News
DealsCorvex agrees USD 103 million take-or-pay deal to perform AI inference with Mid-Atlantic GPUsSovereignNeysa scales AI infrastructure in India with AMD Instinct MI350X GPUs on TytanInferenceMicrosoft Foundry adds Baseten, DeepSeek AI inference (MSFT:NASDAQ)ClustersDataOne and Rowan University sign engineering co-op program agreement for DataOne's Vineland AI factoryClustersLook inside Cal Poly's $3 million AI Factory (Video)DealsOpenAI Taps AMD Chips for 457 MW of AI Capacity in Six-Year TensorWave DealClustersCal Poly launches an AI data center on campus. Take a look insideInferenceArista takes aim at the next AI network battleground: distributed inferenceDealsOpenAI Signs 457 MW AI Data Center Deal as AMD Expands Its Challenge to NvidiaClustersBehind Dell AI Factory’s NVIDIA & AMD Enterprise ExpansionDealsCorvex agrees USD 103 million take-or-pay deal to perform AI inference with Mid-Atlantic GPUsSovereignNeysa scales AI infrastructure in India with AMD Instinct MI350X GPUs on TytanInferenceMicrosoft Foundry adds Baseten, DeepSeek AI inference (MSFT:NASDAQ)ClustersDataOne and Rowan University sign engineering co-op program agreement for DataOne's Vineland AI factoryClustersLook inside Cal Poly's $3 million AI Factory (Video)DealsOpenAI Taps AMD Chips for 457 MW of AI Capacity in Six-Year TensorWave DealClustersCal Poly launches an AI data center on campus. Take a look insideInferenceArista takes aim at the next AI network battleground: distributed inferenceDealsOpenAI Signs 457 MW AI Data Center Deal as AMD Expands Its Challenge to NvidiaClustersBehind Dell AI Factory’s NVIDIA & AMD Enterprise Expansion
EveryNeocloud
Microsoft Fairwater

Microsoft Fairwater

Purpose-built AI supercomputer campus in Wisconsin

Microsoft Fairwater is a specialised AI datacenter campus in Mount Pleasant, Wisconsin. It functions as a single massive supercomputer using hundreds of thousands of Nvidia Blackwell GPUs to train frontier models.

Share on WhatsApp

  • Covers 315 acres with 1.2 million square feet of space
  • Uses hundreds of thousands of Nvidia Blackwell GPUs
  • Delivers 10 times the performance of top supercomputers
  • Features closed-loop liquid cooling with zero water waste
  • Processes 865,000 tokens per second in unified clusters

What it is

Microsoft Fairwater is a purpose-built AI datacenter located in Mount Pleasant, Wisconsin. The company describes it as an AI factory designed specifically for training and running large-scale artificial intelligence models. Unlike standard cloud facilities that host independent workloads, this site operates as one unified supercomputer. It connects hundreds of thousands of graphics processing units into a single flat network. This architecture allows the system to function as a giant accelerator for complex mathematical computations required by modern AI.

How it came about

Microsoft announced the Fairwater facility in September 2025 as part of a broader wave of infrastructure investments. The company stated that multiple identical datacenters are under construction in other locations across the United States. This project represents tens of billions of dollars in capital expenditure. It follows the company’s strategy to link these specialized facilities into a distributed network. The goal is to multiply efficiency and democratize access to AI services globally through this expanded infrastructure.

The compute

The core of the facility consists of dedicated AI accelerators, specifically Nvidia GB200 servers. Each rack contains 72 Nvidia Blackwell GPUs linked by NVLink and NVSwitch technology. This setup creates a pooled memory domain of 14 terabytes per rack. The system uses InfiniBand and Ethernet fabrics to ensure low-latency communication between racks. This full fat-tree non-blocking architecture allows every GPU to communicate at full line rate. The cluster can process 865,000 tokens per second, supporting trillion-parameter model training.

Where the GPUs are

The campus covers 315 acres and includes three massive buildings with 1.2 million square feet of roofed space. Construction involved 46.6 miles of deep foundation piles and 26.5 million pounds of structural steel. The site features the second-largest water-cooled chiller plant on the planet. A closed-loop liquid cooling system circulates cold liquid directly into servers to extract heat. This design eliminates operational water waste by reusing water continuously without evaporation losses.

Customers and deals

The datacenter powers workloads for OpenAI and Microsoft AI products. It supports Copilot capabilities and other leading artificial intelligence applications. The infrastructure is designed to handle the demands of frontier AI models. By providing this scale of compute, Microsoft aims to anchor enterprise AI workloads within its Azure cloud platform. The facility serves as a primary engine for the company’s global AI services.

Its place in AI

Fairwater represents a shift toward purpose-built AI factories in the industry. It delivers ten times the performance of the world’s fastest traditional supercomputer. The facility is part of Microsoft’s global AI wide area network. This network links over 400 datacenters across 70 regions. The project demonstrates how hyperscalers are adapting physical infrastructure to meet the unique networking and cooling needs of large language models.

Questions people ask about Microsoft Fairwater

What makes Fairwater different from a normal datacenter?

Normal datacenters run many small, independent tasks like hosting websites. Fairwater is built to work as one massive supercomputer. It connects hundreds of thousands of GPUs into a single network so they can train one large AI model together.

How does the facility manage heat from so many chips?

It uses a closed-loop liquid cooling system. Cold liquid circulates directly into the servers to remove heat. The water is reused continuously without evaporation, which means the system requires no additional water during operation.

What kind of AI chips does Fairwater use?

The facility runs Nvidia GB200 servers. Each rack holds 72 Nvidia Blackwell GPUs. These chips are linked together to share memory and bandwidth, allowing them to act as a single giant accelerator for AI training.

Good reading and sources

    Where this page comes from

    There is no English Wikipedia article on Microsoft Fairwater yet, so the facts on this page come from the sources listed above: blogs.microsoft.com/blog/2025/09/18/inside-the-worlds-most-powerful-ai, convergedigest.com/microsoft-unveils-fairwater-ai-data-center-in-wisco. Some are the subject's own pages, so its claims about itself are reported as its claims. EveryNeocloud wrote the text in its own words, with AI assistance, and compares it with the sources to make sure no passage is copied.

    It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.

    Photo credits on this page

    The small pictures above come from Wikimedia Commons under free licences. Some are cropped to fit.

    The weekly digest

    Every Monday: the provider of the day, new pages on EveryNeocloud, and the conferences and anniversaries of the week ahead.