Overview
Microsoft unveiled its Fairwater AI Data Centre in Wisconsin on 3 June 2026. The $7.3 billion facility features a closed-loop cooling system that is filled with water just once and then continuously recycled. The centre aims to show how AI can grow sustainably with minimal water use.
A Bold New Approach
On 3 June 2026, at the Microsoft Build 2026 developer conference, CEO Satya Nadella made a big announcement. Microsoft had unveiled its new Fairwater AI Data Centre in Mount Pleasant, Wisconsin. This is not an ordinary data centre. It is designed to solve one of the biggest problems facing AI today: the massive use of water and energy. Fairwater is Microsoft’s most ambitious AI infrastructure project yet. It spans about 315 acres and represents an investment of nearly $7.3 billion . The facility is already live and running. It will support the rapidly growing demand for AI computing power while trying to minimise harm to the environment.
What Makes Fairwater Different?
Most data centres are built horizontally. They spread out over large areas. Fairwater does something different. It uses a two-storey vertical structure . This means the computing systems are stacked on top of each other instead of spread out. This design has several benefits:
-
Dense GPU deployment – Hundreds of thousands of NVIDIA GB200 GPUs are packed closely together.
-
Improved operational efficiency – Shorter connections mean faster communication between processors.
-
Reduced land requirements – A smaller physical footprint for the same computing power.
-
Lower latency – Data travels shorter distances, so responses are faster.
By placing the computing infrastructure closer together, Microsoft aims to optimise performance while reducing the physical footprint of large-scale AI operations.
The Core Innovation: Closed-Loop Cooling System
The real star of Fairwater is its closed-loop liquid cooling system . This is the centrepiece of Microsoft’s sustainability strategy. Here is how it works: the cooling system is filled with water only once during construction. After that, the same water is continuously recycled within the system. It circulates between servers and chillers, absorbing heat and releasing it without ever needing fresh water.
In simple words:
-
Water circulates in a closed loop.
-
Minimal freshwater intake is needed during operations.
-
Water wastage is almost eliminated.
-
Heat is managed efficiently without constant new water.
Satya Nadella explained this clearly during his keynote. He said that the Fairwater facility consumes no more water annually than a typical neighbourhood restaurant . That is a huge improvement compared to traditional data centres, which can use millions of gallons of water every day.
Why Data Centre Cooling Matters
AI models are becoming larger and more complex every day. Training these models requires enormous computing power. Powerful AI systems generate huge amounts of heat. If that heat is not removed, the processors will overheat and fail. Cooling is essential. But cooling itself comes with a cost.
Traditional data centres use evaporative cooling . This method consumes large amounts of water. Water is evaporated to remove heat. That water is lost forever. As AI grows, so does the water problem. A recent United Nations report warned that by 2030, AI and data centres could use 9.3 trillion litres of water every year. That is enough to provide drinking water to over 8 billion people.
Microsoft’s closed-loop system addresses this problem directly. By recycling the same water continuously, Fairwater avoids the need for constant freshwater intake. The system also uses chip-level liquid cooling , which is more efficient than cooling the entire room.
Beyond Water: Energy and Design
Fairwater is not just about water. It is also designed to be energy-efficient. The vertical stacking of GPUs reduces the distance data must travel. This improves latency and effective bandwidth. It also reduces the energy needed for data transfer.
Additionally, Microsoft has committed to powering its data centres with renewable energy wherever possible. The company has pledged to be carbon negative by 2030. Fairwater is a key part of that promise.
A Model for the Future
Fairwater is not a one-time experiment. Microsoft plans to build more data centres using the same design. The company has already announced a second facility of similar scale in Wisconsin. Together, they represent Microsoft’s total commitment of over $7 billion to the state.
The closed-loop cooling system will be deployed in other locations as well. Facilities in Phoenix, Arizona, and other sites will pilot the same technology starting in 2026. This marks a major shift in how Microsoft builds its AI infrastructure.
Environmental Concerns and Microsoft’s Response
In recent years, tech companies have faced growing criticism over the environmental impact of AI. A report from the United Nations University released on the same day as Microsoft’s announcement warned that AI’s water and energy use could double by 2030. Communities near data centres have protested against the strain on local water resources.
Microsoft’s Fairwater project is a direct response to these concerns. By designing a facility that uses water only once and then recycles it, Microsoft hopes to show that AI can grow without destroying local water supplies. Satya Nadella emphasised this point during his Build 2026 keynote: “We need a new approach to our data centres. That new approach is Fairwater.”
What the Numbers Say
Here are the key facts about Fairwater:
-
Investment: Approximately $7.3 billion (first phase $3.3 billion, second phase similar scale)
-
Location: Mount Pleasant, Wisconsin
-
Size: About 315 acres (three large buildings)
-
Cooling system: Closed-loop liquid cooling, filled once during construction
-
Annual water use: Equivalent to a neighbourhood restaurant
-
Status: Already live and operating (went live ahead of schedule in April 2026)
-
GPUs: Hundreds of thousands of NVIDIA GB200 chips
Conclusion: A Blueprint for Sustainable AI
Microsoft’s Fairwater AI Data Centre is more than just a building. It is a statement. It says that AI can be powerful and sustainable at the same time. The closed-loop cooling system is a genuine innovation. It solves one of the biggest environmental problems of AI infrastructure. If