Microsoft at 14x EBITDA: The Market Has Not Priced the Cloud Margin Inflection

Microsoft is no longer primarily a software company. It is the second-largest cloud infrastructure provider on the planet, an enterprise productivity monopoly, and an AI platform business that controls the distribution layer for OpenAI's technology.

The three reporting segments tell the story. Intelligent Cloud (Azure, Azure AI, server products) is now the dominant growth engine and margin contributor. Productivity and Business Processes (Office 365, LinkedIn, Dynamics) is a mature, high-margin annuity that requires minimal incremental capital. More Personal Computing (Windows, Xbox, Surface, Search) is largely ex-growth but generates cash.

The critical insight is that Azure has crossed the profitability inflection point that every hyperscaler eventually reaches: the fixed cost base is paid, and each incremental dollar of cloud revenue flows through at margins far above the corporate average. Microsoft does not break out Azure operating margins directly, but the trajectory of the Intelligent Cloud segment's contribution leaves little room for ambiguity.

This is the machine that generated $305 billion in trailing revenue in FY2025, up from $168 billion in FY2021. A 67% increase in the top line over four years at a 45% operating margin. That combination does not happen by accident.

The revenue growth story is well-understood. The margin story is not appreciated enough.

In FY2021, Microsoft reported $168.1 billion in revenue and $69.9 billion in operating income. Operating margin: 41.6%. By FY2025, revenue reached $281.7 billion and operating income hit $128.5 billion. Operating margin: 45.6%. Four percentage points of margin expansion on a $282 billion revenue base is not incremental improvement. It is an additional $11 billion in annual operating income that appears from scale alone.

Gross margins have held steady in the 68-70% range throughout this period, which means the operating leverage is coming from the operating expense line. Microsoft is growing SG&A and R&D more slowly than revenue. On a business of this size, that discipline compounds into extraordinary numbers.

EBITDA reached $160.2 billion in FY2025, up from $85.1 billion in FY2021. Nearly doubling EBITDA in four years while maintaining exceptional product quality and executing the largest enterprise AI transition in history. The earnings record backs this up: Microsoft has beaten consensus EPS estimates in every single quarter over the past two years, with beats ranging from 1.6% to 7.9%.

The structural argument for Microsoft is that cloud computing is not a commodity and Azure is not losing the race for second place. The enterprise IT migration from on-premises to cloud is still roughly halfway through. The workloads that moved first were the easiest: storage, development environments, productivity applications. What remains is the harder, stickier, and more valuable core: ERP, data warehousing, mission-critical applications, and now AI workloads.

Microsoft has a distribution advantage that Alphabet and Amazon cannot replicate in the enterprise segment. Every large company in the world already pays Microsoft for Office, Azure Active Directory, and Windows. The sales motion for an additional Azure workload or a Microsoft 365 Copilot seat is an expansion conversation with an existing customer, not a land-and-expand cold start. That reduces customer acquisition cost to near zero on a marginal basis.

The Azure OpenAI service has added a new monetisation vector. Microsoft charges per token for API access to GPT-4 and its successors, and enterprise customers are consuming these tokens at scale for document processing, code generation, customer service automation, and workflow integration. Microsoft's April 2026 announcement of expanded AI alliances reframing its role in patents and security signals further commercial deployment of AI capabilities across the enterprise stack.

The key number to watch is Azure revenue growth rate quarter over quarter. When it reaccelerates above 35%, the multiple compression story collapses entirely. The consensus estimate for Q3 FY2026 (the March quarter, reporting next month) is $4.09 EPS. The history of the last eight quarters suggests that number is a floor, not a ceiling.

The bear case on Microsoft starts and ends with capital expenditure. In FY2021, capex was $20.6 billion. In FY2025, it reached $64.6 billion. That is a tripling in four years, and it has compressed free cash flow even as operating cash flow exploded.

Free cash flow in FY2021: $56.1 billion. Free cash flow in FY2025: $71.6 billion. Operating cash flow, however, went from $76.7 billion to $136.2 billion over the same period. The machine is generating cash at an extraordinary rate. The capex is consuming it before it reaches shareholders.

The critical question is whether this capex is maintenance or growth. The evidence strongly suggests the latter. Hyperscaler data centre spending follows a predictable cycle: heavy upfront investment in GPU clusters and power infrastructure, then a long tail of revenue monetisation as customers commit to long-term contracts. The revenue from AI workloads signed today will show up in Azure growth numbers for years. Today's FCF compression is tomorrow's FCF expansion.

Buybacks have slowed as a result: $32.7 billion in FY2022 versus $18.4 billion in FY2025. The dividend has grown steadily, from $16.5 billion in FY2021 to $24.1 billion in FY2025. Microsoft is not ignoring shareholders during the capex cycle, but it is prioritising infrastructure build over capital return. The bet is that the infrastructure earns more than the buyback multiple would have. Given Azure's growth trajectory, that bet looks correct.

Microsoft 365 Copilot is priced at $30 per user per month on top of existing Microsoft 365 subscriptions. The commercial Microsoft 365 base runs to hundreds of millions of seats. If Copilot achieves even 10% penetration at full price, the incremental annual revenue exceeds $10 billion with very high incremental margins.

Penetration is not yet at 10%. The product launched broadly in late 2023, and enterprise adoption follows a familiar cycle: pilot, evaluate, negotiate enterprise pricing, deploy, expand. Most large enterprises are still in the evaluate-to-negotiate phase. The meaningful revenue contribution from Copilot is a 2026-to-2028 story, not a FY2025 story.

This is worth understanding when reading the current valuation. The market is pricing Microsoft on what it has earned. It is not pricing in a Copilot deployment cycle that has not yet started in earnest. The forward estimates for FY2027 and FY2028 embed some Copilot contribution, but consensus AI penetration assumptions tend to be conservative in the early innings.

The Azure OpenAI service adds another layer. Enterprise customers consuming AI via API are on consumption-based contracts, which means usage growth translates directly to revenue without requiring additional sales cycles. The AI workload mix is shifting toward inference (running completed models at scale) from training, which is a more predictable and stickier revenue stream.

The AI competition narrative frames Microsoft against Google, Amazon, and increasingly Anthropic and other foundation model providers. This framing misses the point. Microsoft is not primarily competing on model quality. It is competing on distribution.

Every Fortune 500 company has an existing Microsoft enterprise agreement. Every IT department has Microsoft 365 admins who already manage the tenant. Every developer at a large company already uses GitHub, which Microsoft owns. Every Azure customer already has a billing relationship. The incremental friction of adopting Azure AI services, Copilot, or GitHub Copilot Enterprise is near zero compared to adopting a stand-alone AI vendor.

Google has a comparable cloud platform and strong model capability. It does not have the enterprise productivity installed base. Amazon has the largest cloud platform and excellent AI services. It does not have the enterprise seat density in productivity software. Microsoft has both, and both are growing.

The competitive moat is not technological. It is relational and contractual. Enterprise IT procurement is slow, risk-averse, and heavily influenced by vendor consolidation trends. Microsoft is the obvious consolidation beneficiary. When a CTO is trying to reduce the number of vendors, they are more likely to deepen a Microsoft relationship than to add a net-new AI vendor.

Microsoft's trailing PE of 23.4x is near the low end of its five-year range. The EV/EBITDA of 14.6x is a number associated with mature, slow-growing industrials, not a business compounding EBITDA at 17% annually.

On a price-to-sales basis at 9.1x, Microsoft is cheaper than almost every high-growth software company and at a discount to Nvidia. On a price-to-book basis of 7.1x, the intangible asset intensity of the business (software, cloud infrastructure, brand) means this number understates economic value significantly.

The analyst community is not missing the picture: 41 strong buy ratings, 15 buy ratings, and 5 holds, with a consensus price target of $587. At a current implied price around $374 based on market cap and shares outstanding, that represents approximately 57% upside to analyst consensus. This level of buy-side conviction with a 23x trailing PE is unusual. It reflects a view that current earnings do not reflect the earnings power of the business once the capex cycle normalises.

For reference, if capex returned to FY2022 levels of $23.9 billion and operating cash flow held at FY2025 levels of $136.2 billion, free cash flow would be approximately $112 billion. At the current enterprise value, that implies a free cash flow yield above 4% on a business growing its earnings at double digits. That is not an expensive business.