Nvidia has unveiled a new liquid-cooling design that it says can reduce on-site water consumption in AI data centres by up to 100% under suitable conditions. The system uses a closed-loop liquid cooling approach capable of operating at temperatures as high as 45°C (113°F), eliminating the need for traditional evaporative cooling towers that consume large amounts of water. The company argues that the technology will make future AI data centres more efficient while lowering both water and electricity use.
However, the article argues that reducing cooling water addresses only part of AI's environmental footprint. The largest share of AI's water consumption is indirect, stemming from the enormous electricity required to power AI servers. Power plants—particularly fossil fuel and some nuclear facilities—consume significant amounts of water for electricity generation, meaning AI's expanding energy demand continues to carry a substantial hidden water cost even if data centres themselves use little or no water for cooling.
The rapid expansion of AI infrastructure is also driving unprecedented growth in electricity demand. As companies build larger AI clusters, concerns are shifting from water efficiency alone to the broader sustainability of powering AI at scale. Experts argue that improving cooling technology must be accompanied by increased investment in renewable energy, more efficient computing hardware, and smarter data centre design to reduce AI's overall environmental impact rather than simply relocating resource consumption.
The article concludes that Nvidia's innovation is an important step toward making AI infrastructure more sustainable, but it is not a complete solution. While advanced cooling systems can significantly reduce direct water usage, the long-term environmental challenge lies in meeting AI's rapidly growing electricity needs with cleaner and more efficient energy sources. Achieving sustainable AI will therefore require progress across the entire infrastructure ecosystem—from chip design and cooling technologies to power generation and grid modernization.
