Data centers are thirsty. The kind of thirsty that drinks roughly 2.6 million gallons of water per megawatt of power every single year. Nvidia’s next-generation Rubin AI servers are designed to turn that firehose into a trickle, potentially reducing facility-level water consumption to near zero in the right climates.

The Vera Rubin NVL72 systems represent a full commitment to liquid cooling, not the hybrid approach Nvidia used with its Blackwell generation. These machines have no fans. Zero. Every watt of heat gets pulled away by a closed-loop coolant system running through cold plates mounted directly onto the processors.

How the plumbing actually works

The coolant is a mix of 75% water and 25% propylene glycol, the same stuff used in food-grade antifreeze. It circulates through cold plates that sit right on top of the processors, absorbing heat at the source before carrying it away.

The system operates at coolant temperatures up to 45 degrees Celsius, which is warm enough to reject heat into ambient air using dry coolers. Instead of evaporating millions of gallons of water through traditional cooling towers, the system just pushes warm air outside. No evaporation, no water loss.

The elimination of fans is worth pausing on. Cooling-related power draw has historically accounted for up to 40% of a data center’s total electricity consumption.

Why this matters beyond engineering elegance

Nvidia timed the announcement strategically. Key executives Ali Heydari and Josh Parker presented the environmental and efficiency benefits during London Climate Week in June 2026.

Manufacturing partners are already moving. Supermicro and other server partners began ramping up production and support for Rubin liquid-cooled systems as of January 2026.

One important asterisk: Nvidia’s water conservation claims focus exclusively on facility-level consumption. The water used in generating the electricity that powers these data centers is a separate conversation entirely, and it’s one that Nvidia’s announcements don’t address.

What this means for investors

The competitive landscape is shifting as well. AMD and Intel have been racing to match Nvidia’s AI compute performance. But cooling infrastructure is a different kind of moat. If Nvidia’s liquid cooling systems deliver meaningfully lower total cost of ownership, including water and energy savings, that creates a procurement advantage that raw chip benchmarks alone don’t capture.

The risk to watch is execution. Liquid cooling at this scale is more complex than traditional air cooling. Leaks, maintenance requirements, and retrofit costs for existing facilities could slow adoption. The closed-loop design mitigates some of those concerns, but any technology that puts liquid next to processors worth tens of thousands of dollars needs to work flawlessly.