Microsoft plans to more than triple its data centre capacity to about 38 gigawatts (GW) by 2032, Bloomberg News reported, citing people familiar with the plans.

The company currently has about 12GW of capacity, according to Bloomberg’s reporting. Around 2GW is centred on AI-specific chips, while that share is expected to reach about one-third of the planned 38GW footprint by 2032. Microsoft has not publicly confirmed the reported 38GW target.

If that reported capacity mix is reached, around 12.7GW of Microsoft’s 38GW footprint would be centred on AI-specific chips, based on Bloomberg’s reported one-third share. That would be more than six times the approximately 2GW of AI-specific capacity reported today, compared with a little over a threefold increase in Microsoft’s overall data centre capacity.

Azure capacity stays tight

Microsoft added 1GW of data centre capacity during its fiscal fourth quarter ended June 30 and said it remained on track to roughly double its overall capacity within two years.

The company added 31 data centres across five continents during the quarter, taking the number added during fiscal 2026 to 88. It also reduced the time between receiving new GPUs and bringing them into service in its largest regions by nearly 50% during the financial year.

Azure and other cloud services revenue increased 43% year-on-year in the fiscal fourth quarter. Microsoft also said customer demand continued to exceed available capacity.

Azure surpassed $100 billion in annual revenue during fiscal 2026, representing growth of 41%. Microsoft Cloud revenue, which includes Azure and other commercial cloud products, surpassed $214 billion for the year.

Microsoft CFO Amy Hood said efficiency gains across the company’s CPU and GPU fleet, along with process improvements that brought new capacity online earlier, contributed to Azure’s fourth-quarter performance. Hood said the additional Azure capacity delivered during the quarter was quickly monetised.

AI raises infrastructure demands

Microsoft’s Fairwater AI data centre design uses a network built to connect hundreds of thousands of Nvidia GB200 and GB300 GPUs across its Azure AI infrastructure.

Microsoft said Fairwater can operate at about 140 kilowatts per rack and 1.36 megawatts per row. The facility uses a closed-loop liquid cooling system that continuously reuses coolant after its initial fill.

Fairwater’s backend uses a two-tier Ethernet network capable of 800Gbps GPU-to-GPU connectivity. Microsoft has also built a dedicated AI wide-area network to connect AI data centres across different locations.

Microsoft is also deploying its own AI silicon alongside accelerators supplied by companies including Nvidia and AMD. Its Maia 200 inference system uses a two-tier Ethernet architecture designed to scale clusters to as many as 6,144 accelerators and incorporates Microsoft’s second-generation closed-loop liquid cooling hardware.

Based on Bloomberg’s reported figures, moving from 12GW to 38GW would add about 26GW to Microsoft’s current data centre footprint. Microsoft is already developing a 2GW data centre campus in Pecos, Texas.

Microsoft said in June that the Pecos development would add approximately 2GW to its global data centre capacity, describing it as one of the largest single capacity additions in its history. The multibillion-dollar campus is expected to be developed over five to seven years.

The planned Pecos capacity is equivalent to less than one-tenth of the roughly 26GW of additional capacity implied by Bloomberg’s reported 2032 target. Data centre developments vary in size and design, so the comparison does not translate directly into a specific number of additional campuses.

Power and spending scale up

Microsoft said in February that it had contracted 40GW of new renewable energy capacity across 26 countries since 2020 through more than 400 agreements with utilities and developers.

Of that contracted renewable capacity, 19GW was operational as of February, with the remaining projects scheduled to come online over the following five years. Microsoft also said it had reached its target of matching its annual global electricity consumption with renewable energy in 2025.

The reported 38GW data centre figure and Microsoft’s 40GW renewable energy portfolio are not directly comparable. The former refers to reported data centre power capacity, while the latter refers to the rated generating capacity of renewable energy projects contracted across multiple grids and countries.

The Pecos development will also include dedicated power infrastructure. Microsoft said the 2GW campus will initially use a co-located natural gas power facility operating behind the meter, supplying electricity directly to the site rather than drawing that power from the public grid.

Microsoft expects the power facility and data centre eventually to connect to the wider regional grid. The campus will also use closed-loop cooling, which requires an initial water fill but is designed to avoid additional water consumption during normal cooling operations.

The International Energy Agency estimates that data centres consumed about 415 terawatt-hours of electricity globally in 2024 and projects consumption to reach around 945TWh by 2030 under its base case.

The IEA expects electricity use from accelerated servers, primarily associated with AI, to grow by around 30% annually through 2030 and account for almost half of the net increase in global data centre electricity consumption. Conventional server electricity use is projected to grow at about 9% annually.

The IEA said data centres and the energy infrastructure supporting them also operate on different development timelines. A data centre can become operational in around two to three years, while new power generation and grid infrastructure generally require longer planning and construction periods.

Microsoft also has a long-term power purchase agreement with Constellation Energy supporting the restart of the 835MW Crane Clean Energy Center in Pennsylvania. The nuclear facility was previously known as Three Mile Island Unit 1.

Infrastructure spending remains a large part of Microsoft’s investment programme. Capital expenditure reached $41 billion in its fiscal fourth quarter.

About two-thirds of that spending went to shorter-lived assets, primarily CPUs and GPUs. Microsoft also recorded $5.6 billion in finance leases during the quarter, primarily for large data centre sites.

Microsoft expects capital expenditure of about $175 billion for calendar 2026 after an accounting change that will move more future data centre leases from finance leases to operating leases. The company said the change affects the accounting treatment of those leases rather than its underlying calendar-year investment plans.

(Photo by NASA)

See also: AWS and Microsoft launch multicloud link between AWS and Azure

Want to learn more about Cloud Computing from industry leaders? Check out Cyber Security & Cloud Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.

Cloud Computing News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.