Microsoft

GPUs buy silicon; returns hinge on utilization. Collectives, HBM, schedulers, power/thermal, and inference batching decide effective output from hundred-billion AI CapEx. Hub NVIDIA; complements AVGO/ANET; constraints TSM/MU; physical layer VRT. See CPU industry.

Microsoft is among the few with hyperscale cloud, enterprise AI apps, and office-suite dominance; FY2025 Intelligent Cloud ~$106B (~+22%). Pre-earnings draft for the 2026-07-29 print—watch Azure growth and FY2027 Capex.