NVIDIA Announces Solution for AI Water Consumption Problem

NVIDIA, the world's largest chip manufacturer, has introduced a new cooling system aimed at drastically reducing water consumption in data centers. Company representatives claim that this technology can bring one of the biggest environmental challenges of AI systems — massive water consumption — down to nearly zero. However, industry experts recommend approaching this statement with caution. This is reported by Techcrunch.com news reports.

According to an interview with Axios given by NVIDIA's Director of Sustainability Josh Parker, the water consumption problem in data centers is nearly solved with the help of the new system. The technology is based on a closed-loop liquid cooling system that efficiently removes heat from devices and eliminates the need for external water sources. This is seen as a revolutionary development, especially for massive server farms equipped with NVIDIA chips.

Closed-loop and warm water technology

The operating principle of the new system is unique: it pumps liquid at a temperature of 45°C into the server racks. While this may seem hot to a human, it is considered an ideal temperature for computer chips. The liquid passes through the servers and exits at 55°C, carrying away a large amount of heat. Most importantly, at this temperature, external air can perform cooling via passive radiators, eliminating the need for evaporative cooling methods.

According to ixbt.com, such a system not only saves water but also makes data centers significantly quieter and more energy-efficient. This is because the need for massive fans and chiller units used in traditional cooling systems is significantly reduced. In certain climatic conditions, this can reduce water consumption at the facility by up to 100%.

The invisible side of the problem

However, TechCrunch analysts believe that NVIDIA is only accounting for consumption within the data center building. In reality, the "water footprint" of AI is much broader. The problem is that the power plants supplying electricity to data centers and the chip manufacturing process still require enormous amounts of water.

Experts emphasize that NVIDIA's solution covers only a quarter or a third of total water consumption. If a data center receives energy from plants running on fossil fuels (coal or gas), the external environmental water consumption may offset the facility's savings. Additionally, the process of preparing chips in factories remains one of the most water-intensive industrial sectors in the world.

In conclusion, the technology presented by NVIDIA is an important step toward making AI infrastructure environmentally friendly. But for the industry to become fully "green," not only server cooling but also energy sources and the production chain must be fundamentally reformed. For now, AI's thirst for water remains a pressing global issue.