The business
Microsoft sells the tools a company runs on. Microsoft 365 (Office, Teams, Exchange, SharePoint) is where work happens; Entra ID decides who can log in to what; Windows and GitHub are where software gets written; Azure is where much of it runs. Dynamics and LinkedIn sit on the same customer relationship.
Most of it is sold as seats on multi-year enterprise agreements, which makes revenue unusually predictable. Azure is metered: customers pay for compute, storage and, increasingly, AI inference, which is where the fastest growth now comes from.
In FY2026 (year ending June) revenue was $331.8B, up 17.8%, with an operating margin of 46.8%.
Moat and switching costs
The moat is integration, not any single product. An enterprise that standardised on Entra ID for identity, Microsoft 365 for documents and email, and Azure for workloads faces a switching project that touches every employee and every compliance control. Alternatives exist for each piece (Google Workspace, Okta, AWS), but replacing the set is rare.
AI raises the stakes in both directions. Copilot is sold into the existing seat base, which few rivals can match. But model capability is moving fast, and the partnership with OpenAI means some of the value created sits outside Microsoft's control.
How the technology shows up in the numbers
Software at this scale is extraordinarily profitable: gross margin near 68% and operating margin near 47% in FY2026. What changed is capital intensity. Free cash flow was $67.0B in FY2026, down from $74.1B two years earlier while net income rose to $133.7B, because data-centre spending for AI absorbed the difference.
That gap is the key number to watch: if AI revenue earns a return on the capacity being built, free cash flow catches back up with earnings; if not, Microsoft has turned part of a software business into a utility.
Thesis
Microsoft owns the distribution for enterprise AI: the seats, the identity layer and the cloud. If AI workloads earn software-like returns on the capacity being built, today's depressed free cash flow margin (20%) recovers toward its historical 30%+ and earnings keep compounding in the mid-teens.
We would own it below our stated price, which leaves room for the capital cycle to disappoint.
Risks
- AI capex without matching returns. The largest risk is spending tens of billions a year on capacity that is later priced as a commodity.
- Model dependence. Much of the AI value depends on a partner whose interests do not always align.
- Regulation of bundling (Teams was already unbundled in Europe) and of cloud licensing.
- Seat growth slowing as enterprises automate roles that used to need a licence.
What would change our mind
- Free cash flow failing to recover toward net income over the next two to three years while capex keeps rising.
- Azure growth falling below the growth of the overall cloud market for several quarters.
- Evidence of enterprises consolidating away from Microsoft 365 seats.
