Microsoft Forecasts Show Data Center Crunch Persisting Into 2026
(Bloomberg) — Microsoft Corp.’s data center crisis will continue for longer than the company previously indicated, underscoring the software giant’s struggle to keep up with cloud demand.
Many of Microsoft’s U.S. data center regions are experiencing shortages of physical space or servers, according to sources familiar with the company’s internal forecasts. New subscriptions to Azure cloud services are restricted through the first half of next year at some key server farm hubs, including Northern Virginia and Texas, said the people who requested anonymity to discuss internal estimates.
That’s a longer time frame than the company had previously outlined. In July, Chief Financial Officer Amy Hood said current restrictions would continue until the end of 2025. He said the capacity shortage also affects data centers dominated by central processing units, which have long been the core chips of traditional cloud services, as well as machines running graphics processing units often used for artificial intelligence.
Azure is Microsoft’s most important growth engine; The cloud unit generated more than $75 billion in revenue in fiscal 2025. Its expansion comes at the expense of its biggest rivals, Amazon.com Inc. and surpassed Alphabet Inc.’s Google.
Lack of servers to rent to customers has become a recurring concern for cloud providers in recent years. During its last six-quarter earnings call, Microsoft said it was unable to meet all of its customers’ cloud demand. Amazon and Google announced similar restrictions.
A Microsoft spokesperson said the majority of Azure services and regions in the US “have available capacity so existing customers with distributed workloads can continue to grow.” In the event of some unplanned demand spikes, the company will implement “capacity preservation methods” to balance customer demand across its data center fleet, the spokesperson said.
Azure customers choose data center regions based on physical proximity and available software. When there isn’t enough space at a preferred facility, Microsoft salespeople refer customers to other customers with capacity based on internal referrals. But those workarounds can add complexity and increase the time it takes for data to travel between the server farm and the client, people familiar with the work said.
In some cases, customers experiencing Azure capacity issues are taking their business elsewhere, according to Apurva Kadakia, who helps companies build cloud workloads as global head of cloud and partnerships at Hexaware Technologies. Some touch multiple Azure regions or only send critical workloads to the cloud until more capacity becomes available.
“Our teams periodically work with large customers to plan for spikes in demand, such as holiday periods, and direct them to the most appropriate regions and products,” a Microsoft spokesperson said. “In unusual cases where customers face increased costs or delays, Microsoft will reimburse them for additional costs.”
Microsoft was on a historic construction spree to bring its data centers online; added more than two gigawatts of capacity last year; which was roughly equivalent to the power output of the Hoover Dam.
“Since the launch of ChatGPT and GPT-4, it has been nearly impossible to build capacity fast enough,” Chief Technology Officer Kevin Scott said in early October, referring to OpenAI’s popular chatbot and the AI model that runs it. “Even our most ambitious forecasts regularly fall short.”
Read more: Microsoft Bets $33 Billion on Neoclouds Like Nebius to Ease the AI Challenge
The intense computing demands of AI have driven much of the need for new data centers. But Microsoft also faces a shortage in demand for the traditional cloud infrastructure that underpins applications and websites on the internet.
For these CPU-based workloads, OpenAI is Microsoft’s largest customer, according to people familiar with the company’s operations.
Microsoft also uses significant computing resources to host its own workloads and applications such as the Office suite. Some Microsoft employees have been told to shut down internal projects in affected regions to save capacity, according to people familiar with the matter.
Bringing a data center online can take years, from initial planning to turning on servers. Many key components for data centers, from semiconductors to electrical infrastructure such as transformers, are subject to long delivery times, according to people familiar with the process.
People familiar said exceptions could be made for major customers who want additional capacity due to supply tightness in Azure regions. Availability is much better outside the US. For example, most of Microsoft’s European regions can handle new customer subscriptions without restrictions.
During the July earnings call, finance chief Hood said the persistent supply shortage was due to rising demand. “Oh my gosh, I talked about this in January and said I thought we would be in a better situation supply-demand wise by June,” he said during the July call. “And now I’m saying this: I hope I’ll be in better shape by December.”
–With help from Matt Day and Dina Bass.
More stories like this available Bloomberg.com
