Results: Microsoft 4q26 (calendar 2q26)
Excellent results, confirming major thesis points.
Original review: Part 1, Part 2, Part 3.
Tag for finding my other articles on this stock: MSFT
Key takeaways
17% constant currency revenue growth, with operating leverage despite pressure on gross margins from the infrastructure buildout. Ebit grew 18%.
Cloud and Azure grew 31% and 43% respectively, and they’re now roughly two-thirds and one-third of Microsoft’s overall revenue, so overall revenue may accelerate as they increasingly dominate the mix.
Copilot paid seats now 30m, up 50% q/q. This is perhaps the simplest single measure of whether Microsoft is succeeding as the distributor of commoditised intelligence to enterprise.
Additional thoughts
Excellent results, especially given the More Personal Computing segment is struggling on weak PC sales and Xbox results.
In Part 2 of Can Microsoft Compete? (see links above) I discuss my view that intelligence will commoditise, with multiple models competing for workloads behind the frontier. I think the evidence for this is hardening. Microsoft is building a model-agnostic harness that will automatically route demand to the cheapest model for each particular use case, and will protect an enterprise’s knowledge and data in the process. That’s obviously more appealing for a customer than being locked into one frontier LLM provider that might not remain at the frontier, will always want higher pricing, and might use customer data for its own ends.
The change in depreciation schedules is noise. Microsoft have extended useful lives for data centre and office buildings from 15 years to 25 years. They mostly lease these buildings. When a lease accounts for a large portion of an asset’s life it is a finance life. When it does not, it is an operating lease. This change therefore shifts building leases from finance to operating. This does not change actual cash flows one iota, but operating costs will rise marginally and financing costs will fall marginally. In addition, finance leases are considered capex and operating leases are not, the logic being that if you lease a building for most of its useful life you are effectively borrowing money to buy the asset. Therefore, as buildings shift from finance to operating leases, reported capex will fall. Some analysts are suggesting this is a red flag, implying that Microsoft chose to change depreciation schedules in order to massage its capex number down. I disagree. A 15 year asset life is too short for buildings and I have never understood it. 25 years is more reasonable, especially in light of Andy Jassy’s disclosure that Amazon expect to be monetise their data centre buildings over “30 plus” years. If anything, the implication is that Microsoft has overstated past capex.
ROI on capex. I see a lot of negative commentary on this. Many investors seem to have a huge issue with the fact that we do not know that the ROI will be strong. But when do you ever know? All growth projects involve risk. I think the risks here are manageable for a number of reasons. AI is a transformational technology and I think underestimating long term demand is easier than overestimating it. Certainly demand is well ahead of supply today - this is clear from the comments of all the hyperscalers, the fact that Azure revenue accelerates whenever Microsoft has capacity to meet it, and spot pricing being well ahead of contract pricing. There are also big braking factors on supply, including power and the capacity to manufacture chips. As with all major buildouts, capacity may exceed demand occasionally, and when it does the bears will dance a little jig, but it is fairly easy to slow capex and I expect demand to keep growing, so I would not expect those periods to last long. There are concerns over the asset lives of GPUs and CPUs, but these are mainly driven by obsolescence risk, and chip depreciation schedules are far shorter than time until burnout, so significant obsolescence risk is already baked into the accounting. (In fact the real risk may be to the upside, if chip lives can be extended by disaggregating prefill and inference - Gavin Baker discusses this on a recent Invest Like the Best podcast.) Finally, as I discuss in my review, the hyperscalers are effectively building a new capex-heavy business alongside their existing and growing capital light one. If infrastructure commoditises and there is no ROI, then they don’t need to own it - they will have created a new, separate, and huge utility industry focused on delivering cheap compute. That’s not a bad outcome for the capital-light orchestration and agentic layers, which will enjoy massive TAM expansion as intelligence gets cheaper.
In short I think this was a strongly thesis-confirming quarter and Microsoft remains good value, trading at 23-24x P/E for a company likely to grow in the teens with deep moats.
Notes
FY revenue $331bn, +18%, with cloud $214bn + 27% and Azure $100bn +41%. Ebit +21%.
4q revenue +17% c/c, ebit +18%. EPS +23% - this excludes the mark-to-market for OpenAI, but EBIT better represents underlying growth.
CFOPS +30% to $55bn.
$41bn capex split 1/3 long term and 2/3 short term.
$20bn FCF on a cash basis i.e. excluding finance leases from capex.
Commercial RPO +84%, with all q/q growth coming from non-frontier customers.
Extending useful life of data center and office buildings from 15 to 25 years, “reflecting our operating history and expected use of these assets”. This has a minimal impact on 2027 ebit but reduces reported capex by $15bn from $190bn to $175bn without actually impacting money spent.
Productivity and Business Processes revenue +14%, ebit +14%, with margins depressed by investments in Copilot.
Paid M365 Commercial seats grew 6% year-over-year, with ARPU growth from Copilot and premium tiers.
M365 consumer cloud +22%, 7% seats and rest ARPU.
LinkedIn +10% in constant currency.
>30m paid Copilot seats, with net seat adds more than doubling q/q and the number of customers with >50k seats up 7x y/y.
Copilot Cowork is now generally available and they have introduced Autopilots: autonomous, long-running agents with full enterprise compliance.
A Copilot superapp is coming, bringing chat, Cowork, Autopilot, and Code into one place, alongside skills for the other applications in the enterprise like the CRM and ERP, giving an enterprise-wide AI tool that is pre-wired into existing IT Ops and Security setups - this will “change what people think of M365 capabilities” and massively expand the TAM.
Copilot performance is improving, with satisfaction rates 2x over 3 quarters, conversations per user up 2x y/y, weekly engagement on par with Outlook and Teams, and the time from deployment 80% MAU across a customers’ user base down from months to days.
New E7 suite brings together Copilot, E5, Entra, and Agent 365 and has sold millions of seats in 2 months since launch.
Seat + consumption models expand TAM and are driving “significant revenue”.
Intelligent Cloud revenue +31% (Azure +43%), ebit +31%, with gross margins down on the mix shift to Azure, the infrastructure buildout, and higher Copilot usage, partially offset by strong operating leverage.
Revenue acceleration driven by efficiency gains across the CPU and GPU fleet, earlier delivery of new capacity which was quickly monetized, and GitHub Copilot’s shift to consumption-based pricing (which also started improving gross margin on this product).
Azure has “an incredibly diverse book of business by geo, by segment, by industry” plus a big first party app business which allows them to manage a slowdown in demand.
Customers building with multiple model providers up 5x y/y. Microsoft is building “a new model system where the harness, context, memory, and action space are separate from any one model family”, reducing cost per outcome and improving business continuity and resilience because every model is substitutable.
“Every firm is going to evaluate who are the providers who are helping them with their outcomes and their knowledge creation… This is not going to be about come in and take all my knowledge and benefit yourself…[so] you’ve got to keep your harness separate from the model…the harness will ensure that your memory, your context, all of that is external. That means any given model at any given time is swappable… We ourselves are using [this architecture]. Copilot is built that way. GitHub Copilot is built that way. Our Security Copilot is built that way. And we want to democratize that design pattern so that every enterprise can use it.”
>12 new proprietary models this year across image, voice, transcription, coding, security, and reasoning all targeting cost-efficient inference for enterprise use cases and all designed for proprietary silicon, giving 40% better performance per watt when running MAI models on Maia 200. E.g. in Excel, MAI-Code-1-Flash is delivering comparable quality to GPT-5.6 for the most common task while operating at significantly lower costs and in security, MAI-Cyber-1-Flash achieves better performance than Mythos at half the cost.
“The data estate is evolving from primarily supporting apps used by people to supporting agents”. AI-optimized databases like Cosmos DB and PostgreSQL give agents fast, secure access to real-time data and context. PostgreSQL revenue +55% and accelerating, paid Fabric customers +60%, Foundry revenue more than doubled.
Azure demand continues to exceed supply. If this changes they can slow the purchases of shorter-lived assets like GPUs and CPUs immediately. These are now the majority of capex and the driver of COGs.
On track to 2x compute in 2 years, and efficiency rising, with Copilot throughput up 4x. Maia 200 scaling and delivers 30% better performance per dollar than the latest generation hardware in the fleet. Cobalt CPUs also scaling.
Component pricing is spiking, hurting capex, but efficiency helps offset this and on-prem pricing is rising even faster, so cloud is more competitive. Contract pricing reflects input pricing, so there is a direct transfer to revenue growth, and I infer that when component pricing comes back down there might be margin and FCF improvements.
More Personal Computing revenue -5% on lower PC market demand on rising component and device prices against a high prior-year comp for Windows, and weak revenue for XBox, offset by growth in Search.
Ebit decreased 15% including impairment charges in Xbox.
Xbox and gaming: aim to return to growth in 2027.
Windows: making it the best place to run secure edge AI. “We see significant opportunity for Windows to become the offload for unmetered intelligence, combining powerful on-device compute with enterprise-grade security.”
Thanks for reading - if you enjoyed reading this please like and restack, and do get in touch if you have questions.
Pete
Disclaimer: This post is for informational and educational purposes only. Building Arks is not licensed or regulated to provide any financial advisory service and nothing published by Building Arks should be taken as a recommendation to buy or sell securities, relied upon as financial advice, or treated as individual investment advice designed to meet your personal financial needs. You are advised to discuss your personal investment needs and options with qualified financial advisers. Building Arks uses information sources believed to be reliable, but does not guarantee the accuracy of the information in this post. The opinions expressed in this post are those of the publisher and are subject to change without notice. The publisher may or may not hold positions in the securities discussed in this post and may purchase or sell such positions without notice.