Summary

Microsoft's Q3 earnings revealed a critical capacity constraint in the AI infrastructure supply chain, with the company planning to spend $190 billion on capex but still unable to meet demand. The core issue lies not in GPU availability, but in the complex manufacturing of chips with sufficient memory to support modern AI workloads. This signals a fundamental shift in how AI vendor contracts will be structured, moving from traditional software agreements to more complex supply-based allocations that prioritize capacity and technical constraints.