Microsoft has set a goal to expand its data center capacity by more than threefold by the year 2032. This plan aims to solve one of the most significant barriers facing the corporation: the lack of computational power needed to meet the growing demand for cloud computing and artificial intelligence (AI) services.
According to sources informed to Bloomberg, Microsoft's global network of data centers is expected to exceed 38 gigawatts of capacity in 2032, compared to the current approximately 12 gigawatts. This projected capacity represents an electrical consumption higher than that recorded by the State of New York (USA) during its peak demand.
The planning includes both Microsoft-owned and leased facilities. However, it does not account for the contracted computational capacity of specialized cloud companies, such as CoreWeave. Analysts warn that these estimates are subject to change, given that building large data centers takes several years, and market changes or the emergence of new technologies can modify projections.
This expansion comes after a period of computational power shortage that became a primary obstacle for Microsoft. The situation was intensified by a decision made around early 2025, when Chief Financial Officer Amy Hood suspended part of data center development, fearing the company was investing in excessive infrastructure. This action led to regret among several executives.
Due to server insufficiency, Microsoft even imposed restrictions on new cloud service subscriptions in crucial regions in the United States and Europe. On some occasions, customers opted to migrate their businesses to competitors. A notable example is the Chinese retailer Temu, which signed an important agreement with Oracle the previous year after failing to obtain the desired additional capacity from Microsoft in certain areas.
The lack of capacity also affected Microsoft's salespeople, who were trying to increase sales of cloud and AI services, and caused anxiety among investors about when the vast investments in AI would begin to generate returns.
New complex in Atlanta signals strategic shift
A key indicator of this strategic change is a new set of data centers under construction in the Atlanta region (USA), named East US 3. Alistair Speirs, Microsoft's cloud infrastructure executive, stated that this complex will help mitigate the lack of capacity in Virginia (USA), which concentrates the company's largest data center hub. Speirs emphasized that 'essentially, this makes more capacity available everywhere when we open a new data center.'
The region is expected to receive approximately 300 megawatts of capacity this year alone, with a forecast increase to over one gigawatt in the coming years. Sources close to the project indicate that the development of this complex will cost tens of billions of dollars.
Despite the advancement of AI, East US 3 will be primarily focused on CPU-based computing, using servers from companies like Intel, in contrast to the AI chips typically supplied by Nvidia. Furthermore, Microsoft is reallocating part of the development of its own products and other internal operations to less saturated data centers, thus freeing up space in the most contested locations for customers.
Even more than three years after the start of the current AI growth, Microsoft's data centers remain predominantly geared towards general-purpose computing. Most facilities employ central processing units (CPUs) to support cloud applications and databases, rather than training AI models. Of the current 12 gigawatts of capacity, only about two gigawatts are dedicated to specific AI chips, with expectations that this share will grow to approximately one-third of the 38 gigawatts projected for 2032.
The evolution of AI systems itself contributes to the need to expand various types of computing. As Speirs stated, newer tools require increasing volumes of servers equipped with CPUs, in addition to GPUs, because 'GPUs alone do not constitute a large AI infrastructure.'
External reactions and business impact
Microsoft's expansion and that of other major technology companies occur in a context of growing resistance to the installation of data centers in the United States. Surveys indicate that most American citizens oppose the implementation of these structures in their localities. State governors, such as those in Texas and New York, have already determined interruptions in new data center projects.
Simultaneously, the four largest companies involved in the competition for computational capacity—Microsoft, Alphabet, Amazon, and Meta—have collectively committed to spending nearly $2.4 trillion (approximately R$ 12.6 trillion) in the coming years, focusing mainly on equipment and data center contracts.
The capacity shortage has also affected other areas of Microsoft's business. In August, GitHub was down for almost eight hours, largely due to data center capacity limitations. With the exponential use of the platform for code generation via AI, GitHub began depending on other data center providers, including Amazon Web Services (AWS).
In some instances, products used in the US had to be operated from international data center regions because the nearest facilities did not have sufficient capacity, which can result in software slowdowns. Kyle Daigle, GitHub's Chief Operating Officer, reported that the company is increasing its capacity and restructuring its systems to optimize efficiency.
Signs of pressure have also reached the gaming sector. Xbox, which previously offered unlimited cloud game streaming, notified subscribers last week that it would begin limiting playtime and charge users who exceed these new limits.
