Microsoft's $7bn Wisconsin AI Megasite: Liquid Cooling, Clean Energy, Local Jobs

Microsoft's AI Megasite: Redefining Data Centres

Microsoft is building a next-generation AI campus in Mount Pleasant, Wisconsin. The first facility is slated to go live in early 2026, backed by US$3.3bn already committed and a further US$4bn planned for a second site.

For real estate and construction teams, this project shows how hyperscale AI demand reshapes site selection, utility agreements, cooling design, workforce pipelines and community benefits. It also sets expectations for schedule certainty, grid coordination and long-lead procurement.

Project snapshot

• Location: Mount Pleasant, Wisconsin
• Investment: US$3.3bn (phase one) + US$4bn (phase two); total >US$7bn
• Go-live: Early 2026 for the first facility
• Scale: Hundreds of thousands of NVIDIA GPUs in clustered configurations; enough fibre cabling to encircle the Earth four times
• Compute: Tenfold performance over today's top supercomputers, according to Microsoft
• Cooling: >90% closed-loop liquid cooling; remaining areas use outside air and switch to water only during extreme heat
• Water: Annual use projected to be minimal-comparable to a single restaurant, or an 18-hole golf course over a summer week
• Energy: Renewable matching commitment; a 250MW solar project in Portage County is underway; Microsoft is prepaying for energy and electrical infrastructure to stabilise local pricing
• Jobs: >3,000 at peak construction; ~500 full-time roles at the first site, growing to ~800 with the second
• Community: Ecological restoration funding via the Root-Pike Watershed Initiative Network; broadband expansion for rural residents and Sturtevant businesses
• Workforce: Datacenter Academy with Gateway Technical College (1,000 students targeted within five years); 114,000 people trained in AI skills across Wisconsin, including 1,400 in Racine County; a manufacturing-focused AI Co-Innovation Lab at UW-Milwaukee

What this means for developers, owners and GC/MEP teams

• Secure electric capacity early. Prepayment models and bespoke tariffs can de-risk community concerns about rate impacts and help lock timelines.
• Design for liquid cooling at scale. Plan for integrated cold-plate loops, heat-exchanger rooms, dry coolers, and leak detection. Structural provisions must support high-density racks and heavier piping.
• Engineer for thermal flexibility. Use outside-air economisation where climate allows; hold contingency for extreme heat events without major water draw.
• Treat water as a strategic asset. Closed-loop systems cut consumption, but water treatment quality, redundancy and monitoring remain critical.
• Overbuild fibre pathways. High-core, diverse routes with strict separation reduce risk for AI training clusters.
• Phase the campus. Separate compute, storage and network buildings improve operations, resilience and future expansion.
• Anticipate long-lead items. Switchgear, transformers, heat-rejection equipment and GPUs require early procurement and logistics planning.
• Bake in community benefits. Workforce academies, ecological restoration and broadband upgrades create local support and smoother approvals.

Cooling and water: practical notes

Closed-loop liquid cooling is the default across most of the campus. It's filled once during construction and recirculates, trimming waste and enabling higher rack density.

Remaining areas rely on outside-air cooling and only switch to water in extreme heat. For Wisconsin's climate, account for freeze protection, fluid chemistry management and service clearances around modular heat-exchange skids.

Energy strategy and grid coordination

Microsoft states it will match the facility's consumption with carbon-free energy on the grid and is advancing a 250MW solar project in Portage County to support that goal. The company is also prepaying for energy and electrical infrastructure under transparent tariffs with WE Energies to stabilise local pricing and maintain reliability.

For project teams, this points to early engagement with utilities, tariff customisation and clear cost-recovery models that protect residents while unlocking capacity for AI loads.

Workforce and community impact

At peak, more than 3,000 tradespeople have been on site. Operations will create ~500 full-time jobs initially, scaling to ~800 with the second facility.

Through Gateway Technical College's Datacenter Academy, Microsoft expects to prepare 1,000 students in five years. Statewide, 114,000 people have received AI skills training, with 1,400 in Racine County. The University of Wisconsin-Milwaukee Co-Innovation Lab focuses on AI for manufacturers such as Regal Rexnord and Wiscon Products.

Voices from Microsoft

"What sets Microsoft's infrastructure apart is the relentless pursuit of innovation and sustainability," says Noelle Walsh, President for Cloud Operations + Innovation at Microsoft. "Our data centres use advanced closed-loop liquid cooling systems - integrated pipes circulate cold liquid directly into servers, extracting heat efficiently and ensuring zero water waste."

Microsoft's Vice Chair and President, Brad Smith, adds: "Mount Pleasant isn't just becoming a hub for AI - it's becoming a blueprint for how innovation can serve everyone. We're not just investing in an AI data centre; we are investing in a community."

Key dates and next steps

• Now: Construction approaching completion; hiring underway for operations teams
• Early 2026: First facility scheduled to go live
• Next phase: Second Wisconsin site planned, bringing total full-time roles to ~800

Useful resources

• U.S. DOE: Data Center Efficiency
• ASHRAE TC 9.9: Thermal Guidelines for Data Processing Environments
• Complete AI Training: Courses by Job