Nvidia, a leading chip manufacturer, has introduced a warm-water cooling system that it claims will significantly cut water usage inside data centres. The company suggests this technology could eliminate 'pretty much all water usage' within these facilities, with an executive stating that the 'water consumption challenge for data centres is largely solved' at the facility level.
The new system works by circulating coolant through server racks at approximately 45°C. After absorbing heat from the chips, the coolant exits at around 55°C. This elevated temperature allows for passive heat dissipation into the outside air in many climates, potentially removing the need for energy-intensive evaporative cooling towers or fans. Such a closed-loop system, once filled, recirculates the same water, meaning no new water is continuously consumed for cooling purposes directly within the data centre.
While this innovation represents a notable step forward for on-site efficiency, it addresses only a portion of the overall water footprint associated with artificial intelligence (AI) infrastructure. Critics point out that Nvidia's measurement of water use primarily focuses on the data centre's immediate perimeter, overlooking the substantial water consumption involved in generating the electricity that powers these facilities and in the manufacturing of the chips themselves.
The core issue lies in the reliance on fossil fuels for electricity generation. Power plants, particularly those burning coal and natural gas, are major consumers of water, largely for cooling processes. For instance, natural gas plants can use approximately 1.17 litres of water per kilowatt-hour of electricity generated, with coal plants being even more water-intensive at 2.2 litres per kilowatt-hour. Given that fossil fuel plants currently supply about half of all data centre power, the water used externally for electricity generation can more than double or even triple a data centre's total water footprint.
This means that even with Nvidia's advanced cooling, which tackles roughly a quarter to a third of a data centre's total water consumption, the broader environmental challenge remains. Renewable energy sources like wind and solar power, in contrast, have a significantly lower water footprint, using negligible amounts per kilowatt-hour generated. However, projections indicate that natural gas and coal are still expected to provide over 40% of the new electricity needed for data centres through 2030, according to the International Energy Agency (IEA).
For UK businesses and consumers, the implications of AI's growing energy and water demands are significant. As data centres proliferate to support AI development, their environmental impact becomes a critical consideration for corporate sustainability targets and national resource management. The UK's Information Commissioner's Office (ICO) has a role in ensuring data protection and ethical AI, but the broader environmental regulatory landscape, including potential future water usage mandates, could impact how these facilities operate. Expert commentary suggests that while technological advancements like Nvidia's are welcome, a holistic approach focusing on renewable energy integration is crucial for the UK to mitigate the environmental costs of its burgeoning AI sector and align with net-zero ambitions.