Microsoft Corporation is reported to be planning an expansion of its global data center footprint to more than 38 gigawatts by 2032, more than tripling the roughly 12 gigawatts it runs today. The figure, attributed to Bloomberg News reporting that cited people familiar with the plans and relayed by Reuters on September 10, 2026, has not been confirmed in any public filing or statement by the company.
The number matters because Microsoft has been turning away some cloud and AI business for lack of capacity. Azure and other cloud services revenue rose 43 percent year over year in fiscal Q4, and Microsoft Cloud revenue reached $59.3 billion for the quarter and $214.4 billion for fiscal 2026. A backlog of committed demand gives the company a reason to build. Commercial remaining performance obligations climbed 84 percent year over year to $678 billion, with roughly 30 percent expected to be recognized as revenue within the next 12 months.
Right now Microsoft has a capacity problem, not a demand problem, and the reported plan is the company's answer to that problem. Whether a 38-gigawatt footprint is achievable by 2032 depends on power, land, and permits in dozens of jurisdictions, and on whether customers still want that much compute eight years out. It is not clear this will hold.
What the 38-Gigawatt Plan Actually Says
Only about 2 gigawatts of Microsoft's current capacity is dedicated to AI-specific chips. Under the reported roadmap, that share would grow to roughly one-third of the much larger 38-gigawatt total by 2032. That implies a shift in the mix of workloads, not just more of the same servers.
Microsoft has already shown it can move fast. The company added another gigawatt of capacity in fiscal Q4 and says it remains on track to roughly double overall capacity within two years. It has also signed a 20-year agreement with Constellation Energy tied to the restart of the 835-megawatt Three Mile Island Unit 1, which gives it another long-term electricity source as AI workloads drive sharply higher power requirements.
The company has also been expanding its government footprint. In early 2026 it planned to add three Availability Zones in the US Government Arizona region, a smaller but telling sign that even regulated customers are pushing for more capacity. Azure Storage is being retooled for agentic workloads and Kubernetes-based cloud-native applications, which will consume more of whatever compute gets built.
The Money Behind the Buildout
The plan carries a large price tag. Microsoft expects more than $50 billion of capital expenditures in fiscal Q1 2027 and about $175 billion during calendar 2026 under its broader capex definition, which includes finance leases. The company reported roughly $145 billion of capital expenditures in fiscal 2026, while cash additions to property and equipment totaled about $116 billion.
The bull case rests on committed demand. With $678 billion in remaining performance obligations, Microsoft has revenue visibility that few companies can match, and about 30 percent of that backlog converts to revenue within a year. If new capacity comes online on schedule, it can be filled by business the company already has under contract.
The bear case is simpler: the money has to be spent before the revenue arrives, and free cash flow takes the hit in the meantime. Multi-year construction timelines, changing customer demand, and improvements in computing efficiency could all reduce how much physical capacity Microsoft ultimately needs. The 2032 target should be read as a planning framework, not a promise.
Capital of this scale also crowds out other spending. Every dollar committed to a data center shell or a power purchase agreement is a dollar not spent on product development or acquisitions. Microsoft's management is betting that the marginal return on compute exceeds the marginal return on the alternatives, which is a defensible position today and a risky one if AI workloads consolidate onto fewer, larger clusters.
Power and Politics Are the Real Constraints
Power availability is the binding constraint, not silicon. Texas temporarily halted approvals for new data center grid connections while officials review the industry's electricity and water demands. New York imposed a moratorium on large new data centers amid concerns about electricity costs. Microsoft will need enormous amounts of power, land, and regulatory approval across multiple jurisdictions to more than triple its reported footprint.
The water question is getting louder too. Data centers of this scale draw millions of gallons for cooling, and communities near proposed sites are pushing back on both water and electricity. The Three Mile Island agreement covers 835 megawatts, which is meaningful but small against a 38-gigawatt target. Microsoft will need many more deals like it.
The company is not alone in this race. Alphabet announced its own major data center expansion in Finland the same week, and institutional sentiment shifted slightly: Microsoft's hedge fund count slipped to 273 in the second quarter from 282 in the first, even as position value rose to $66.51 billion from $63.58 billion. Alphabet's holders climbed to 275 from 265, with position value jumping to $93.74 billion from $72.58 billion.
What to Watch Next
Two data points will tell whether the reported plan is real: whether Microsoft confirms the 38-gigawatt target in a filing or earnings call, and whether its capacity growth keeps pace with the 43 percent Azure growth rate. The first gigawatt of AI-specific capacity already exists; the harder part is the next 10.
For more context on cloud and edge infrastructure, see Reuters. The reported roadmap is a bet that demand for AI compute will keep compounding for the rest of the decade. If it does, Microsoft's early spending looks prescient. If efficiency gains outpace demand, the company will be holding gigawatts of expensive, underused capacity in 2032.
Readers can track Microsoft's quarterly filings for updates on data center investment. The company reports fiscal Q1 2027 results in late October, and any change to the capex guidance will be the first real signal of whether the 38-gigawatt figure is still live.
The broader cloud market is moving the same direction, with edge computing pulling workloads closer to users even as hyperscalers centralize AI training. Microsoft's bet assumes both trends keep running in parallel. If edge deployments eat into centralized demand faster than expected, the 38-gigawatt number starts to look like a ceiling rather than a floor.