Nvidia says its next-generation data center design slashes water use to near zero by running AI servers hotter than ever. According to The Verge AI, the company’s Rubin-generation reference design for a fully liquid-cooled data center has, in Nvidia’s words, “eliminated massive amounts of power usage and pretty much all water usage.” The trade-off is heat: these servers can run as high as 113 degrees Fahrenheit (45 degrees Celsius).
What stands out here is the scale of the claimed savings. Josh Parker, Nvidia’s head of sustainability, told The Verge AI the design takes water use “from roughly 2.6 million gallons per megawatt per year for conventional cooling-tower-based systems to near zero, up to a 100 percent reduction.”
What’s actually changing
Traditional data centers lean on cooling towers that evaporate huge volumes of water to shed heat. Nvidia’s approach skips most of that.
- Heat gets captured directly at the chip, not blown away by air conditioning.
- Liquid loops carry that heat at much higher temperatures than older systems allow.
- Outdoor “dry coolers” then reject the heat into the air, which works for much of the year without water.
- Running hotter gives operators more flexibility on ambient air temperature, so they’re less dependent on cool surroundings.
The key insight: hotter liquid is easier to cool with plain outside air. That’s why the higher operating temperature is a feature, not a bug.
Why this matters
Data centers have become a public flashpoint. Communities near these facilities have pushed back hard on how much water and power they swallow, especially in drought-prone regions. A credible path to near-zero water use is exactly the kind of answer the industry has been scrambling for.
This also isn’t an isolated bet. The Verge AI notes that Amazon recently touted higher heat tolerances as part of making its mostly air-cooled data centers more efficient. Running hardware hotter is becoming a shared industry strategy, not a one-off Nvidia experiment. And Nvidia claims “every cloud provider and data center operator building for [Rubin] is making the transition.”
The catch
This is significant, but it’s not the whole story, and The Verge AI is clear on the gaps.
- Construction impact stays. The water and resource cost of building these massive facilities isn’t addressed.
- Power generation is untouched. Cutting water use doesn’t shrink the enormous electricity demand these data centers place on the grid.
- No cost figures. As Gizmodo points out, Nvidia’s blog post doesn’t say what this design costs to build versus a less efficient air-cooled one. That silence matters when Nvidia claims everyone is switching.
So the water win looks real, but the energy and construction concerns that drive most of the public pushback remain on the table.
What to watch next
If Nvidia’s claim holds up in real deployments, expect liquid cooling at high operating temperatures to become the default spec for AI buildouts, not a premium option. The open questions are economic and practical: how much more these facilities cost upfront, how the higher heat affects hardware lifespan and reliability, and whether “near zero” water holds during the hottest stretches of the year when dry coolers struggle.
For practitioners planning AI infrastructure, the signal is clear. Cooling design is moving from an afterthought to a core part of the build, and water efficiency is now a number worth asking vendors about. You can find the full details in the original report at The Verge AI.