Microsoft Targets 38 GW Data Center Capacity by 2032
Cloud giant plans to more than triple operating infrastructure to power artificial intelligence workloads.
Microsoft is planning to more than triple its global data center capacity over the next five years, targeting 38 gigawatts of capacity by 2032. The aggressive expansion underscores the staggering energy and infrastructure commitments required to sustain the artificial intelligence boom.
Key details
Microsoft currently operates approximately 12 gigawatts of data center capacity worldwide. Reaching its 38-gigawatt target by 2032 will require the company to bring online an average of more than 1 gigawatt of new capacity every quarter. The expansion plan includes a mix of self-owned facilities and leased capacity from specialized neocloud providers, such as CoreWeave, Nscale, Lambda, Iren, and Nebius.
To fund this scale, Microsoft's capital expenditures have climbed steeply: total capex rose from $55.7 billion in 2024 to $115.9 billion in 2025 and $145.3 billion in 2026, with estimates projecting $175 billion for 2027. In the fourth quarter of FY2026 alone, Microsoft brought online 31 data centers, totaling 88 new facilities for the full fiscal year. Part of this growth includes its flagship "Fairwater" campuses—massive, AI-dedicated facilities housing hundreds of thousands of GPUs each.
Why this matters
An expansion to 38 gigawatts represents a monumental surge in power demand for a single technology corporation. For context, 38 gigawatts is equivalent to the peak electrical demand of a medium-sized nation or tens of millions of homes. Microsoft's trajectory highlights how AI compute requirements are reshaping utility planning, forcing energy providers to scramble for massive blocks of continuous power while accelerating grid transition risks.
Context
Microsoft's capacity surge reflects a broader hyperscale race where capital expenditure is increasingly dictated by hardware availability and power grid access rather than software adoption alone. As hyperscalers compete for gigawatt-scale power allocations, the scale of Microsoft's lease commitments—spending an estimated $60 billion on third-party AI cloud leases by late 2025—demonstrates how legacy tech giants are relying on external infrastructure partners to bypass internal construction backlogs.
Risks and open questions
The primary constraint on Microsoft's 38-gigawatt ambition is grid availability and rising public and regulatory resistance. State and local authorities across the United States and Europe are increasingly enacting data center moratoriums or imposing requirements for developers to self-fund grid upgrades and water infrastructure. Whether regional power grids can supply 26 additional gigawatts by 2032 without driving up consumer energy costs or jeopardizing corporate climate targets remains an open question.
What happens next
Microsoft will continue building out its Fairwater AI campuses while expanding its long-term power purchase agreements and microgrid investments. Industry observers will monitor upcoming utility rate cases and regional transmission planning filings to see how grid operators accommodate Microsoft's quarterly 1-gigawatt deployment pace.
Source: Data Center Dynamics Published on AI Usage Global, author: AUG Bot
