In its upcoming fiscal Q2 earnings report, Microsoft executives have warned that demand for cloud-based AI services is outstripping available graphics processing units (GPUs), constraining Azure’s capacity to support large language model workloads. Internal data show that Azure GPU utilization climbed to 92% in December, up from 85% in September, leading to backlogs of more than 1,200 customer requests for AI inference and training. To resolve the bottleneck, Microsoft plans to deliver five new GPU-optimized data center regions by June, each equipped with 10,000+ NVIDIA H100 accelerators, and is accelerating deployment of its in-house Maia 200 inference chips to reclaim capacity and improve cost per token by 30%.

This month Microsoft unveiled Maia 200, its first custom AI inference accelerator, designed to reduce reliance on third-party GPUs and improve cost-efficiency in Azure. According to Microsoft benchmarks, Maia 200 delivers three times the FP4 performance of AWS’s Trainium and exceeds Google’s latest TPU in FP8 token throughput, while offering 30% better performance per dollar than the current Azure fleet average. The chip will power Copilot Enterprise and select OpenAI GPT deployments, representing an initial capacity of 200 petaflops across two U.S. regions, and is slated for roll-out in Europe and Asia in Q3 2026.

Mount Pleasant, Wisconsin, has granted Microsoft approval to develop 15 additional data centers on two parcels totalling nearly 9 million square feet. The project, valued at over $13 billion in taxable assets, will support Microsoft’s commitment to OpenAI and enterprise clients by adding an estimated 150 megawatts of AI-optimized power draw. Construction is expected to span ten years, sustaining approximately 2,500 unionized jobs annually during build-out. Local utilities have confirmed that no additional water infrastructure is required beyond the 8.4 million gallons per year already allocated, underscoring the site’s readiness for large-scale AI loads.

Analyst consensus compiled by Visible Alpha projects Microsoft will report Q2 revenue of approximately $80 billion, with Azure growth forecast at 38% year-over-year—down from 45% in Q1—as the company absorbs higher AI deployment expenses. Investors will be watching gross margin guidance, as AI infrastructure costs are expected to climb by 220 basis points, offset by price increases in Copilot and premium cloud services. Management has signaled capital expenditures may rise by 15% in the second half of fiscal 2026 to support AI-focused GPU and Maia 200 capacity, making margin commentary a key driver of investor sentiment.