Nvidia says its new data center design will fix AI's water problem, according to a report published by Fortune.

The company has not yet released full technical specifications through this channel, and the report as carried here consists of the headline claim alone. No figures on expected water savings, cooling architecture, deployment dates, or named data center partners accompany the announcement in the material available.

What does the claim address?

AI training and inference workloads drive significant water use in two ways:

• Direct cooling of servers and GPUs in data center facilities
• Indirect water consumption at power plants generating the electricity those facilities draw

Water use per megawatt of IT load varies widely by cooling method, climate, and facility design. Any design claiming to "fix" the problem would need to quantify reductions against a stated baseline, which the current material does not do.

What is missing from the announcement?

Readers evaluating this claim should look for:

• Measured or projected liters of water consumed per kilowatt-hour, or per training run
• The cooling technology involved (liquid cooling, direct-to-chip, immersion, or closed-loop systems)
• Whether the claim covers direct on-site use only, or includes electricity-related consumption
• Deployment timelines and which Nvidia platforms the design applies to

Die Signal will update this story as Nvidia publishes engineering details, benchmark data, or documentation supporting the claim.

Editor's note: the syndicated item underlying this report contained a headline only, without body text. This article reports the claim as stated and does not add specifications that the source did not provide.