Microsoft: The Backlog Is the Moat

Microsoft sells three things: the software an enterprise runs its people on (Microsoft 365, LinkedIn, Dynamics), the cloud it runs its systems on (Azure, server products, GitHub), and the consumer surface of Windows, Xbox and search. The company reports them as three segments, and in the quarter ended 31 March 2026 they diverged sharply: Productivity and Business Processes revenue was $35.0 billion, up 17%; Intelligent Cloud was $34.7 billion, up 30%; and More Personal Computing was $13.2 billion, down 1%.¹⁶

Total revenue was $82.9 billion, up 18%, with operating income of $38.4 billion (up 20%), net income of $31.8 billion, and diluted earnings of $4.27 a share.¹ Microsoft's own summary attributes the $12.8 billion of added revenue to "growth in Microsoft Cloud," with Intelligent Cloud "driven by Azure" and Productivity "driven by Microsoft 365 Commercial cloud."⁴ The headline numbers are strong. The point of this dossier is that the interesting questions sit underneath them: what actually protects Microsoft's cloud business, and what it is spending to defend it.

The engine is Azure, and it is still accelerating

Microsoft Cloud — the company's roll-up of Azure, the commercial portions of Microsoft 365 and Dynamics, and LinkedIn's commercial services — reached $54.5 billion in the quarter, up 29%, the first time it has cleared $50 billion in three months.¹ Inside that figure, Azure and other cloud services grew 40% year over year, a rate that has, unusually for a business this large, risen: the prior quarter grew 39% on a Microsoft Cloud base of $51.5 billion, and Intelligent Cloud stepped up from $32.9 billion to $34.7 billion in a single quarter.³ A $200-billion-a-year run-rate accelerating rather than fading is the single most important fact about the company. It matters because acceleration at scale usually signals genuine demand rather than a one-off migration: the base is already so large that only a broad, repeated buying pattern — new AI workloads layered on top of existing Azure consumption — can move the growth rate up rather than letting the law of large numbers drag it down.

The chart shows why the mix matters. Intelligent Cloud has drawn level with the historically dominant Productivity segment, and the gap between them and the stagnant consumer business is now enormous. Growth and the growth narrative both run through Azure.

The moat: switching costs you can measure

"Strong business" is not analysis. The precise moat mechanism here is contracted lock-in, and it has a number. Microsoft's commercial remaining performance obligation the dollar value of contracted revenue a company has sold but not yet delivered — future Azure consumption, multi-year Microsoft 365 seats, signed but unrecognised bookings rose 99% year over year to $627 billion in the quarter.¹ That backlog is roughly twice the company's annual revenue, already under contract. It is the clearest evidence that customers are not renting Azure month to month but committing to it for years — the definition of a switching cost. The near-doubling matters as much as the level: a backlog growing far faster than reported revenue means customers are signing longer and larger commitments than they are currently consuming, which is exactly the pattern you would expect if enterprises were reserving AI capacity ahead of deploying it.

The Productivity segment shows the same lock-in from the other side. At $35.0 billion and 17% growth, it is driven by Microsoft 365 Commercial, where AI features are sold as an add-on to a seat an enterprise has already standardised on for identity, mail and documents.⁴ Displacing that means re-tooling how a company's employees work, which is why so little of it churns.

The steelman for the bear is Microsoft's own language: it competes in markets where technology shifts fast and entry is cheap, and it says so in its annual report.⁶ The backlog is the counter-evidence — but only if the company can physically serve it.

The bet: build the capacity or forfeit the backlog

Serving that demand is extraordinarily capital-intensive. Microsoft spent $30.9 billion on property and equipment in the single quarter, and $80.1 billion over the nine months to 31 March 2026 — sums that would have been an entire year's capital budget a few years ago.² The cost is already visible in margins: Microsoft Cloud gross margin slipped to 68%, which management attributes to "scaling our AI infrastructure and growing usage of AI product features," down from 69% a year earlier.⁵⁷

What keeps the story from being a simple margin-compression trade is operating leverage: even while absorbing that capex and the lower-margin AI mix, operating income still grew 20% — faster than the 18% on revenue — so the software business is expanding profit fast enough to fund the build without the whole model de-rating.¹ The depreciation from today's data centres, however, will land on the income statement for years, whether or not the AI revenue arrives on schedule; that timing mismatch is the crux of the bet.

A large part of that AI demand — and a large part of the spend — runs through Microsoft's relationship with OpenAI, whose models power Copilot and much of Azure's AI revenue and whose training and inference run on Azure hardware. Microsoft accounts for its OpenAI stake under the equity method, so its share of OpenAI's losses lands in "other income (expense), net," which the company reported at $942 million for the quarter.² The paper's read: this is the central tension in the stock. The capital is being committed years ahead of the revenue it is meant to earn, and a meaningful share of the demand thesis rests on a single partner. If the backlog converts, the spend is the cheapest moat money can buy; if AI consumption softens, Microsoft will own the most expensive under-utilised infrastructure in the industry.

The tape

Walled off from the business above: the market's reaction to this quarter turned not on the growth line but on the capital line. A clean beat on revenue, Azure and margins was met with a soft share reaction, because investors are now underwriting tens of billions of quarterly capex against revenue that is contracted but not yet delivered — the $627 billion backlog is precisely the promise the spend is meant to fulfil.⁸ The tape, in other words, is pricing conversion risk, not demand risk. That is a narrower worry than it looks, and a real one: the market has largely conceded that Microsoft can sell the capacity, and is now asking only whether it can build and depreciate it at an acceptable return. The business is the story; the tape is the footnote.

Go deeper

Read the FY26 Q3 10-Q's cash-flow statement and its investments note to see the capex line and how the OpenAI equity-method losses flow through other income.² Then read the segment and competition sections of the FY2025 10-K to hold Microsoft's own account of its moat against the market's.⁶ The numbers to track next quarter: whether Azure growth holds near 40%, whether Microsoft Cloud gross margin keeps sliding as AI capacity comes online, and whether the commercial backlog keeps compounding — that trio, together, tells you if the capital bet is working.