Zero water usage where it really matters: your data center

AI workloads are reshaping where and how data centers are built. Rack densities and thermal loads are rising and the locations driving the next phase of AI growth are the same areas where water is already under pressure.

For more than a decade, the industry optimized around Power Usage Effectiveness (PUE) with leading AI-ready facilities now targeting PUE values of 1.2 or below. However, some of the lowest PUE figures come from evaporative cooling – an approach that consumes significant water for heat rejection. As AI infrastructure scales, that trade-off no longer holds in certain geographies. Improving power and water efficiency together is now essential for the long-term sustainability of AI Factories.

Global average Water Usage Effectiveness (WUE) in data centers sits around 1.8 L/kWh. However, efficient sites operate closer to 0.2 L/kWh and zero water consumption (0.0 L/kWh) now achievable through the right thermal architecture.

Industry frameworks reflect that shift. ISO/IEC 30134 and EN 50600 include WUE alongside PUE as part of an overall infrastructure assessment. The Green Grid, which created both metrics, positions WUE as a companion to PUE rather than a secondary consideration.

Johnson Controls is driving the response. Our advanced thermal innovation is engineered across the full thermal chain – from chip to heat rejection. That depth is captured in a comprehensive set of system-level reference design guides that give operators the clarity to plan and scale. And it's delivered through the largest global field organization in the industry, redefining what data centers can achieve as AI scales.

The site sets the cooling strategy

Cooling architecture continues to decide the water footprint of a data center.

Evaporative systems can deliver strong energy performance but depend on continuous water use. Air-cooled systems remove that dependency but require a different approach to maintain capacity and efficiency at AI-scale loads. The right choice depends on where the facility is built.

In water-constrained regions such as the U.S. Southwest or the Middle East, limiting water use can become the overriding priority. In cooler, wetter climates like Northern Europe, different cooling strategies may be more effective.

Zero water usage at AI density with Johnson Controls

In regions where water availability constrains growth, eliminating on site water use can fundamentally change what is possible. Air cooled and water chiller architectures play a central role in enabling this outcome, particularly for high density, multistory data centers that support AI training and inference at scale.

The Johnson Controls YORK YDAM magnetic bearing centrifugal chiller is designed for that role in high-density data center applications. Its advanced cooling capacity enables faster, more-efficient scaling without consuming water.

With up to 3.5MW of cooling to support AI workloads, it delivers up to 20% increased capacity density than comparable solutions and fits more performance into constrained footprints. It also utilizes magnetic bearing compressors to support the higher chilled-fluid setpoints required for 45°C warm-water cooling of the latest generation of inference and training GPUs.

Where the climate makes water-cooled architecture the more efficient route, the Johnson Controls YK-HT two-stage centrifugal chiller paired with dry coolers achieves the same zero water outcome.

Control that spans the full thermal chain

Zero water use depends on the full thermal chain, not just the chiller.

The Johnson Controls Global Reference Design Guides provide validated cooling architectures for gigawatt-scale data centers that are engineered to enable industry-leading PUE and WUE.

Guide 401 sets out an air-cooled chiller plant capable of serving a gigawatt-scale AI Factory. The reference design integrates fan coil walls, CDUs and high efficiency air-cooled centrifugal chillers to manage both air and liquid cooled IT loads and deliver a fully water-free heat rejection process.

Guide 402 addresses water-cooled architectures where climate makes that route more efficient. Together, they give operators a validated path to zero water usage, regardless of site conditions.

Zero water usage where it really matters

Where freeing up power gives operators the option for more compute, removing water removes the resource constraint that determines whether AI infrastructure can scale.

We create the conditions AI infrastructure depends on. It’s engineered through advanced thermal innovation, validated at AI factory scale through Global Reference Design Guides, and built on 140 years of engineering the systems that mission-critical environments depend on.

FAQs

1. What is zero water usage in a data center?

Zero water usage in a data center refers to the elimination of onsite water consumption in the cooling process. It is achieved through advanced thermal architectures – including air-cooled chillers and water-cooled chillers paired with dry coolers – that remove evaporative heat rejection from the cooling system entirely. Johnson Controls delivers zero water usage at gigawatt scale through YORK thermal technologies and validated reference designs for AI factories.

2. Can data centers achieve zero water usage without compromising performance?

Yes. Advanced thermal architectures now deliver zero onsite water use while sustaining the cooling density AI workloads require. The YORK YDAM air-cooled magnetic bearing centrifugal chiller supports up to 3.5MW of cooling and up to 20% higher capacity density than comparable solutions. It also enables AI training and inference workloads at scale without consuming water on site.

3. How does Johnson Controls enable zero water usage at AI factory scale?

Johnson Controls enables zero water usage at AI Factory scale through YORK thermal technologies and validated system-level reference designs. Reference Design Guide 401 provides a blueprint for a fully water-free heat rejection process using air-cooled centrifugal chillers, integrated with fan coil walls and CDUs to manage both air- and liquid-cooled IT loads at gigawatt scale.