Nvidia Claims New Chip Cooling Tech Solves Data Center Water Use

Nvidia has announced that its latest AI systems can operate with a liquid cooling technology that runs warm enough to eliminate or drastically reduce the need for water-intensive chilling equipment in data centers.

The claim, made by Josh Parker, Nvidia's chief sustainability officer, at London Climate Week represents a significant assertion from the world's leading AI chip manufacturer as data centers face mounting criticism over their environmental footprint. "The water consumption challenge for data centers is largely solved," Parker said in an interview.

Why it matters

Data centers powering the AI boom are drawing intense scrutiny from local communities and regulators over their consumption of water and electricity. If Nvidia's technology delivers on its promise at scale, it could reshape the economics and environmental profile of AI infrastructure buildout — though efficiency gains may also enable even more rapid expansion.

How the cooling system works

Nvidia's approach uses a recirculated liquid mixture containing water and propylene glycol, similar to automotive antifreeze, that can operate at 113 degrees Fahrenheit. Previous liquid cooling systems required much lower temperatures, necessitating mechanical chillers that consume substantial energy and water to maintain those conditions.

Because the new coolant runs warmer, data centers could potentially operate without chillers in many environments. Steve Solomon, Microsoft's vice president of data center engineering, said the technology could eliminate mechanical chiller requirements "in most climates most of the time — even in hot places such as Arizona."

Adoption challenges remain

Several factors will determine whether Nvidia's technology transforms industry practice. The company declined to disclose system costs, and widespread adoption depends on the economics of retrofitting existing facilities or building new ones designed for fully liquid-cooled infrastructure. Nvidia maintains the systems will reduce overall cooling costs for operators.

The transition will also take years. Many data centers currently running older cooling technologies will continue operating, meaning the environmental benefits won't materialize immediately across the industry.

The broader water equation

Even dramatic reductions in cooling-related water use don't eliminate all concerns. Generating the electricity required to power AI systems can itself consume significant water, depending on the energy source. Coal and natural gas plants, as well as many nuclear facilities, use water for cooling.

Nvidia acknowledges that efficiency improvements are designed to support continued growth rather than reduce absolute resource consumption. "AI workloads are not getting lighter," Parker wrote in a blog post, arguing that without efficiency gains, energy demand would rise even faster alongside AI adoption.

The announcement comes as Google and Amazon have faced local opposition to data center projects and defended their water management practices. Tech companies increasingly argue that efficiency improvements will mitigate the environmental impact of AI infrastructure expansion.

These details were first reported by Axios.