Nvidia says its upcoming data center reference design for its next AI “Rubin” generation significantly reduces water consumption. The company states that, with its new liquid cooling approach, it has eliminated “pretty much all water usage” in the fully liquid-cooled data center setup it describes.
Both outlets focus on Nvidia’s effort to respond to broader public concerns about data center resource use—particularly water and energy—associated with scaling AI compute. The Verge notes that Nvidia’s messaging highlights water reductions while also implying improvements in power usage, but it suggests the company’s claims do not fully address all environmental and infrastructure concerns that critics raise around AI data centers.
Fortune reports on Nvidia’s specific claim that the design removes nearly all water usage compared with prior assumptions about how liquid cooling might still require water for certain processes or heat rejection steps. Nvidia’s statements are presented as part of its planned reference design guidance, rather than independently verified measurements shared by external organizations. The extent of real-world performance and impacts beyond the reference design is not detailed in the provided reports.
