Microsoft Aims to Triple Data Centre Capacity by 2032
Microsoft is reportedly planning a massive expansion of its data centre infrastructure, targeting 38 gigawatts of capacity by 2032. This would more than triple its current footprint, driven by surging demand for AI workloads.
Microsoft is reportedly targeting a data centre capacity of 38 gigawatts (GW) by 2032, a figure that would more than triple its current infrastructure. According to reports from Bloomberg News, the company currently operates around 12GW of capacity. The expansion is largely driven by the growing demand for artificial intelligence, with AI-specific hardware expected to constitute about one-third of the new total footprint.
The company has not officially confirmed these specific targets, but internal reporting suggests a significant shift in how Microsoft allocates its resources. If the reported mix holds true, approximately 12.7GW of the future capacity would be dedicated to AI chips. This represents a sixfold increase in AI-specific power compared to today, far outpacing the overall growth rate of the data centre estate.
Demand Outpaces Current Supply
Microsoft’s cloud division, Azure, continues to face tight capacity constraints despite rapid growth. In the fiscal fourth quarter, the company added 1GW of new capacity and opened 31 data centres across five continents. Revenue for Azure and other cloud services rose 43% year-on-year, surpassing $100 billion in annual revenue for fiscal 2026. However, executives noted that customer demand still exceeds available capacity, indicating that supply remains the limiting factor.
To address this bottleneck, Microsoft has focused on efficiency and speed. The company reported that it reduced the time between receiving new graphics processing units and bringing them online by nearly 50% in its largest regions. CFO Amy Hood attributed this to improvements in fleet efficiency and faster deployment processes, allowing the company to monetize new capacity more quickly.
Specialized Infrastructure for AI Workloads
The expansion relies on highly specialized infrastructure designed to handle the intense power demands of modern AI. Microsoft’s Fairwater design supports hundreds of thousands of high-performance GPUs, operating at approximately 140 kilowatts per rack. This setup utilizes closed-loop liquid cooling systems to manage heat efficiently, a critical requirement for maintaining performance in dense computing environments.
Beyond off-the-shelf hardware, Microsoft is deploying its own custom silicon, such as the Maia 200 inference system. This system is designed to scale clusters to thousands of accelerators while integrating advanced cooling technology. The company is also building a dedicated wide-area network to connect these AI data centres across different geographic locations, ensuring seamless data flow for complex models.
Power Grids Face New Pressure
Scaling from 12GW to 38GW requires a substantial increase in energy procurement. Microsoft has contracted 40GW of new renewable energy capacity across 26 countries since 2020, with 19GW already operational as of early 2026. This massive investment in green energy is essential to power the new data centres, but it also highlights the strain on existing electrical grids.
The trade-off for this rapid expansion is a significant increase in capital expenditure and logistical complexity. A single large campus, such as the planned 2GW site in Pecos, Texas, represents only a fraction of the total additional capacity needed. The sheer scale of the build-out means that Microsoft must manage not just technology, but also physical space, power distribution, and cooling systems on an unprecedented scale.
