§ 01 Business Overview
Microsoft is the last Magnificent Seven company to appear in MCI, and the timing is deliberate. The story has changed materially in the last 90 days in ways that make the analysis richer than it would have been at any prior point.
Microsoft organizes its business into three segments. Productivity and Business Processes, which includes Microsoft 365, LinkedIn, and Dynamics, generated $35.0 billion in Q3 FY2026, up 17% year over year. This is the legacy software franchise, the business that built Microsoft into a $3 trillion company, and it is still growing at a rate most companies would celebrate as their primary growth driver.
Intelligent Cloud, which houses Azure and GitHub, generated $34.7 billion in Q3 FY2026, up 30% year over year, with Azure specifically growing 40%. This segment is now the company's growth engine, and it is where the AI infrastructure bet lives. Microsoft Cloud revenue as a whole reached $54.5 billion in the quarter, up 29%.
More Personal Computing, which includes Windows, Xbox, and Search, generated $13.2 billion and declined 1%. This segment matters less each year as the revenue mix shifts toward cloud.
Microsoft's fiscal year ends June 30. For fiscal year 2025, the company generated $281.7 billion in revenue, up 15% from $245.1 billion in FY2024. Operating income was $131.0 billion on a 46.5% operating margin. Net income was $104.9 billion. These are the numbers of the most profitable software business in history.
Then came the decision that changed the character of the company. Microsoft guided $190 billion in capital expenditure for calendar 2026, approximately $55 billion above what analysts expected. CFO Amy Hood confirmed on the Q3 FY2026 call that two-thirds of that spending goes into GPUs and CPUs for Azure and Copilot. The company that spent a decade being celebrated for capital efficiency and margin expansion has voluntarily transformed itself into a capital-intensive infrastructure business. Whether that transformation generates the returns that justify the cost is the central debate in Microsoft right now.
The most recent print, Q3 FY2026 reported April 29, 2026, came in at $82.9 billion in revenue, up 18%, with Azure at 40% growth, EPS of $4.27 against $4.06 consensus, and a commercial remaining performance obligation of $627 billion, up 99% year over year. The stock fell after earnings anyway, because the Q4 capex guidance came in above $40 billion, and investors focused on the spending rather than the results.
Two days before that earnings report, Microsoft and OpenAI announced a restructured partnership that changes the economics of the relationship in ways the market is still working through.
§ 02 Competitive Moat · Strong
Microsoft's moat operates across four distinct layers, each reinforcing the others. It is the only company in this blog that has held a dominant market position across three consecutive technology platform shifts: client-server computing in the 1990s, the transition to cloud in the 2010s, and now AI in the 2020s. Each transition should have disrupted Microsoft. None of them did, because the moat is not built on any single product. It is built on distribution.
The enterprise distribution moat. Microsoft sits inside the workflow of virtually every knowledge worker in the world. Microsoft 365 has over one billion commercial seats. Active Directory manages identity and access for the majority of enterprise networks globally. Azure runs under a significant portion of enterprise cloud workloads. This installed base is not just a customer list. It is a distribution channel that makes it possible to sell new products - including AI products - to customers who already trust Microsoft, already pay Microsoft, and already run Microsoft software in every corner of their organizations. No competitor has this channel. Google Workspace has significant consumer and education share but meaningfully less enterprise penetration. AWS has cloud infrastructure but no productivity software to bundle. Salesforce has CRM but not the OS-level access Microsoft has through Windows and Active Directory.
The Copilot monetization model. Microsoft 365 Copilot adds $30 per user per month on top of existing Microsoft 365 commercial subscriptions. At Q3 FY2026, there were over 20 million paid Copilot seats, up from 15 million in January and up 250% year over year. Weekly Copilot engagement now matches Outlook levels per CEO Satya Nadella, meaning it has crossed from optional add-on to habitual workflow tool for active users. At 20 million seats and $30 per month, Copilot generates approximately $7.2 billion in annualized revenue that did not exist before FY2025. The theoretical ceiling, if Copilot penetrates the full one billion-plus commercial seat base at $30 per month, is $360 billion in annualized revenue on top of existing subscriptions. The realistic scenario is a fraction of that, but even 10% penetration of the installed base at current pricing would add $36 billion in high-margin recurring revenue.
The business model is also shifting. Nadella said on the Q3 call that any per-user business Microsoft runs, whether productivity, coding, or security, will become a per-user and usage business. GitHub Copilot moved to consumption-based pricing effective June 1, 2026. Nearly 60% of customer service customers are already purchasing usage-based credits. A consumption layer on top of seat revenue expands free cash flow per customer without requiring new seat sales.
Azure's data gravity and platform moat. Azure grew 40% in Q3 FY2026, its fastest rate in years, against a base that already runs at over $135 billion annualized. The commercial remaining performance obligation of $627 billion, up 99% year over year, is the clearest signal of demand durability. Over 10,000 customers have used more than one AI model on Azure Foundry, with the number using both Anthropic and OpenAI models doubling quarter over quarter. Microsoft's competitive advantage in cloud AI is increasingly the platform breadth - offering customers access to every major frontier model without being dependent on any single provider. That is a different moat than Google's, which is deeply integrated with Gemini, and different from Amazon's, which is betting heavily on Anthropic and its own Trainium silicon.
The OpenAI relationship. The original deal gave Microsoft exclusive API access to OpenAI models in exchange for capital and compute. The restructured deal announced April 27 ends Microsoft's revenue-sharing payments to OpenAI, caps Microsoft's total revenue from OpenAI at $38 billion, makes the OpenAI license non-exclusive so OpenAI can serve products on rival clouds, and extends Microsoft's IP rights through 2032. Wedbush analyst Daniel Ives called it a net positive, noting Microsoft will now receive approximately $6 billion from OpenAI in 2026 compared to $4 billion previously expected. The non-exclusivity clause is the concession. Microsoft no longer controls where OpenAI deploys. That narrows the Azure advantage from OpenAI demand specifically, while preserving the broader Azure AI platform position through Foundry alternatives.
The financial snapshot connects directly to the moat. Microsoft's 46.5% operating margin in FY2025 is only possible if customers pay premium prices for software they are unwilling to migrate away from. The mechanism is enterprise distribution lock-in: Microsoft sells at premium prices because switching costs across email, productivity, identity, cloud, and security are so high that the cost of leaving exceeds the cost of staying even when a competitor offers a better individual product.
§ 03 Financial Snapshot
The three-year direction of travel is up on every line. Revenue grew from $198.3 billion to $281.7 billion, a 42% increase. Operating income grew from $83.4 billion to $131.0 billion, a 57% increase. Net income crossed $100 billion for the first time in FY2025. Operating margin expanded from 42.1% to 46.5% across three years - remarkable for a company simultaneously absorbing the beginning of a $190 billion annual capex ramp. The FY2023 year is important context: revenue grew only 7% as the post-pandemic cloud optimization cycle slowed Azure growth. That year tested the thesis that Azure had durable growth. It came through, and Azure re-accelerated from there.
| Fiscal Year | Revenue | Operating Income | Op. Margin | Net Income | Net Margin |
|---|---|---|---|---|---|
| FY2022 (Jun 2022) | $198.3B | $83.4B | 42.1% | $72.7B | 36.7% |
| FY2023 (Jun 2023) | $211.9B | $88.5B | 41.8% | $72.4B | 34.2% |
| FY2024 (Jun 2024) | $245.1B | $109.4B | 44.6% | $88.1B | 35.9% |
| FY2025 (Jun 2025) | $281.7B | $131.0B | 46.5% | $104.9B | 37.2% |
Q3 FY2026 results (most recent):
| Metric | Q3 FY2026 | Change YoY |
|---|---|---|
| Total Revenue | $82.9B | +18% |
| Productivity and Business Processes | $35.0B | +17% |
| Intelligent Cloud | $34.7B | +30% |
| Azure and Other Cloud Services | - | +40% (+39% constant currency) |
| More Personal Computing | $13.2B | -1% |
| Microsoft Cloud Revenue | $54.5B | +29% |
| Operating Income | $38.4B | +20% |
| Operating Margin | 46.3% | - |
| Net Income | $31.8B | +23% GAAP |
| Diluted EPS | $4.27 | vs. $4.06 consensus |
| Commercial RPO | $627B | +99% |
| AI Business Annualized Run Rate | $37B | - |
| Copilot Paid Seats | 20M | +250% |
| GitHub Copilot Organizations | 140,000 | - |
| Capital Expenditure | $31.9B | - |
| Q4 Capex Guidance | >$40B | - |
| Microsoft Cloud Gross Margin | ~64% | from ~69% a year ago |
Valuation: At approximately $414 to $420, Microsoft trades at 25x forward earnings, a 26% discount to its five-year median P/E of 34x. ROIC on a trailing twelve-month basis sits at 27.4%. Analyst consensus is 41 Buys, 11 Outperforms, 3 Holds, and zero Sells, with a mean 12-month price target of approximately $562 - implying roughly 35% upside from current prices. FY2030 consensus revenue estimate sits at approximately $601 billion.
§ 04 Risk Rating
The capex commitment is the central risk. $190 billion in calendar 2026, growing further in FY2027, while Microsoft Cloud gross margin compresses from 69% to approximately 64% as AI infrastructure costs hit depreciation before AI revenue fully offsets them. CFO Amy Hood acknowledged the compression directly and guided Q4 Microsoft Cloud gross margin at approximately 64%. Data centers, GPUs, and networking equipment carry 5 to 6 year depreciation schedules. Microsoft is booking the cost today on infrastructure that will generate its full revenue over the next several years. Free cash flow stays compressed in the near term. The question is whether the $627 billion RPO converts to revenue on the timeline the market requires.
The OpenAI relationship is a double-edged asset. Microsoft invested billions in OpenAI and received exclusive API access that gave Azure a distinctive advantage in enterprise AI. The restructured deal makes that access non-exclusive. OpenAI can now deploy its models on Google Cloud, AWS, and other platforms. The near-term financial improvement - $6 billion in 2026 payments versus $4 billion previously - is real but modest relative to the strategic narrowing. Microsoft's AI platform advantage increasingly depends on Azure Foundry's breadth and enterprise distribution, not on OpenAI exclusivity. That is still a strong position, but it is a different one than investors priced in when the original partnership was announced.
Copilot adoption is slower than the headline numbers suggest. Twenty million paid seats sounds large. Against the one billion-plus commercial Microsoft 365 seat base, it represents roughly 2% penetration. Copilot adoption remains in the low single digits of the Office 365 seat base, below the expectations that justified early Copilot pricing enthusiasm. The seat-plus-consumption model is the right long-term structure, but consumption revenue requires users to actually use the product intensively enough to exceed their seat allocation. Weekly engagement matching Outlook is a positive signal, but engagement does not translate linearly to consumption billing.
Antitrust exposure is real and broadening. FTC investigations into cloud practices and OpenAI partnership dynamics, European Commission scrutiny of Teams bundling with Microsoft 365, and concerns about Azure's market power in AI infrastructure all represent regulatory risk that did not exist at meaningful scale three years ago. Forced unbundling of Teams from Microsoft 365 alone would remove a competitive advantage the company has held for five years. A requirement to license Azure AI capabilities to competitors on equal terms would narrow the Copilot moat meaningfully.
The risk rating is 5 because the underlying business is executing at near-record levels, the $627 billion RPO provides the most revenue visibility in enterprise technology, and the operating margin is expanding even during the capex build, which validates that the core software franchise generates enough profit to absorb the infrastructure investment without fundamentally breaking the financial model.
§ 05 Bull vs. Bear
Bull case: Microsoft is the only company in this blog that has compounded its way through three successive technology platform shifts without losing its dominant position. Client-server computing, then cloud, now AI. Each time, skeptics argued the existing model would be disrupted. Each time, Microsoft used its enterprise distribution to sell the new paradigm to the same customers who bought the old one. Copilot is doing this again. It is being sold to enterprises that already pay for Microsoft 365, already trust Microsoft security, and already run Azure workloads. The incremental sale does not require winning a new customer. It requires getting an existing customer to add $30 per seat per month for a productivity tool that demonstrably reduces time spent on email, documentation, and code review.
At 25x forward earnings, Microsoft trades at a 26% discount to its five-year median multiple of 34x. A business with this margin profile, this RPO, and this installed base trading at a 26% discount to its own history is not obviously expensive. The $562 analyst consensus target implies 35% upside. If the capex cycle peaks in FY2027 as Hood has guided and operating leverage reasserts, the free cash flow recovery from that peak could drive multiple expansion simultaneously with earnings growth, compounding the return.
The consumption pricing shift is also underappreciated by current consensus models. When GitHub Copilot moves to consumption-based pricing in June 2026 and the seat-plus-consumption model matures across the Microsoft 365 base, revenue per user expands without requiring new seat sales. That is a margin-accretive revenue layer that is not fully captured in any current model because the pricing shift happened after most FY2026 estimates were set.
Bear case: Microsoft spent a decade being valued as a capital-light software compounder. It commanded a 34x median P/E because investors believed operating leverage would continue expanding margins indefinitely. The $190 billion capex commitment breaks that thesis structurally, at least for the duration of the build cycle. A software company with 46% operating margins that voluntarily compresses its cloud gross margin from 69% to 64% in a single year while spending $190 billion on infrastructure is not the same business that earned the 34x multiple. The appropriate multiple for a capital-intensive infrastructure utility with embedded software margins is lower. The market is working out what that number is.
The OpenAI non-exclusivity concession matters more over a three-year horizon than the market is currently pricing. OpenAI's models are the most capable and widely adopted in the enterprise market. When OpenAI can deploy on Google Cloud or AWS with equal terms, the Azure advantage from GPT-4o and future OpenAI models becomes a feature rather than a moat. Google Cloud has Gemini deeply integrated. AWS has Trainium and Anthropic. Microsoft had OpenAI exclusively. Now it has OpenAI non-exclusively plus everything else on Azure Foundry. That is still a strong platform, but the differentiation has narrowed.
Copilot at 2% penetration of the installed base after 18 months of aggressive enterprise marketing is a slower ramp than the bull case requires. If penetration plateaus at 5 to 8% of the commercial seat base rather than converging toward 20 or 30%, the Copilot revenue contribution over five years is a meaningful business but not the transformative margin driver the current multiple implies.
Buy on Weakness. Entry interest: $375 to $400. At approximately $414 to $420, Microsoft is fairly valued for the business it is today. Twenty-five times forward earnings on 18% revenue growth, 46% operating margins, and a $627 billion RPO is not expensive in absolute terms. But it is not cheap either, and the capex cycle creates a specific risk: if cloud gross margin continues compressing toward 62 to 63% in FY2027, the multiple the market is willing to pay contracts simultaneously with near-term earnings, creating a double compression that the current price does not fully compensate for.
At $375 to $400, Microsoft trades at approximately 22 to 23x forward earnings on a business that has compounded through three technology cycles, holds the largest enterprise software distribution channel in the world, and is running its Copilot monetization at a $7.2 billion annualized run rate growing 250% year over year. That range prices in meaningful capex headwind while still paying for one of the most durable business models in large-cap technology.
At $375 to $400, the capex risk is more adequately compensated, and the upside from operating leverage reasserting in FY2028 is significant.
§ 06 What to Watch
Microsoft 365 Copilot net paid seat additions in Q4 FY2026. Hood guided for sequential growth in net seat adds. If the Q4 report, expected July 30, 2026, shows acceleration from the 5 million net adds implied between the January and April data points, the monetization thesis gains conviction. If seat adds decelerate, the penetration ceiling concern becomes the dominant narrative.
Microsoft Cloud gross margin trajectory. At approximately 64% in Q3 and guided at approximately 64% in Q4, the question is whether margin stabilizes there or continues compressing into FY2027 as more capex hits depreciation. Any quarter where Microsoft Cloud gross margin recovers toward 66 to 67% signals the infrastructure build is ahead of schedule in converting to billable utilization.
GitHub Copilot consumption revenue in Q4 FY2026. The move to consumption-based pricing went live June 1. The first quarter that includes meaningful consumption billing data will be Q1 FY2027, reported October 2026. That print will be the earliest hard data point on whether the seat-plus-consumption model generates the per-customer revenue expansion that justifies the pricing shift.
Commercial RPO conversion rate. $627 billion in RPO growing at 99% year over year is the most important forward-looking indicator in Microsoft's results. Watch whether the ratio of RPO to actual revenue holds or expands. If RPO grows faster than reported revenue, customers are committing further ahead than they are consuming, which is a sign that supply, not demand, is the constraint. If revenue begins catching up to RPO growth, it signals the infrastructure build is coming online and converting contracted backlog to actual revenue.
OpenAI's cloud deployment decisions. Now that OpenAI can deploy on any cloud provider, where OpenAI chooses to run the most compute-intensive workloads becomes a real-time signal of which cloud has the best price-performance. If OpenAI expands materially on AWS or Google Cloud for specific workloads, it narrows Azure's distinctive advantage in the most important AI workload category.
This analysis introduced the concept of the capex cycle and why it creates a temporary valuation dislocation that patient investors have historically been rewarded for tolerating.
Most software companies are valued on a multiple of earnings or free cash flow. The implicit assumption is that earnings today are a reasonable proxy for earnings in the near future. When a company enters a period of intense capital investment, that assumption breaks down. Current earnings are depressed by depreciation on infrastructure that is not yet generating its full revenue potential. Current free cash flow is depressed by cash outflows that will generate returns over a 5 to 10 year asset life.
The historical parallel is Amazon from 2013 to 2016. AWS was consuming capital at a rate that compressed Amazon's reported earnings to near zero. Investors who evaluated Amazon on current earnings concluded the business was barely profitable. Investors who evaluated Amazon on the contracted AWS backlog and the margin profile of the cloud business at scale concluded the stock was mispriced. They were right.
Microsoft's situation is structurally similar. The $627 billion commercial RPO represents contracted future revenue that has not yet appeared in the income statement. The $190 billion in capex represents investment in infrastructure that will depreciate over 5 to 6 years but will generate revenue for the duration of that useful life and beyond. Valuing Microsoft on current earnings during the peak capex year is like valuing Amazon on 2014 earnings. The number is real but it is not representative of the earnings power the business will demonstrate once the build cycle normalizes.
Looking across the full MCI portfolio, this analysis completes the capex comparison that Week 07 Meta, Week 08 Alphabet, and Week 10 Amazon started. Meta is spending $135 billion, Alphabet $185 billion, Amazon $200 billion, and Microsoft $190 billion. The combined hyperscaler capex in 2026 approaches $725 billion. Nvidia is the primary beneficiary of that spending. Micron's HBM business exists because of it. The entire AI infrastructure economy, every stock in this blog from Palantir to Amazon, sits downstream of whether that $725 billion generates the returns that justify the investment. Microsoft's $627 billion RPO is the single largest piece of evidence that it does.