Home/Deal Room/Microsoft brings Wisconsin Fairwater AI superfactory facility online

Data centers · Evidence brief

Microsoft brings Wisconsin Fairwater AI superfactory facility online

Microsoft + QTS / Blackstone + NVIDIA

Primary sourcePress release only2026-06-23
Deal structureOperational AI superfactory / distributed hyperscale AI data-center network
GeographyMount Pleasant, Wisconsin + Fayetteville / Atlanta, Georgia
Infrastructure layersData centers + networking + cloud capacity
Disclosed scaleFirst Mount Pleasant, Wisconsin Fairwater datacenter facility is fully operational after April equipment startup; Microsoft estimates $4.7B of Wisconsin hyperscale construction spend from 2024-2028, with a second adjacent Mount Pleasant facility under construction for 2028, plus Atlanta / Fayetteville Fairwater site linked by dedicated AI WAN

Why it matters

The strategic read

Turns Fairwater from a future AI-superfactory architecture into an operating capacity milestone. Microsoft now has the first Wisconsin facility online while it continues the second Mount Pleasant build and links Fairwater campuses into a distributed training network for OpenAI, Microsoft AI, Copilot and other frontier workloads. The strategic signal is that AI data centers are moving from announced megasites into live, multi-site supercomputer infrastructure with local construction, workforce, power, cooling and networking supply chains attached.

Source-supported terms

What was disclosed

  1. 01

    Microsoft announced on June 23, 2026 that it completed construction of its first datacenter facility in Mount Pleasant, Wisconsin, after bringing equipment online and conducting startup activities in April.

  2. 02

    Microsoft said the first Mount Pleasant Fairwater facility is now fully operational and described Wisconsin as home to the world’s most powerful supercomputer.

  3. 03

    The facility was announced in May 2024 and completed ahead of schedule after nearly 10,000 construction workers contributed over two years.

  4. 04

    Microsoft said nearly 550 full-time employees are currently on-site supporting the first facility, with hiring expected to continue.

  5. 05

    Microsoft estimates it will have spent $4.7B between 2024 and 2028 locally on hyperscale construction in Wisconsin.

  6. 06

    Construction of Microsoft’s second adjacent Mount Pleasant facility is ongoing, with foundation installation, steel erection and underground utility placement underway and completion scheduled for 2028.

  7. 07

    Microsoft expects full-time employment in the Village of Mount Pleasant to grow to around 800 once the second facility is operational, with further expansion to the north representing hundreds more.

  8. 08

    Microsoft says its Fairwater AI datacenters are directly connected to each other and eventually to other U.S. sites through a dedicated network.

  9. 09

    The Atlanta Fairwater site is part of Microsoft’s first AI superfactory, linked with the first Fairwater site in Wisconsin.

  10. 10

    Microsoft describes the Fairwater architecture as purpose-built for hundreds of thousands of cutting-edge AI chips to work together on frontier workloads.

  11. 11

    The Wisconsin Fairwater site covers 315 acres and 1.2M square feet across three buildings, according to Microsoft’s September 2025 technical announcement.

  12. 12

    Local reporting places the Atlanta/Fayetteville Fairwater facility within a 615-acre QTS-developed campus.

Disclosure boundary

What is known—and what is not

Primary Microsoft technical, product and operational milestone announcements. Public materials disclose the Mount Pleasant first-facility operational status, April startup activities, local hyperscale construction spend estimate, worker and employee counts, second-facility 2028 schedule, dedicated Fairwater AI WAN and broad NVIDIA GPU architecture. Full QTS/Microsoft lease terms, power contracts, capex split by site, exact site-level IT load, GPU counts by facility, customer allocation, cooling-vendor details, network capacity, delivery remedies and workload-level economics are not public.