Microsoft predicts data center crunch will continue into 2026

Microsoft is restricting new Azure cloud subscriptions in key US regions due to ongoing data center shortages, with constraints expected to last through the first half of next year, according to sources familiar with the company’s internal forecasts.

Some of Microsoft’s largest server hubs, including Northern Virginia and Texas, are affected, leading the company to direct customers to other regions with available capacity.

The shortage impacts both GPU-powered machines for AI and traditional CPU-based cloud workloads.

Capacity issues have been a recurring topic for Microsoft, which has cited inability to meet all customer demand on recent earnings calls.

Microsoft reported Azure generated over US$75 billion in its 2025 fiscal year, outpacing rivals Amazon and Google in cloud growth.

The company continues to expand its data center footprint, adding over two gigawatts of capacity in the past year.

Microsoft said most US Azure regions still support existing workloads, and that it sometimes compensates customers for added costs or delays caused by capacity issues.

🔗 Source: Bloomberg

🧠 Food for thought

Implications, context, and why it matters.

Azure capacity limits reveal regional and SKU allocation

Azure shortages vary by place and size. A request can fail for a specific VM family or availability zone (physically separate data centers within an Azure region). In a region, certain SKUs (specific VM sizes/configurations) or zones can be tight 1 2.
Microsoft asks customers to try other sizes, zones, or regions when supply runs short 2. Latency matters because Microsoft targets about sub-2 milliseconds within a region across zones 4. One test saw sub-1 ms in the same zone and more than 5 ms across regions 5. Compliance and multi-region operations add more hurdles.

Cross-region deployment services rise as planning turns strategic

Consultancies and managed service providers can help teams move workloads across regions with plans that factor in availability and inter-region latency. Microsoft publishes region-to-region median (P50) round-trip time tables 6. High cross-region RTTs often push teams away from synchronous writes 4 5.
Orchestration and traffic tools can suggest regions based on live availability signals. They can weigh inter-zone latency, about 70 percent higher than same-zone in one test 5, plus customer proximity.
WAN optimization and replication vendors can speed cross-region transfers for multi-region rollouts. Resilio says its peer-to-peer architecture hit 100+ Gbps per server in tests 7.