Microsoft Stock’s 25% correction looks “cheap” — but the AI demand it’s tied to raises uncomfortable questions
microsoft stock is down 25% from its all-time high and described as the cheapest it has been in more than three years, yet the sell-off is unfolding alongside two competing realities: Azure is described as Microsoft’s fastest-growing business, while Copilot adoption inside paid enterprise software remains modest so far.
What is the market really pricing into Microsoft Stock after a 25% correction?
Microsoft is positioned as one of a small group of hyperscalers offering vast cloud computing capacity through hundreds of centralized data centers around the world. That capacity is rented to business customers through Azure, and many of those customers use it to develop and deploy artificial intelligence software.
The tension for investors is straightforward. On one side, Azure is described as consistently the fastest-growing piece of Microsoft’s overall business. On the other, investor focus has sharpened around Copilot, the company’s AI virtual assistant embedded into flagship products including Windows, Bing, and 365 (Word, Excel, and Outlook). Copilot is framed as a massive financial opportunity, but adoption has been modest so far, and that modest adoption is explicitly linked to the recent sell-off in microsoft stock.
The public narrative implied by the price move is that the AI story is being tested in real time. The open question is whether the market is reacting primarily to near-term adoption data around Copilot, or to broader concerns about how quickly AI-driven cloud demand can be converted into durable revenue.
Why modest Copilot uptake is colliding with a massive addressable base
Copilot can function as a regular chatbot, but it is also positioned as a productivity tool when packaged with enterprise software. While it can be used for free through Windows or Bing, Microsoft charges a fee when businesses want to embed it in the 365 application suite.
The scale of the addressable market is spelled out in licensing terms: companies pay for over 400 million 365 licenses for employees globally. Against that backdrop, Microsoft’s recent fiscal 2026 second quarter (ended Dec. 31) showed businesses had bought 15 million Copilot licenses for 365, implying a penetration rate of 3. 7%.
That figure is low in context, and it sits at the center of the current debate about microsoft stock. The same dataset also contains a counterweight: Copilot for 365 licenses grew 160% year over year, the number of businesses with more than 35, 000 Copilot for 365 licenses tripled during the quarter, and daily active users increased tenfold. Those points suggest a pattern in which adoption, once it begins, can broaden inside organizations over time and usage can ramp sharply.
In practical terms, the contradiction is that Microsoft can point to strong growth rates and expanding usage intensity, while investors can point to the still-small share of 365 seats paying for the product today. Both can be true at once—and that coexistence is now being reflected in valuation pressure.
How Azure’s AI infrastructure push creates both confidence and risk
AI development is described as happening largely inside data centers filled with thousands of specialized chips supplied by companies including Nvidia. Because this infrastructure can cost billions of dollars to build, many developers rent computing capacity from cloud providers such as Microsoft Azure instead.
Demand signals in that ecosystem are presented as extraordinary. As of Dec. 31, Microsoft had a $625 billion order backlog from customers waiting for more data centers to come online, up 110% year over year. Microsoft also invested $118 billion to build more infrastructure over the last four quarters, with the expectation that spending will increase further.
For microsoft stock, that combination—huge backlog paired with heavy investment—can be read as a bullish indicator that demand is real and capacity-constrained. But it also increases scrutiny of what, exactly, sits inside the backlog and how concentrated that demand may be.
What isn’t being explained clearly: the concentration inside the $625 billion backlog
The backlog includes a concentration risk that is explicitly flagged: 45% of the $625 billion backlog, or $281 billion, is attributable to OpenAI alone. The context also states OpenAI has around $20 billion in annualized revenue at the moment and recently secured $110 billion from investors in a capital raise, while also noting that funding still would not be enough to fulfill its obligations to Microsoft—let alone other cloud providers with large outstanding commitments.
Those details create a pointed line of inquiry for investors watching microsoft stock: if nearly half of the backlog is tied to a single customer, what is the practical meaning of “backlog” as a visibility metric for future revenue? The numbers presented do not resolve that question; they sharpen it.
Verified fact: The backlog size, its year-over-year growth, the share attributable to OpenAI, and the infrastructure investment figures are all stated directly in the provided context, along with OpenAI’s annualized revenue and its recent capital raise amount.
Informed analysis: A backlog that is both rapidly growing and heavily concentrated can simultaneously support bullish demand narratives and amplify uncertainty about timing and collectability, especially when fulfillment depends on the pace at which new data center capacity comes online and on the financial capacity of the largest named customer.
Where the debate goes next for microsoft stock
In the current framing, the stock decline is tied to modest Copilot penetration, while the company’s cloud segment is described as a critical engine for AI development and deployment. The same set of facts also highlights a capacity buildout cycle with enormous capital intensity and a backlog that is both massive and unusually concentrated.
The next phase of investor scrutiny appears to hinge on whether Copilot adoption inside 365 continues to expand beyond early pockets of heavy usage, and how Microsoft’s backlog converts as new data centers come online. Until the balance between those factors is clearer, the debate embedded in the price action remains unresolved: whether microsoft stock is “cheap” because the market is overreacting to early adoption data, or whether the market is applying a discount for execution risk hidden inside a backlog dominated by a single customer.
