Microsoft admits AI "fever" will overload cloud infrastructure by 2026

Microsoft's data center outages will continue longer than the company previously announced, with many regions in the US facing a shortage of space or servers.

New sign-ups for Azure cloud services will be restricted to a few key server centers until the first half of next year, according to people familiar with the matter.

This is a longer timeframe than the company had previously forecast. In July, CFO Amy Hood said the current restrictions would be extended through the end of 2025.

The capacity shortage affects both machines running graphics processing units (GPUs) commonly used for artificial intelligence (AI), as well as centers that primarily use central processing units (CPUs) that are the "backbone" for traditional cloud services.

Azure is Microsoft’s most important growth engine. The cloud unit generated more than $75 billion in revenue in fiscal 2025, a growth rate that has outpaced even its biggest rivals, Amazon.com Inc. and Alphabet Inc.’s Google.

But in its last six earnings reports, Microsoft has repeatedly admitted it can't meet all of its customers' demand for cloud services. Amazon and Google have also described similar difficulties.

A Microsoft spokesperson said the majority of Azure services in the US have enough capacity to ensure existing customers can continue to develop the workloads they have already deployed.

In the event of a surge in demand, the company will apply “capacity conservation methods” to balance customer demand across the entire system.

Azure customers choose data centers based on geographic location and software availability. When their preferred facility reports capacity is full, Microsoft staff directs customers to other options.

However, sources familiar with the matter said the solution could increase the complexity and delay of data transmission, and customers who run into problems could then switch to a rival service.

To address the problem, Microsoft has been on a data center buildout campaign, adding more than 2 gigawatts of capacity over the past year.

However, Chief Technology Officer Kevin Scott admitted in early October that Microsoft has been unable to build capacity nearly fast enough since the launch of the ChatGPT chatbot and the GPT-4 large language model.

Even their most ambitious forecasts have often proven insufficient, he said.

AI's intense computing demands are a major factor, but Microsoft also faces a demand crunch for its traditional cloud infrastructure, which underpins applications and websites across the internet.

Additionally, Microsoft also uses a large amount of resources to run its own applications such as the Office suite.

Some Microsoft employees have even been asked to pause internal projects in affected areas to save capacity, sources familiar with the matter said.

The process of bringing a data center to life can take years, from planning to turning on servers. Many key components, from semiconductors to electrical infrastructure like transformers, have long lead times.

The situation outside the US is looking better – many of Microsoft’s European regions can handle new registrations without restrictions. But the domestic difficulties that Microsoft’s own success in pushing the AI race are creating are creating its biggest challenge.

This “big guy” has to admit the reality that they are not keeping up with the pace of innovation that they help create./.

(TTXVN/Vietnam+)

Source: https://www.vietnamplus.vn/microsoft-thua-nhan-con-sot-ai-gay-qua-tai-ha-tang-dam-may-den-nam-2026-post1069576.vnp