Elon Musk has endorsed NVIDIA's "green" manifesto: the water footprint of AI infrastructure under the microscope.
Tech giants are increasingly coming under public scrutiny due to their massive consumption of natural resources. The issue of electricity and fresh water usage for cooling data centers is particularly acute. NVIDIA's management firmly rejects these accusations, and this time, Elon Musk has publicly supported their stance.
Facts and Figures: How Much Water Do Data Centers Actually Use?
In a recent statement, NVIDIA relies on data from the Manhattan Institute from March 2026. According to these estimates, all U.S. data centers consume only about 0.2% of the country's fresh water. Moreover, the lion's share of this volume comes from indirect consumption — the generation of electricity needed for their operation.
The company's key argument is the transition to closed-loop liquid cooling systems. Modern solutions operate at a stable temperature of around 45 degrees Celsius. In northern latitudes, this allows for a complete shift to economical dry coolers, eliminating the need for bulky cooling towers. As early as 2025, NVIDIA claimed that the Blackwell platform is 300 times more water-efficient than traditional air cooling.
Given that maintaining optimal temperature conditions consumes up to 40% of a data center's total energy, such innovations radically change the economics and environmental footprint of AI infrastructure.
Musk's Support and Hidden Risks
Elon Musk, whose company xAI actively uses clusters based on NVIDIA solutions, publicly endorsed this manifesto. His support significantly strengthens the position of partners seeking to dispel the myth that neural networks deplete regional water supplies. The cooling system is indeed a sealed closed loop where fluid circulates without constant replenishment from the water supply.
"The NVIDIA DSX reference architecture for AI centers does not consume water. We have managed to minimize energy consumption and almost completely eliminate water usage," said Ali Heydari, head of cooling and data center infrastructure at NVIDIA.
However, optimistic reports often mask regional issues. Indirect costs are rapidly rising amid the large-scale deployment of neural networks. According to forecasts from analysts at the Berkeley Lab, direct water consumption by data centers could surge to 38-73 billion gallons by 2028.
Furthermore, the efficiency of dry coolers is directly tied to climate. In arid desert zones, their productivity drops sharply. A telling example is the giant xAI Memphis Colossus complex, which daily drew up to 1.3 million gallons of clean drinking water from underground sources. The company also launched dozens of gas turbines before obtaining official permits, sparking a wave of lawsuits from local residents.
Expert Opinion: NVIDIA's environmental manifesto is an important step in shaping a positive industry image, but it does not eliminate fundamental problems. The further development of AI will entirely depend on the strictness of environmental oversight by authorities and the ability of companies to scale truly "dry" cooling technologies, rather than simply shifting the water footprint to the energy sector.