Executive Briefing: Microsoft has $190 billion to spend on AI and still can’t get enough

Source: Nate’s Newsletter Date: 2026-05-24 URL: https://natesnewsletter.substack.com/p/ai-big-tech-industrial-business

Summary

Nate reframes Big Tech AI spending as industrial infrastructure, not software economics. Microsoft’s $190B 2026 capex (guidance exceeding $40B next quarter) plus four hyperscalers’ combined ~$700B (nearly double 2025) signals a structural shift: AI is a manufacturing process where every inference consumes physical capacity. Two-thirds of quarterly spend goes to short-lived assets (GPUs, CPUs). Microsoft expects to remain capacity-constrained through 2026 — a novel problem for cloud providers. The key question shifts from “which model is best” to “who can operate the factory producing intelligence at scale?”

Implications

• AI vendor contracts become supply contracts. Allocation, capacity reserves, and fallback provisions replace seat-based licensing. Token-based forecasting replaces headcount planning. Enterprise procurement teams need supply-chain expertise, not just IT evaluation.
• The $700B combined capex creates an infrastructure race with no clear ceiling. If Microsoft commits $190B, rivals must match or accept capability lag. This compounds the margin pressure already visible in OpenAI’s -122% Q1 operating margin.
• Feeds threads: token economics competition (capex as the new moat), enterprise deployment battleground (contract restructuring), Anthropic distribution machine ($303.8B+ compute commitments are part of this same race).