Microsoft is rapidly expanding its data centers, more than tripling their capacity to 38 GW, to meet surging demand for AI and cloud services.

Microsoft plans to increase its global data center capacity from approximately 12 GW to more than 38 GW by 2575, while its latest annual investment budget reached $1.45 billion.
On September 11, 2569, at 04.00:3 AM, Bloomberg News reported that Microsoft plans to more than triple the computing power of its global data center network to address computing capacity limitations that have forced the company to decline some of its artificial intelligence (AI) and cloud businesses, amid rapidly increasing demand for these services.
Sources familiar with the plan revealed that Microsoft aims for its global data center network to have more than 38 gigawatts (GW) of power by 2575, up from approximately 12 GW currently. This level of power would exceed the peak electricity consumption of New York State.
These figures reflect the size of the computing power Microsoft expects to gain from its massive investment in data centers. The company's capital expenditures in its most recent fiscal year reached $1.45 billion, and analysts expect spending to continue increasing in the coming years.
The plan covers both Microsoft-owned data centers and leased locations, but excludes computing power leased from newer cloud providers, or “Neocloud,” such as CoreWeave. However, sources warn that the plan is still subject to change, as developing data centers takes years, and customer behavior and new technologies can quickly alter demand assumptions.
Meanwhile, Microsoft and other major technology companies are facing increasing opposition to data center construction. Surveys show that most Americans don't want data centers in their areas, and several state officials, including the governors of Texas and New York, have ordered delays to some new data center projects. Microsoft declined to comment on details of its construction plans.
The shortage of computing power has been one of Microsoft's biggest obstacles in recent years, and it has worsened after the company decided to delay the development of some data center projects until early 2568. The server shortage has caused some customers to switch to competitors' services and has put pressure on Microsoft's sales team, which needs to expand sales of cloud and AI services, as well as on investors who want to see a return on the billions of dollars the company has invested in AI.
Microsoft isn't alone in this problem; Alphabet's Google is also struggling to keep up with demand in terms of computing power. Meanwhile, the four major players in the computing infrastructure competition—Microsoft, Google, Amazon, and Meta—have pledged a combined investment of nearly $2.4 trillion over the next few years, largely focused on equipment and data center leasing.
However, many Microsoft executives are beginning to have hope that the company is overcoming its computing power limitations. Microsoft CEO Satya Nadella stated in July that the company is bringing new computing power to the system faster than ever before and is finding ways to use existing hardware more efficiently and profitably.
One of the key projects. คือ A new data center complex is under construction in the Atlanta area, codenamed “East US 3”. This will help alleviate the computing power shortage in Virginia, home to Microsoft's largest data centers.
The project will bring approximately 300 megawatts of power for data centers into the system this year, before increasing to more than 1 GW in the coming years. One source indicated that the entire development of the area is worth tens of billions of dollars. The project will focus on general-purpose computing using server-based CPUs from companies like Intel, rather than the AI chips typically supplied by Nvidia.
Even though the AI trend has been around for more than three years, Microsoft's data centers still largely focus on general-purpose computing, using CPUs for databases and cloud applications rather than training AI models. Currently, only about 2 GW of Microsoft's total 12 GW of power is used specifically for AI chips, but this proportion is expected to increase to about one-third of the 38 GW of power the company aims to bring to the system by 2032.
Alistair Speirs, Microsoft's head of cloud infrastructure, stated that new AI tools require increased server CPUs alongside GPUs, necessitating companies to increase all types of computing power. He pointed out that GPUs alone cannot build an efficient AI infrastructure.
Microsoft is still struggling to mitigate the impact of its decision to delay data center construction nearly two years ago, initiated by Chief Financial Officer (CFO) Amy Hood due to concerns that the company might build more capacity than it needs. However, many senior executives have since acknowledged that this decision contributed to the computing capacity bottleneck the company still faces today.
Hood said on Wednesday that Microsoft could get data centers up and running faster than in the past, with the company seeking to reduce the time from bringing infrastructure to the point where the hardware is ready to generate revenue.
Due to computing power shortages, Microsoft has had to limit new cloud subscriptions in several key data centers in the US and Europe, and in some cases, customers have opted to sign contracts with competitors instead.
For example, Temu, a Chinese online retailer and major cloud customer, secured a large contract with Oracle's cloud services last year after being unable to obtain additional computing power from Microsoft in the regions it needed.
Besides impacting sales, the server shortage also affects system stability. In August, Microsoft's software development platform, GitHub, experienced a nearly eight-hour system outage, largely due to limitations in its data center capacity. GitHub relies on other data center providers, including Amazon Web Services, to support the rapidly increasing usage driven by the AI trend, which has led to more people using coding tools.
At times, GitHub has had to route its products serving US customers through data centers overseas because nearby centers lacked sufficient capacity, which could slow down software deployment. GitHub is currently increasing its processing power and restructuring its systems to be more efficient in order to support the exponential growth in usage.
Signs of computing power limitations are also appearing in Microsoft's gaming business, with Xbox, which previously offered unlimited cloud gaming, informing subscribers last week that it will begin limiting playtime and charging additional fees for those who exceed the new limits.
The key challenge for Microsoft, therefore , lies in whether the company can build and deploy data centers quickly enough to support the accelerating demand for cloud and AI services. Speirs stated that Microsoft needs to plan its infrastructure well in advance and strive to scale responsibly to keep pace with the increasing demand.
Reference: www.bloomberg.com






























