Inside the AI Questions Behind Microsoft's Stock Selloff

Investors think Microsoft’s latest earnings report may signal a faultline in the hyperscaler’s AI business—and possibly AI in general. And some signs are pointing that way, but the negativity may be premature.

Microsoft's shares fell nearly 10% on Thursday after the company's earnings results were released Wednesday night. Investors may have been spooked by growth numbers that didn't live up to expectations, as well as rising capital expenditures (capex). Microsoft’s Q2 2026 report posted revenues of $81.3 billion, up 17% year-over-year (y/y), with Intelligent Cloud, the segment containing Azure, accounting for $32.9 billion, up 29% y/y. Azure and related cloud services specifically grew 39%, compared to 40% for the preceding quarter.

And Microsoft is guiding for a further decrease in Azure growth next quarter, with revenues expected to increase 37% to 38%.

AI Capex Questions Are Key

While the numbers on the face of it seem fine, there are more questions about the future. Microsoft, like just about everybody else, is spending more of its cash flow on capex to fuel AI growth. And that growth doesn't seem to be a slam dunk.

To add to the complexity, Microsoft has to balance capex for both internal AI needs as well as for its infrastructure for customers in the cloud segment. CFO Amy Hood pointed out on the conference call that results would have been higher if more of the capex were directed toward infrastructure to customers, rather than in-house investment. Hood said that it’s important to note that not all AI capex is going into Azure but is spread across other parts of the business, such as Microsoft 365 Copilot and GitHub Copilot, which Hood termed “our first-party apps.”

“[I]f I had taken the GPUs that just came online in Q1 and Q2, in terms of GPUs, and allocated them all to Azure, the KPI would have been over 40,” Hood said, apparently indicating that guidance in Azure would have been 40% if the segment had access to all Microsoft's purchased GPUs.

The bottom line is that costs are increasing, but there are many questions as to what will happen with spending for the rest of the year—and how those investments pay off. Microsoft posted capex of $37.5 billion this quarter, up from $34.9 billion last quarter. And while the company is guiding for an unspecified decrease in capex next quarter, investors on the earnings call voiced concerns that capex growth isn’t balanced by growth in Azure services. They want to see Azure ROI for the capex invested.

Despite showing adjusted net income of $30.9 billion (up 23% y/y) and earnings per share of $4.14 (up 24% y/y), investors punished Microsoft's stock. As noted, shares lost nearly 10% of their value this week.

Where Microsoft Is Losing Out in AI

While Microsoft continues to post solid growth in Azure revenues, the mismatch between capex and those revenues can’t be easily explained away. If GPUs aren’t going to Azure AI services, that means demand for those services hasn’t grown in step with investments. And Azure AI services are designed to feed enterprise inference workloads, which by all accounts are exploding.

There are other concerning signs. A year ago, Microsoft famously reduced its datacenter buildout schedule, causing concerns that the hyperscaler had misjudged demand. A year later, CEO Satya Nadella said this week that the company had added “nearly a gigawatt of capacity” in the quarter. And CFO Hood noted that Microsoft is working hard to expand its datacenters “as quickly as we can.” But some observers seem skeptical about the pace of the rollouts and what that says about infrastructure demand.

There is also the matter of OpenAI, which readjusted its relationship with Microsoft in 2025, eliminating Microsoft’s right of first refusal to offer compute services to OpenAI. Microsoft has a $625 billion backlog, but much of that (45%, according to some estimates) is attributed to OpenAI, which is losing money and needs to raise more money for that spend.

While Microsoft’s partnership with OpenAI remains tight, questions are growing about OpenAI. It’s pledged to consume $1.4 trillion in datacenter resources in its quest to develop artificial general intelligence (AGI). In the meantime, competition in the AI arena is fierce, with competitor Anthropic making significant progress.

Microsoft is indeed diversifying. It has expanded its relationship with Anthropic, adding Claude to Microsoft Foundry and integrating it with Microsoft 365 Copilot—the latter being one of the “first party apps” designated for capex instead of Azure. Still, the Anthropic relationship may also be fraught: As The Information recently reported, Anthropic’s launch of Cowork is viewed by Microsoft as competitive to Copilot.

Taken together, the slight drop in expectations for Azure’s growth and the anticipated reduction in capex for next quarter signal that Microsoft isn’t weighing in as heavily in AI as it did last quarter. Its ongoing reliance on OpenAI and its potentially rocky relationship with Anthropic also hint at future challenges.

Still, there are signs of progress. Microsoft’s recent release of its Maia 200 inference chip, which is already deployed in key datacenters in Iowa and Arizona, could boost the hyperscaler’s AI prospects. And the company’s profitability this quarter was notable. It may be too soon to give up on Microsoft’s AI future.

Futuriom Take: Microsoft's reduction in guidance for Azure growth plus its shifting capex plans have investors in a tizzy. But release of the Maia 200 chip, sizable Azure growth, and solid profitability indicate there may be better news ahead.