NVIDIA DSX Targets Near-Zero Cooling Water; Nscale Reports 49% More Inference
NVIDIA says its DSX AI-factory designs could save 54 million liters of water a day at 2 GW, and early tests at Nscale and Lambda raised output within fixed power budgets.
1 min read
NVIDIA's illustration for the DSX AI factory platform; a rendering, not a photo of a site.
Image: NVIDIA
Source · Press image from the NVIDIA newsroom, used for news reporting.
NVIDIA said on October 21 at GTC Berlin that European AI clouds are adopting its DSX AI Factory platform, which pairs liquid-cooled computing with dry coolers to cut facility cooling water to near zero and adds software to get more output from a fixed power budget. According to NVIDIA's press release, a 2-gigawatt deployment could save 54 million liters of water a day compared with cooling-tower designs, equal to the daily water use of 180,000 households.
Nscale's 49% inference gain is the largest measured result, all inside fixed power budgets.
Illustration (AI-generated)
The cooling design relies on closed-loop liquid cooling and outdoor dry coolers that release heat into the air without evaporating water. NVIDIA said Vera Rubin infrastructure supports 100% liquid cooling with coolant temperatures of up to 45 degrees Celsius, and that local climate determines whether supplemental mechanical cooling is needed in the hottest periods.
