Why do analysts remain optimistic about demand after Microsoft cut spending?

  • Microsoft reduced its calendar 2026 capital expenditure guidance by 5 billion dollars to 170 billion dollars. Morgan Stanley attributed the reduction to delays in the delivery of third-party capacity, rather than a demand issue.
  • Microsoft also said capital expenditure would accelerate in the first half of calendar 2027, driven mainly by rising CPU-related memory and storage costs, GPUs and delayed delivery of third-party capacity. The increase in spending therefore includes equipment investment as well as the effects of some cost increases and changes in delivery timing.

What have other cloud giants said about spending in 2027?

  • Meta narrowed its calendar 2026 capital expenditure guidance from 130 billion–145 billion dollars to 135 billion–145 billion dollars, and said it would continue investing aggressively in 2027; Alphabet maintained its guidance of 195 billion–205 billion dollars and reiterated that spending would rise significantly year over year in 2027.
  • Amazon maintained its calendar 2026 capital expenditure guidance of 220 billion dollars and said spending in 2027 would be significantly higher than in 2026. On that basis, Morgan Stanley concluded that demand still exceeds supply and expects strong demand to continue into 2027.
  • However, the companies did not provide quantitative capital expenditure guidance for 2027 this quarter. Morgan Stanley will next focus on the impact of specific guidance on Greater China technology hardware companies’ results for the fourth quarter of calendar 2026.

How does cloud providers’ investment flow through to the supply chain?

  • Morgan Stanley maintains its overweight view on the Greater China supply chain. In server assembly, Wistron is an important L6 partner for Nvidia and AMD and is increasing its exposure to server and rack assembly through new customers.
  • Lenovo supplies general-purpose servers to Microsoft and Oracle and is also expanding its AI server business; as of the second quarter of calendar 2026, its AI server project pipeline was approximately 54 billion dollars. Wiwynn supplies general-purpose servers to Microsoft, Meta and Amazon, and is also an important supplier of Trainium server racks and the AMD Helios racks used by Meta.
  • Spending also flows through to components: general-purpose and AI servers require PCBs and copper-clad laminates, while computing and networking chips require ABF substrates. Morgan Stanley favors suppliers in these areas, as well as Lotes, which is expanding its CPU socket and AI component businesses.