovereign private cloud deployments to thousands of servers in a single sovereign environment, and it also announced an $18 billion multi-year investment to expand AI infrastructure in Australia. At the same time, rivals pushed hard on enterprise AI access: Amazon deepened its Anthropic collaboration with a $100 billion, 5 gigawatts compute deal, while Salesforce and Google Cloud expanded integrations for agentic AI workflows.

The common thread is that enterprise AI is fragmenting by workload and jurisdiction. Microsoft is trying to win the parts of the market where data residency, local operations, and government or regulated-industry requirements matter most, but competitors are making sure customers have alternatives across cloud, workflow, and model layers rather than defaulting to Azure and Copilot.

Microsoft's internal reshaping now looks like part of the AI strategy, not an isolated HR event. The company announced a one-time voluntary retirement program covering about 7% of US employees, and additional reporting said Microsoft is flattening management and cutting compensation layers as AI adoption changes internal workflows.

This deserves attention because it ties the economics of AI directly to the operating model of a large software company. Microsoft is effectively saying that AI expansion requires not only more datacenter capital but also a different cost structure and org design, which is a more concrete sign of AI-driven change than generic claims about productivity gains.