Elon Musk endorsed NVIDIA's "dry" manifesto: a revolution in cooling AI data centers
Large technology corporations are increasingly coming under fire for their colossal consumption of natural resources. The main public grievances are the astronomical amounts of electricity and fresh water required to operate modern data centers. However, NVIDIA resolutely rejects these accusations, presenting compelling arguments in favor of a new cooling paradigm.
Entrepreneur Elon Musk, whose company xAI actively uses clusters based on NVIDIA solutions, has publicly expressed full support for the chip manufacturer's environmental manifesto. The discussion about the carbon and water footprint of advanced technologies is entering a fundamentally new level.
Real Numbers: How Much Water Does the AI Industry Consume
In its blog, NVIDIA cites data from the Manhattan Institute for March 2026. According to these estimates, data centers consume only about 0.2% of all fresh water in the United States. At the same time, the lion's share of this volume comes from indirect consumption — that is, from generating electricity to power the servers.
Modern liquid cooling systems used in NVIDIA's architecture operate at a stable temperature of 45 degrees Celsius. This technological approach allows AI complexes in northern latitudes to fully transition to economical dry radiators, completely eliminating the need for bulky evaporative cooling towers.
It is important to note that back in 2025, NVIDIA claimed that the Blackwell platform is 300 times more water-efficient compared to traditional air cooling. Given that maintaining optimal temperature consumes up to 40% of a data center's total energy, such innovations are radically changing the rules of global competition in the artificial intelligence market.
The Role of Elon Musk and Hidden Risks
Musk, whose xAI is one of the largest consumers of NVIDIA chips, has repeatedly called them a key element of his projects. His public endorsement significantly strengthens the partners' position. Developers aim to completely dispel the popular myth that neural networks deplete regional water reserves.
NVIDIA's cooling system is an absolutely sealed closed loop. The liquid constantly circulates inside the equipment, so the servers do not require regular replenishment from the water supply. As stated by Ali Heydari, head of data center cooling and infrastructure at NVIDIA: "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."
However, optimistic national reports often mask real regional problems. Indirect costs are rapidly increasing amid the large-scale deployment of neural networks. Analysts at the Berkeley Lab predict that direct water consumption by data centers will jump to 38-73 billion gallons by 2028.
The efficiency of dry radiators is directly tied to the surrounding climate. The equipment shows excellent results in cool states, but its productivity drops in arid desert zones.
The startup xAI is also feeling environmental pressure. The giant Memphis Colossus computing complex daily extracted up to 1.3 million gallons of clean drinking water from underground sources. Moreover, the company launched dozens of gas turbines before obtaining official permits, sparking a wave of outrage among local residents and leading to high-profile lawsuits.
Expert opinion. While NVIDIA and Musk promote the narrative of "water neutrality," the reality is that indirect costs continue to grow exponentially. The further development of the industry will entirely depend on the strictness of environmental control by authorities. Investors should closely monitor regulatory risks in water-scarce regions — this is where the main battle for the future of AI may unfold.