Training and running AI models requires enormous amounts of computing power, and keeping those powerful chips cool often means consuming large amounts of water. It's become a growing source of concern as tech companies build massive AI data centers across the United States.
Now NVIDIA says it has a solution. The company recently unveiled a new liquid-cooling system designed for its next-generation AI infrastructure that it claims could reduce cooling water consumption to nearly zero in certain environments.
The announcement arrives as public scrutiny of AI's environmental footprint continues to grow, with communities questioning whether new data centers are placing too much strain on local resources.
So what exactly is NVIDIA proposing, and does it really solve AI's water problem?
Why AI data centers use so much water
AI systems run on specialized chips called GPUs, many of which are designed by NVIDIA. These chips generate enormous amounts of heat when processing AI workloads.
Historically, many data centers have relied on cooling towers that use water evaporation to remove heat. While effective, the process can consume millions of gallons of water annually.
As AI models become larger and more powerful, cooling demands have increased alongside them, turning water consumption into one of the most controversial aspects of the AI boom.
The issue has become especially sensitive in regions already facing water shortages or drought conditions.
NVIDIA's new approach
According to NVIDIA, its latest cooling system circulates warm liquid directly around AI hardware at temperatures reaching approximately 45°C (113°F).
Rather than relying on traditional cooling towers, the warmer liquid can be paired with dry coolers, which function more like large radiators and do not require significant water consumption.
NVIDIA says this could reduce facility cooling water use from roughly 2.6 million gallons per megawatt annually to nearly zero in favorable climates.
The company also argues that operating at higher liquid temperatures improves overall energy efficiency while creating opportunities to reuse waste heat elsewhere.
In other words, NVIDIA's goal is o redesign the entire thermal management process around water conservation.
The important caveat
The company's claims primarily address one specific category of water use: cooling. And while that matters, cooling isn't the only way AI consumes water.
Water is also used throughout the broader AI supply chain, including electricity generation, semiconductor manufacturing and data center construction.
Even if cooling water usage falls dramatically, AI systems still carry an environmental footprint beyond the walls of the data center itself.
Of course, there's also the question of scale. AI demand continues to grow at an extraordinary pace. Critics argue that efficiency gains can sometimes be offset by rapid expansion, meaning total resource consumption may continue rising even as individual facilities become more efficient.
Why this matters
The industry's largest companies are now under pressure to demonstrate that future AI growth can occur without placing unsustainable demands on water and energy resources. NVIDIA is not alone here — Microsoft has made similar zero-water cooling claims for its newest data centers.
NVIDIA isn't claiming that AI no longer has an environmental footprint. Instead, it's targeting one of the most visible concerns surrounding modern data centers: water-intensive cooling system
Whether NVIDIA's approach becomes the new standard remains to be seen, but it offers a glimpse of how the industry may attempt to address one of its most persistent criticisms.
Follow Amanda Caswell and stay ahead of the AI curve
