Skip to content
News

Microsoft Reportedly Weighs Expansion of Data-Center Capacity Beyond 38 GW by 2032

DataNews Editorial Team

Image source: primary source

Microsoft is reportedly considering expanding its global portfolio of owned and long-term leased data-center capacity from about 12 GW today to more than 38 GW by 2032, as Azure capacity remains constrained by AI demand. The figure is not an official Microsoft target and would depend on project delivery, permitting and power availability.

Highlights
  • Microsoft is reportedly evaluating more than 38 GW of global data-center capacity by 2032, versus about 12 GW currently operated.
  • The reported capacity includes owned and long-term leased facilities, but excludes third-party rented compute such as CoreWeave capacity.
  • The 38-GW-plus figure is not an official Microsoft target and remains dependent on power, permitting and project execution.
  • Microsoft reportedly added about 1 GW of phase capacity in the fourth quarter of fiscal 2026.
  • The 20-year Constellation agreement is associated with the planned restart of the 835 MW Crane Clean Energy Center.

Microsoft is considering a long-range expansion of its global data-center capacity to more than 38 GW by 2032, according to S1’s account of Bloomberg reporting. The reported plan would increase the company’s portfolio from approximately 12 GW of currently operated capacity and would cover Microsoft-owned facilities as well as long-term leased data centers.

The more-than-38-GW figure is a scenario under consideration, not a publicly announced Microsoft target. It excludes compute capacity rented from third-party operators such as CoreWeave. Its delivery would depend on long-term development execution, regulatory approvals and access to electricity.

The expansion is being considered as demand for Azure cloud and AI services exceeds available capacity. About 2 GW of Microsoft’s current portfolio is equipment primarily using AI chips. S1 indicates that AI-oriented equipment could account for around one-third of total capacity in a scenario where the portfolio exceeds 38 GW, although this is an estimate rather than a committed deployment profile.

Microsoft has been accelerating infrastructure deployment. In fiscal 2026, the company reportedly operated 88 data centers globally and added 31 facilities across five continents in the fourth quarter, representing about 1 GW of phase capacity. Reported company-wide capital expenditure was approximately $145.3 billion for fiscal 2026, including $41 billion in the fourth quarter. These figures are corporate-level CAPEX and are not identified as data-center-only spending.

The company also reported approximately $329.1 billion in additional lease agreements that had not yet commenced as of June 30, 2026, predominantly related to data centers and scheduled to begin across fiscal years 2027 through 2033. That figure does not represent the cost of operating data centers or recognized balance-sheet lease liabilities.

Power procurement is integral to the buildout. Microsoft has signed a 20-year power purchase agreement with Constellation Energy associated with the planned restart of the 835 MW Three Mile Island Unit 1, known as the Crane Clean Energy Center. The available information does not specify further terms of the agreement or a restart timetable.

Grid and permitting constraints remain material. In Texas and New York, regulatory and interconnection conditions can extend delivery timelines for large data-center projects. Texas has about 474 GW of new grid-interconnection requests, with data centers estimated to account for about 90% of that total; these are connection requests, not confirmed projects under construction or operating capacity.

If the reported plan proceeds, Microsoft could become an especially significant buyer of powered land, grid interconnection, generation, transformers, cooling systems and construction capacity. The constraints described could also increase the strategic value of sites with secured power and shift hyperscaler competition further toward control of physical infrastructure and energy resources.

Why It Matters

If executed, the plan would represent one of the largest known hyperscale infrastructure expansions. It also underscores that AI data-center deployment is constrained not only by GPU supply, but by powered sites, grid interconnection, construction capacity, cooling infrastructure and permitting.

Sources

This article was prepared by the DataNews editorial team based on the sources listed above.