Microsoft Reportedly Plans to Expand Data-Center Footprint Beyond 38 GW by 2032
Редакция ЦОД:Главное
Microsoft is reportedly targeting an expansion of its global data-center footprint from about 12 GW today to more than 38 GW by 2032, spanning facilities it owns and leases. The plan is intended to relieve capacity constraints for cloud and AI workloads, while excluding compute capacity rented from neocloud providers such as CoreWeave.
- Microsoft is reportedly targeting more than 38 GW of owned and leased data-center capacity by 2032, versus about 12 GW currently.
- The reported capacity target excludes compute rented from neocloud providers, including CoreWeave.
- AI-chip capacity is expected to rise from about 2 GW today to roughly one-third of the planned 2032 footprint.
- East US 3 near Atlanta is expected to deploy about 300 MW of equipment in 2026 and is planned to exceed 1 GW over several years.
- Capacity constraints in major US and European hubs are reportedly limiting some new cloud contracts.
Microsoft is reportedly planning to more than triple its global data-center footprint to more than 38 GW by 2032, from approximately 12 GW currently, according to a Bloomberg report relayed by BigGo Finance. The reported target covers data centers owned or leased by Microsoft, but excludes compute capacity the company rents from neocloud providers, including CoreWeave.
The expansion is aimed at addressing shortages of compute capacity for cloud and AI workloads. Microsoft is reportedly restricting new cloud contracts in several major data-center hubs in the United States and Europe because of capacity constraints, indicating that physical infrastructure availability is already affecting commercial deployment.
AI-specific capacity is expected to account for a materially larger share of the estate. About 2 GW of Microsoft’s current 12 GW footprint is allocated to AI chips, while AI workloads are expected to represent roughly one-third of the planned 38 GW-plus footprint by 2032.
East US 3 near Atlanta is cited as an example of the build-out. Around 300 MW of equipment is expected to come online there in 2026, with the campus planned to expand beyond 1 GW over the following several years. The 300 MW figure refers to an expected phase deployment, while the more than 1 GW figure is the planned capacity for the broader complex.
Microsoft has also provided broad CAPEX indicators: $145 billion in the latest fiscal year, $50 billion for the first quarter of fiscal 2027, and $175 billion for calendar 2026. These are corporate-wide capital-expenditure figures; the available information does not specify what portion is dedicated solely to data centers.
If delivered, the roadmap could increase demand for data-center sites, grid infrastructure, generation, facility equipment and AI computing systems. It also indicates that shortages of deliverable capacity can constrain cloud sales and potentially shift customers between providers. However, the capacity target remains a long-term plan, and its delivery may change with data-center development timelines, customer demand, technology shifts and local opposition to new construction.
The reported roadmap illustrates the scale at which hyperscalers are attempting to add physical infrastructure for AI and cloud demand. It also highlights that the binding constraints extend beyond AI-chip availability to data-center delivery, power availability, grid infrastructure, construction lead times and local acceptance of new sites.
- — / 11 сентября 2026 г. · ЯпонскийПервоисточник →
Материал подготовлен редакцией ЦОД: Главное на основе указанных первоисточников.