Microsoft FY2026: a bigger cloud backlog still has to earn a return

How to frame the business

Microsoft closed fiscal 2026 with two unusually large signals of commercial demand. Azure and other cloud services revenue increased 43% in the fourth quarter, while commercial remaining performance obligation reached $678 billion, up 84% from a year earlier. Those figures support the view that enterprise demand is broad and that customers are making long-duration commitments. They do not mean the full backlog is near-term revenue or that every contracted dollar will carry the same margin. The accounting schedule and the cost of delivering the service matter as much as the headline size.

The composition of remaining performance obligation is the first place to slow down. Management said roughly 30% of the $678 billion balance should be recognized as revenue during the following 12 months and that the weighted average duration, including OpenAI, was 2.3 years. It also said RPO growth was 25% when OpenAI was excluded. The reported 84% growth rate is therefore real, but it combines a very large customer commitment with a broader commercial base. A useful analysis keeps both facts visible instead of treating the total as a simple proxy for next year's sales.

Revenue conversion is already visible in the cloud results. Microsoft Cloud revenue was $59.3 billion in the fourth quarter and $214.4 billion for the fiscal year, with annual growth of 27%. Management also said nearly 90% of full-year cloud revenue came from customers outside frontier-model companies. That provides evidence of breadth beyond a small group of AI laboratories. Still, Microsoft Cloud is an aggregate measure covering Azure, Microsoft 365 Commercial cloud, the commercial portion of LinkedIn and Dynamics 365. It should not be read as a pure measure of infrastructure demand.

The capacity bill is substantial. Microsoft reported fourth-quarter capital expenditures of $41 billion, including finance leases, and cash paid for property and equipment of $35.8 billion. Management said roughly two-thirds of capital expenditures went to shorter-lived assets, primarily CPUs and GPUs, with the balance directed to longer-lived assets such as data-center sites. Operating cash flow was $55.4 billion and company-defined free cash flow was $19.6 billion for the quarter. The business funded its investment internally, but free cash flow did not rise in line with revenue because the expansion requires cash before all related revenue is recognized.

Margins show the same timing tension. Fourth-quarter company gross margin was 67%, down from the prior year, and Microsoft Cloud gross margin was 65%, also lower. Management attributed the pressure to a greater Azure mix, continued AI infrastructure investment and rising product usage, partly offset by efficiency gains. At the same time, Intelligent Cloud operating income grew 31% and its operating margin remained about 41%. The evidence is not that AI infrastructure has destroyed profitability. It is that rapid growth can coexist with lower gross-margin percentages and higher operating profit dollars.

Reported earnings also need a clean bridge to operating performance. Fourth-quarter revenue was $90.0 billion and operating income was $40.6 billion, both up 18%. GAAP net income increased 31%, but Microsoft identified investment-related gains and presented non-GAAP results excluding the impact of investments in OpenAI. The distinction matters because changes in the value or accounting treatment of strategic investments can move net income without describing the economics of Azure, Microsoft 365 or the underlying software franchises.

Valuation lens

A valuation should connect three stages rather than assigning one multiple to the word AI. The first stage is contracted demand: how much of remaining performance obligation becomes revenue, over what period and with what concentration. The second is physical capacity: the cash cost, lease commitments, utilization and replacement cycle of CPUs, GPUs, networking and data-center sites. The third is monetization: revenue per unit of capacity, product gross margin and the operating expenses required to win and support customers. Microsoft's mix of subscriptions and consumption services can smooth demand, but it cannot eliminate the economics of infrastructure. Investors should also track management's disclosed change in estimated useful lives for data centers and office buildings from 15 to 25 years beginning in fiscal 2027. A longer useful life changes the timing of depreciation and some lease classification effects; it does not change the cash already committed to the assets.

What can break the thesis

The central risk is a mismatch between the pace of investment and the pace of profitable consumption. Capacity can arrive late and constrain revenue, or arrive early and depress utilization. Large customer commitments can make backlog growth look broader than it is, while contract duration delays the conversion to reported revenue. Hardware prices, power availability, construction schedules and rapid accelerator replacement cycles can all affect returns. Product usage may grow while gross margin falls if inference and other compute-intensive services are not priced efficiently. Outside cloud, Microsoft still carries slower or declining businesses, including weaker fourth-quarter Windows OEM and Xbox results, so consolidated growth should not be assumed to move uniformly across the portfolio.

Bottom line

Microsoft ended fiscal 2026 with strong evidence of cloud demand, a much larger contracted revenue base and operating profit growth alongside heavy infrastructure spending. The open question is not whether customers want capacity. It is whether each new dollar of capacity produces enough recurring revenue and cash over its useful life. The best confirmation would be continued Azure growth, a broader mix within RPO, stable or improving cloud margins as capacity scales and free cash flow that begins to reflect the revenue already contracted. Until then, backlog and capital expenditure should be analyzed as one economic system, not as separate bullish statistics.