
Last 4 earnings calls · through Q4 FY2026
The bull case is that demand for Azure and AI services still exceeds the capacity Microsoft can bring online, so revenue growth is set by construction speed rather than by customer interest.
Copilot seat additions keep accelerating and Microsoft is layering usage-based charges on top of seat licenses, which raises how much each large customer can eventually spend.
Revenue is growing faster than operating expenses and the dividend and buyback continue, which bulls read as evidence that the infrastructure spending is disciplined.
The bear case is that Microsoft is building data centers far faster than AI revenue arrives, so depreciation and component costs weigh on margins before the bet pays off.
A large share of the contracted backlog comes from one AI developer, so the durability of that revenue depends on a single partner's ability to keep paying.
Bulls treat unmet demand as an asset, but bears note that unfinished data centers mean revenue forgone now, which competitors with spare capacity can capture instead.
Windows and Xbox are contracting as higher memory prices lift device prices and the gaming business resets, leaving the company more dependent on cloud and AI growth.
In short: Azure and Copilot demand is outrunning the capacity Microsoft can build, while PC and gaming revenue shrink.
Azure is now Microsoft's largest and fastest-growing product line, and management expects growth to accelerate again in the first half of fiscal 2027.
Management once expected to catch up with demand inside the fiscal year and now says supply stays short, which caps Azure revenue and lets rivals take workloads.
Component cost inflation is pushing device prices up and gaming is being reset, so the consumer segment subtracts from total growth while cloud carries the company.
In short: Operating margins hold near record levels, but AI infrastructure is lowering gross margin and absorbing cash.
Management moved from calling for only slightly higher full-year margins to a full point of expansion while still funding the AI build.
Microsoft returned $48.7 billion to shareholders, about 73% of free cash flow, which shows the AI build has not yet forced it to cut the dividend or buyback.
Sales are shifting toward Azure and AI features cost more to run, so each dollar of revenue carries more infrastructure cost than the older software mix did.
Management guides capital spending higher again next fiscal year, so free cash flow now depends on how quickly that capacity is sold and turned into revenue.
In short: Microsoft offers the widest model choice and agent controls, but one AI partner supplies much of its backlog.
Because customers can mix OpenAI, Anthropic, open-source and Microsoft's own models on one platform, Azure demand does not hinge on any single model winning.
Identity, security and governance for agents is a new category sold into the existing customer base, which raises switching costs for large organizations.
One AI developer supplies a large share of contracted revenue, and its own sales are far smaller than what it has committed to spend with Microsoft.
Microsoft sells AI in many jurisdictions, and new rules can slow deployments or force changes to models and data handling in its largest overseas markets.