Nomura: Microsoft lowers full-year capital spending, but AI computing demand still exceeds supply

Why does Nomura still expect investment to grow after the capital expenditure cut?

Microsoft revised its 2026 calendar-year capital expenditure guidance from 175 billion dollars to approximately 170 billion dollars because of delays in third-party capacity delivery, with some GPU computing capacity originally scheduled for delivery shifted to the first half of 2027; management also said demand for Azure AI computing capacity still exceeds available supply.
In the first quarter of fiscal 2027, ending in September 2026, capital expenditure including finance leases was 45.2 billion dollars, up 30% year over year and 10% quarter over quarter; management expects it to exceed 50 billion dollars next quarter.
Microsoft expects capital expenditure to accelerate further in the second half of fiscal 2027, or the first half of calendar 2027, mainly from CPU and GPU spending and delayed third-party capacity deliveries. CPU spending includes higher memory and storage costs.

In the first quarter of fiscal 2027, Microsoft Cloud revenue was 63.9 billion dollars, up 30% year over year; Azure revenue grew 48%, driven by the AI platform Foundry and improved CPU monetization.
Under the segment reporting structure revised this quarter, Agents & Infra revenue was 76.8 billion dollars, up 25% year over year; Devices & Consumer revenue was 15.5 billion dollars, down 3% year over year.
Commercial bookings fell 12% year over year. Microsoft said large Azure commitments in the prior-year period raised the comparison base, while healthy growth in commercial sales this quarter partly offset the effect.

Foundry has more than 120 thousand customers, and revenue more than tripled year over year; Microsoft cites model choice, depth of capabilities and integration across its product stack as differentiators.
Microsoft combines per-seat subscriptions and usage-based billing for Copilot: customers using usage-based Copilot grew by more than 75% this quarter; the number of customers with more than 50 thousand Copilot seats rose to more than 8 times its year-earlier level.
On the infrastructure side, Maia 200 can process up to 1.7 times as many tokens per watt in Teams and Copilot AI applications; Microsoft expects Maia 300 to deliver 40% more on this measure than leading GPU systems. Management believes infrastructure efficiency gains help improve returns on capital expenditure.

Microsoft management expects free cash flow to remain positive in fiscal 2027. Considering this alongside continued capital expenditure growth, Nomura sees solid growth prospects for AI infrastructure stocks.
Nomura favors optical module supplier Zhongji Innolight and PCB and copper-clad laminate suppliers Shengyi Tech and WUS PCB, assigning buy ratings to all three. Microsoft's investment supports demand for related infrastructure; outcomes for individual companies still depend on demand for their products and competitive conditions.
Nomura's listed downside risks for Zhongji Innolight include weaker-than-expected demand for high-end optical modules, slower product upgrades and intensifying price competition.