Microsoft’s AI Moment: Prove the Economics

Microsoft spent 2023 and 2024 convincing investors that artificial intelligence would change its business. In April 2026, the argument is narrower: can the company turn Copilot into paid seats fast enough to justify the servers, chips, and data centers it is building. (microsoft.com) (bnnbloomberg.ca) Microsoft’s stock closed at $372.88 on April 10, 2026, well below its 52-week high of $555.45 shown on Microsoft’s own investor page. That gap is why the market is no longer grading Satya Nadella on demos and launch events. (microsoft.com) Copilot is Microsoft’s paid artificial intelligence assistant for work, and the flagship version is sold as an add-on to Microsoft 365 for $18 per user per month when billed yearly. That price turns adoption into a simple math test: every employee seat that upgrades has to cover not just software costs, but the computing bill behind the model. (microsoft.com) Microsoft has also widened the funnel by giving enterprise customers a lower-tier secure AI chat at no additional cost with eligible Microsoft 365 subscriptions. That helps usage, but it also raises the pressure on the premium Copilot product to prove why companies should pay extra. (microsoft.com) The company’s own numbers show the machine behind the story is already huge. Microsoft reported fiscal second-quarter revenue of $81.3 billion on January 28, 2026, up 17%, and Nadella said Microsoft had already built an artificial intelligence business “larger than some of our biggest franchises.” (microsoft.com) But the same quarter also came with a supply warning. On the fiscal 2026 second-quarter call, Microsoft said it had to balance incoming capacity for Azure cloud demand, first-party use in Microsoft 365 Copilot and GitHub Copilot, research needs, and replacement of aging servers and networking gear. (microsoft.com) That is the bottleneck investors are staring at now. If Microsoft buys too little infrastructure, Copilot and Azure customers wait; if it buys too much infrastructure, the company is left defending tens of billions of dollars of spending before the revenue fully arrives. (microsoft.com 1) (microsoft.com 2) The spending is not theoretical. In July 2025, Microsoft told investors it expected first-quarter capital expenditures to be over $30 billion, driven by strong demand signals and continued scaling of artificial intelligence infrastructure. (microsoft.com) Microsoft is trying to make that infrastructure earn money from both sides. Azure sells the computing power to outside customers building their own artificial intelligence tools, while Copilot uses the same broad investment to sell finished assistants directly inside Word, Excel, Outlook, Teams, and other Microsoft software. (azure.microsoft.com) (microsoft.com) There are real customer wins, but they are still anecdotes until they show up cleanly in recurring revenue. Microsoft’s 2025 annual report highlighted Barclays rolling out Copilot to 100,000 employees, which is the kind of deployment Wall Street wants to see repeated across many companies, not just a few showcase names. (microsoft.com) So the next phase of the Microsoft story is less about whether artificial intelligence is useful and more about whether the bill and the subscription line move in the right order. If paid Copilot seats keep rising and Azure demand stays tight, the infrastructure looks disciplined; if adoption lags, the same buildout starts to look like a very expensive promise. (microsoft.com 1) (microsoft.com 2)