Microsoft is learning a painful lesson about the cost of hesitation in the AI era. After pausing some data center development in spring 2025, the company now finds itself unable to keep up with surging demand for AI services — and customers are feeling the squeeze.
According to Bloomberg reporting from April 1, Microsoft has restricted new subscriptions for Azure cloud services in critical hubs including Northern Virginia and Texas. The bottleneck is not compute hardware but something more fundamental: physical space and power.
A Self-Inflicted Capacity Gap
The shortage traces back to Microsoft's decision to slow data center construction roughly a year ago, a move that now looks poorly timed given the explosive growth in AI workloads. Demand for AI services has raced past the company's ability to provide them, creating what analysts are calling a "capacity wall" that could persist through mid-2026.
The restrictions affect new enterprise customers trying to onboard Azure AI services, potentially pushing them toward competitors like AWS and Google Cloud at a critical moment in the enterprise AI adoption cycle.
Billions in Emergency Deals
To close the gap, Microsoft has embarked on an aggressive campaign of infrastructure partnerships. The company spent $11.1 billion on data center leases alone in Q1 2026 and has signed roughly $33 billion in compute capacity deals with so-called "neoclouds" — smaller, specialized cloud providers.
The largest of these is a $19.4 billion agreement with Nebius, the AI cloud spinoff from Yandex. Microsoft also signed a $14 billion deal with Nscale and additional agreements with CoreWeave and Lambda. These partnerships represent a significant departure from Microsoft's traditional strategy of building and owning its own infrastructure.
Planned Expansion
Microsoft expects to increase its AI capacity by more than 80 percent throughout fiscal year 2026 and aims to roughly double its overall data center footprint over the next two years. The company has committed $7.3 billion to its Fairwater campus in Wisconsin, which it says will become the "world's most powerful data center."
What It Means for the AI Industry
Microsoft's capacity crunch highlights a tension running through the entire AI sector: the gap between the pace of AI model development and the physical infrastructure needed to run those models at scale. While frontier models continue to advance rapidly, the power-hungry data centers required to serve them remain constrained by real-world factors like energy supply, permitting, and construction timelines.
For enterprise customers, the immediate impact is limited AI service availability in key regions. For the broader industry, it serves as a warning that the AI boom's biggest bottleneck may not be algorithmic — it may be concrete, copper, and kilowatts.
