Microsoft's first purpose-built AI datacenter in Mount Pleasant, Wisconsin, entered full operation on June 23, 2026, after months of startup work that began in April. The Fairwater facility—part of a $3.3 billion campus—marks a major expansion of Azure's infrastructure in the US North Central region, specifically engineered to handle the immense compute demands of large-scale AI training and inference.
The Datacenter Is Real—and Running
The June 23 announcement confirmed that construction was complete and all systems were live. The datacenter occupies a portion of the 1,030-acre site Microsoft acquired in 2023, taking over land originally slated for a Foxconn manufacturing plant that never fully materialized. While Microsoft hasn't disclosed the exact megawatt capacity, the campus's scale suggests hundreds of thousands of servers, with power sourced from renewable energy agreements that the company has been signing across the Midwest.
This is not a conventional datacenter retrofitted for AI. Microsoft designed Fairwater from the ground up for the extreme power density and cooling requirements of GPU clusters running models like those behind Copilot, Azure OpenAI Service, and the thousands of custom AI applications enterprises are building. Liquid cooling, high-speed InfiniBand networking, and direct-connect options for Azure ExpressRoute are almost certainly part of the build—though the company rarely details the internals for security reasons.
For anyone tracking Microsoft's timeline, the Fairwater project moved from first shovel to production in under three years. That speed reflects the urgency of the AI boom: demand for cloud AI compute has outstripped supply for most of 2024 and 2025, driving long wait times for GPU instances. This datacenter is a pressure release valve.
What Fairwater Means for Different Microsoft Users
For Everyday Windows Users
If you use Windows 11's Copilot, Microsoft 365's AI features, or even Bing Chat, some of that inference work may now route through Fairwater, especially if you're in the central United States. You won't see a sudden performance jump—Microsoft's global edge network distributes loads intelligently—but the added capacity should help reduce the occasional "AI is at capacity" errors that have popped up during peak usage. The real benefit is behind the scenes: more headroom for Microsoft to roll out AI features to all users without throttling.
For IT Professionals and Cloud Architects
Fairwater extends the US North Central Azure region, which already includes a handful of datacenters in Illinois and Iowa. The new facility likely adds availability zones or at least increases the capacity within existing zones, which matters for anyone running virtual machines, Azure Kubernetes Service clusters, or AI batch jobs. If you're designing high-availability architectures, check your Azure region list: you may now see new zone options for North Central US, or at minimum, you'll find better VM core quotas available as the datacenter ramps up.
Latency to the Midwest from Chicago, Minneapolis, and even Toronto should improve for workloads placed in the new facility. If you've been holding off on migrating AI pipelines to Azure because of capacity constraints, this is your signal to test again.
For Developers and AI Startups
The most immediate impact is on GPU-intensive VMs. Microsoft has been steadily expanding its NC-series and ND-series instances, which use NVIDIA H100 and AMD MI300X accelerators. Fairwater almost certainly houses a large fleet of these, plus Microsoft's own custom Maia 100 AI accelerators that were announced in late 2023. Developers using Azure Machine Learning or the AI Toolkit for Visual Studio Code can manually target the North Central region when creating compute targets. Don't expect a separate "Fairwater" region—it's integrated into the existing fabric, but you may notice faster provisioning times and fewer "insufficient capacity" errors.
Startups in the Midwest that have suffered from high latency to Virginia or West Coast Azure regions can now deploy closer to home. This is especially relevant for real-time AI applications like video analytics, recommendation engines, and geospatial AI, where every millisecond counts.
The Road to June 23, 2026
How did a small Wisconsin village become a lynchpin in Microsoft's AI infrastructure?
2022-2023: The Foxconn Fallout. Foxconn's much-hyped LCD factory, promised in 2017, had largely fizzled, leaving Mount Pleasant with a sprawling but empty industrial site. Microsoft, already scouting for Midwestern land with abundant water and power, saw an opportunity. In March 2023, the company announced a $3.3 billion investment to build two datacenter campuses on the site, with the first phase—Fairwater—breaking ground later that year.
2024: Building Through Controversy. Construction wasn't without friction. Environmental groups raised concerns about water usage for cooling and the strain on the local power grid. Microsoft responded with pledges to use closed-loop cooling systems and to match the datacenter's electricity consumption with 100% renewable energy by 2025. It also funded local infrastructure upgrades, including a new substation.
April 2026: Startup Begins. Microsoft quietly began the "commissioning" phase in April, a period where servers are installed, networks are tested, and the facility is gradually integrated into Azure's global backbone. This phase typically takes 6-12 weeks, so the June 23 go-live date aligns with an accelerated but thorough rollout.
Why the AI Datacenter Label? Microsoft operates hundreds of datacenters, but it publicly calls only a handful "AI datacenters." The distinction is more about design philosophy than marketing. These facilities use advanced power distribution that can handle the 700W+ per chip that modern accelerators draw, compared to the 200-300W of typical server CPUs. They also employ direct-to-chip liquid cooling in many cases, which is far more efficient for the sustained, high-utilization workloads of training runs that last weeks.
Now What? Action Steps for Teams
1. Re-evaluate Your Azure Region Strategy
Log into the Azure portal and open the "Resource providers" or "Quotas" blade. Search for "North Central US." If you see new VM SKU quotas appear—especially for NCv4 or NDv6 series—request increases. Even if you don't have an immediate need, having approved quota speeds up future deployments.
2. Run Latency Tests
Use Azure's Network latency test (via az network latency) or tools like psping to measure round-trip times from your offices or existing infrastructure to the North Central region. The new datacenter may not yet have public test endpoints, but any improvement in the region's overall latency profile could change your multi-region design.
3. Check for New Local Edge Zones
With a major core datacenter online, Microsoft often lights up Edge Zones in nearby cities—Chicago, Milwaukee, Madison—to bring ultra-low latency to 5G-connected applications. Developers building with Azure Edge may see new options appear in 2026.
4. Monitor Pricing Changes
New datacenter capacity can gradually lower spot VM prices in a region due to increased supply. If you run interruptible AI training jobs, compare spot pricing for GPU VMs in North Central versus East US. A small delta can add up over thousands of GPU hours.
5. Schedule a Capacity Planning Call
If your organization is planning a large-scale AI deployment—say, a new Copilot extension or a custom model for document processing—talk to your Microsoft account team. They can provide guidance on which specific Azure clusters have the latest accelerators and the shortest wait times.
What to Watch Next
Fairwater is just the first of at least two phases planned for the Mount Pleasant site. A second datacenter, reportedly codenamed "Lakeview," is already under construction next door and expected to come online in 2027. Combined, they will double the AI compute capacity Microsoft can offer in the central United States.
More broadly, this opening is part of a torrent of datacenter investments by Microsoft, Amazon, and Google—collectively projected to exceed $200 billion in 2025-2026 alone. With AI still hungry for every petaflop, the race to build is far from over. For Windows and Azure users, the payoff is a cloud that can finally keep up with its own AI ambitions.