Overview
- Bloomberg reported that Microsoft intends to grow its global data-center footprint from about 12 GW today to roughly 38 GW by 2032, but the company has not publicly confirmed that target.
- The share of capacity purpose-built for AI chips would rise from about 2 GW now to roughly one-third of the planned footprint under the reported plan, which would require many more high-density GPU racks.
- Microsoft has already signaled big spending changes with guidance that includes $50 billion in capex for the fiscal first quarter of 2027 and about $175 billion for calendar 2026 and it is spreading long-term data-center leases over 25 years to smooth reported costs.
- Concrete, confirmed moves that show how the company is scaling include the Pecos, Texas 2 GW campus with on-site natural gas power and closed-loop cooling, the Fairwater AI design for large GPU fleets and Microsoft’s 40 GW of contracted renewable energy.
- Analysts and Microsoft disclosures warn that power and grid upgrades often lag data-center builds, which could tighten local grids, lengthen project timelines and raise the need for on-site generation, PPAs and broader energy planning.