On January 3, 2025, Microsoft published a blog post announcing $80 billion committed to AI datacenter construction for fiscal year 2025 alone — with more than half of that capital targeted at US soil. For vendors selling UPS systems, generators, cooling infrastructure, cabling, and electrical services, understanding where these buildings are going up — and at what stage of construction — is the difference between getting on the bid list and finding out after the contracts are signed.
The Scale of What's Being Built
To put $80 billion in context: that's more than the entire US datacenter construction market was worth just five years ago. Microsoft is building at a pace that requires somewhere between 3 and 5 gigawatts of new power capacity annually. Each gigawatt of datacenter capacity requires roughly 10–15 MW of UPS infrastructure, thousands of battery strings, multiple generator installations, and an entire cooling plant.
Key Campus Projects in the US
Virginia — Boydton and Quincy: Microsoft's Virginia campus in Boydton (Mecklenburg County) remains one of its largest anywhere. Ongoing expansion phases have been permitted continuously since 2019, with the most recent variance filings in late 2025 indicating a 200MW+ addition. The Quincy, Washington campus is similarly expanding, with utility interconnection applications filed with Grant County PUD in Q3 2025.
Wisconsin — Mount Pleasant: The Racine County campus, first announced in 2017, has accelerated dramatically since the AI infrastructure push began. Building permits filed with Mount Pleasant village show multiple 100MW+ structures in various stages of construction and commissioning. This campus is expected to reach 2GW of total capacity by 2028.
Iowa — West Des Moines and Alluvion: Microsoft has had a presence in Iowa since 2012, but recent filings with MidAmerican Energy show interconnection requests for an additional 400MW+ across two sites, suggesting a major phase of new construction.
Texas — San Antonio: A new campus outside San Antonio, first permitted in early 2025, represents Microsoft's largest new footprint in Texas. The project involves multiple buildings totaling an estimated 300MW and is expected to begin equipment installation in mid-2026.
What This Means for Equipment Vendors
The timing of procurement at a hyperscale project is everything. Microsoft, like all hyperscalers, has preferred vendor programs — but those programs don't cover every component at every site. Gap fills, regional contractors, and tier-2 equipment suppliers are consistently used on large projects, especially during rapid expansion phases when primary vendors can't meet demand.
Permit filings are the earliest public signal. Microsoft consistently files electrical permits, mechanical permits (for cooling infrastructure), and structural permits 6–12 months before equipment installation begins. A vendor who identifies a new permit filing in Mount Pleasant in January is positioning themselves 6+ months before the procurement team is actively shopping.
The AI Density Problem
AI training clusters require dramatically more power density than traditional cloud infrastructure — often 20kW to 80kW per rack versus the 5–10kW standard. This forces facilities to upgrade their entire power train: new UPS with higher kVA ratings, battery systems with shorter recharge cycles, and cooling infrastructure capable of handling the waste heat from GPU clusters.
For battery vendors specifically, this is a replacement cycle happening at accelerated speed. Equipment that would normally last 8–10 years is being replaced in 4–5 years in AI-dense environments due to the increased cycling. Facilities managers know this and are actively re-evaluating their battery contracts.
How to Track Microsoft's Expansion in Real Time
The key data sources for tracking hyperscale construction are: county-level building permits (filed before any public announcement), utility interconnection applications (filed with the local power company, usually 12–18 months ahead), and job postings for site-specific facility roles (a surge of Critical Facilities Engineer posts in a specific zip code is a reliable indicator of an upcoming commissioning).
Kova Stack aggregates all three signal types and flags them for vendors by geography. If you're selling into the Virginia, Wisconsin, Iowa, or Texas markets, the signals are already there — you just need the system to surface them before your competitors do.
Sources & Further Reading
Microsoft Official Blog: "Microsoft to invest $80 billion to build out AI infrastructure" (January 3, 2025) — The original announcement detailing the $80B commitment and US-first focus.
Data Center Dynamics: Microsoft $80B Coverage — Industry analysis of the announcement and geographic breakdown.
Racine County Coverage of Mount Pleasant Campus — Local permit filings and construction timeline coverage for the Wisconsin campus.