As artificial intelligence continues to drive unusual growth in computing infrastructure, concerns over the environmental footprint of data centres are becoming increasingly prominent. Amid other concerns, one of the biggest challenges is water consumption. Recent disclosures revealed that Amazon’s global data centres used approximately 2.5 billion gallons (9.46 billion litres) of water in a single year, highlighting the resource demands of modern cloud and AI operations. Against this backdrop, Nvidia claims AI data centres can run with almost no water through a new liquid-cooling architecture designed for its next-generation AI systems.

The company’s latest approach aims to dramatically reduce water usage while improving energy efficiency, potentially reshaping how future AI facilities are designed and operated.

Why Data Centre Water Usage Matters

Data centres generate enormous amounts of heat while processing and storing data. Traditionally, operators have relied on air-conditioning systems, cooling towers, and water-intensive cooling infrastructure to keep servers running safely.

As AI models become larger and more computationally demanding, the amount of heat generated by servers continues to increase. This has intensified concerns about both energy consumption and water use, particularly in regions already facing water stress.

Also Read: America’s Data Center Backlash: Why Communities Are Challenging AI Infrastructure

Nvidia’s New Cooling Strategy

Nvidia’s latest Rubin-generation AI infrastructure introduces a fully liquid-cooled system that removes the need for conventional air cooling.

Instead of cooling entire rooms or server halls, heat is removed directly from the chips using liquid that flows through sealed cold plates attached to processors. The coolant absorbs heat and continuously circulates through a closed-loop system, allowing it to be reused rather than consumed.

Key Features of Nvidia’s Cooling Design

According to Nvidia, this design significantly decreases the need for water-intensive cooling infrastructure while enhancing overall data centre performance.

Also Read: India’s Artificial Intelligence Data Centre Faces Climate Threats As Temperatures Climb

How the Technology Works

A major innovation is the operating temperature of the cooling system.

Traditional data centres often require cooler environments to maintain equipment performance. Nvidia’s liquid cooling system, however, can safely operate with coolant temperatures reaching 45°C (113°F).

The company compares this temperature to a hot tub, noting that while most people can only tolerate water at 38-40°C for a limited time, its AI servers are designed to function efficiently at even higher temperatures.

By operating at warmer temperatures, facilities require less energy for cooling, which can substantially reduce overall electricity demand.

Also Read: China’s Underwater Data Centre Powered By Wind Energy Begins Operations

Near-Zero Water Consumption: Is It Possible?

Nvidia says its DSX AI Factory reference design is built around a completely closed-loop cooling system that continuously recycles coolant without requiring fresh water input during normal operations.

In favorable climates where dry coolers can reject heat efficiently, the company believes water use can be reduced to nearly zero. Ali Heydari, Director of Data Centre Cooling and Infrastructure at Nvidia, stated that the company has effectively eliminated most water consumption and significantly reduced power requirements through this approach.

However, Nvidia acknowledges that some facilities may still require supplementary cooling during periods of extreme heat, depending on local climate conditions.

Also Read: AI Expansion Slows As US Data Center Projects Face Delays And Cancellations

The Energy Efficiency Advantage

Cooling systems have historically represented a substantial portion of data centre energy consumption.

Did You Know?

• Cooling can account for up to 40% of total data centre electricity use.
• Liquid cooling removes heat more efficiently than traditional air cooling.
• Higher operating temperatures reduce the need for energy-intensive chillers.
• AI workloads generate significantly more heat than many traditional computing applications.

By reducing cooling energy requirements, operators can potentially lower operating costs while improving sustainability performance.

Also Read: Artificial Intelligence’s Environmental Footprint: Balancing Innovation With Sustainability

Amazon’s Water Usage Highlights the Industry Challenge

While Nvidia is promoting a future of low-water AI infrastructure, today’s cloud industry remains heavily dependent on water.

• Amazon recently disclosed that its global data centres consumed 9.46 billion litres of water annually, equivalent to approximately 5% of the annual water consumption of the Seattle metropolitan area, where the company is headquartered.
• Amazon Web Services (AWS) maintains that its facilities continue to become more efficient and that water consumption varies significantly based on climate, location, and cooling technologies.

The disclosure nevertheless underscores the growing scrutiny facing the AI and cloud computing sectors as demand for digital infrastructure accelerates worldwide.

Also Read: The Impact Of Artificial Intelligence On Electrical Energy Grids

What This Means for the Future of AI Infrastructure

If Nvidia’s technology performs as promised at scale, it could help address one of the industry’s most significant environmental concerns. Reduced water consumption, lower energy use, and improved cooling efficiency could make future AI facilities more sustainable, particularly in water-scarce regions.

However, widespread adoption will depend on implementation costs, operational reliability, climate conditions, and how quickly operators transition away from traditional cooling methods.

As AI continues to expand globally, innovations like liquid cooling may become essential for balancing technological growth with environmental responsibility.

Also Read: Jeff Bezos On Human Water Consumption: A Growing Challenge For AI Expansion

FAQs

1. Why do AI data centres use so much water?

Most traditional data centres use water-based cooling systems to remove heat generated by servers and computing equipment.

2. How much water did Amazon’s data centres use?

Amazon reported that its global data centres used approximately 9.46 billion litres (2.5 billion gallons) of water in a single year.

3. How does Nvidia’s cooling system reduce water use?

The system uses a closed-loop liquid cooling design that continuously recycles coolant rather than consuming fresh water during normal operations.

4. What is direct liquid cooling?

Direct liquid cooling removes heat directly from computer chips using liquid flowing through specialized cooling plates attached to processors.

5. Can AI data centres really operate with zero water?

Nvidia claims AI data centres can run with almost no water consumption under favorable conditions, though some facilities may still require supplemental cooling during extreme weather.

Also Read: AI Data Center Energy Demand To Quadruple By 2030, Report Finds