Nvidia introduced water-free data centers for cooling chips

6/25/2026, 02:55 PMЕвгения Слив

One of the major reputational challenges for Silicon Valley remains the significant environmental impact of artificial intelligence infrastructure. This week, Nvidia announced that it has eliminated at least one major component of the AI industry’s ecological footprint – its dependence on local water resources. Traditionally, most water in such facilities is used to cool chips that generate intense heat during energy‑intensive computations. Nvidia’s breakthrough lies in a new closed‑loop cooling system that circulates a liquid coolant made of 75% water and 25% propylene glycol – a chemical mix similar to automotive antifreeze.

A key detail: the coolant remains effective at temperatures up to 46 degrees Celsius. This allows the new Nvidia DSX data centers to avoid overheating without using massive amounts of water and fans, reducing operational costs and environmental impact. Although each new generation of infrastructure offers more computing power per watt, a full‑liquid cooling system dramatically cuts energy consumption for cooling, which fundamentally changes overall energy usage at the hyperscaler level.

The new system, which Nvidia claims delivers “up to 100% reduction in water consumption,” is sure to draw attention from the industry amid growing public discontent. A Pew Research Center survey showed that most Americans consider data centers harmful to the environment and to the quality of life in nearby communities. Some politicians have even attempted to blame environmental activists and foreign forces for this public backlash. In response, Google and Microsoft have also announced new efforts to reduce water usage and develop “edge” computing, where AI tools run on local devices rather than in large data centers.

Beyond environmental benefits, the new cooling system will appeal to an industry desperate to lower the cost of AI access. If technology companies can cut overhead by reducing data‑center energy consumption, it could pave the way for lower token prices – the basic unit of AI usage measurement. Water‑related challenges are also pushing investors and developers toward the idea of building data centers in space. Such facilities could simply dissipate excess heat into a vacuum, requiring neither fans nor water, though their practical feasibility remains an open question.

Nvidia’s claim of “zero‑water” cooling comes with several important caveats. The system’s efficiency depends on the local climate. If outside temperatures approach 46°C, as they sometimes do in Arizona and Nevada, additional cooling resources will be required. Nvidia itself acknowledges that in hot climates, industrial chillers will still need to operate on the hottest days. Furthermore, it is too early to talk about scaling the new architecture across the entire industry; the company did not disclose implementation costs. In the foreseeable future, most large AI labs are likely to continue using traditional, more water‑intensive facilities.

It is also worth remembering that water consumption is only a small part of the industry’s overall environmental footprint. Many data centers run on fossil fuels, emitting significant greenhouse gases, and the enormous resources poured into building them often come at the expense of developing more sustainable alternatives like wind and hydro power.

Popular news