NVIDIA Vera Rubin Unveils 100 Percent LiquidCooled Technical Solution
On June 21, 2026, NVIDIA launched its Vera Rubin 100% liquid-cooling solution. It raises coolant temperature to 45°C for full-loop fan-free operation. Adopting large cold-plate and micro-channel technologies, it cuts data center PUE from 1.35 to 1.15 with near-zero cooling water consumption. The compact air liquid CDU unit delivers stable fluid control to match this high-power AI thermal architecture.
As single-rack power surpasses 200 kW, Vera Rubin upgrades liquid cooling into core mainstream infrastructure for hyperscale AI data centers. It features cable-free, hose-free and fan-free layout with rigid metal manifolds for fluid transmission. The air liquid CDU with manifold precisely regulates coolant flow and temperature across the whole cooling loop.
The solution applies 45°C high-temperature coolant. After absorbing chip heat, the outlet temperature reaches around 55°C, while cold plates keep chips running at full capacity within safe temperature ranges. Micron-precision micro-channel cold plates optimize internal heat exchange, managed centrally by the air liquid cooling distribution unit.
Vera Rubin uses 75% water and 25% propylene glycol mixed coolant for thermal conductivity, anti-corrosion and bacteriostasis. Higher return-water temperature enables free outdoor heat rejection in most regions, removing the need for high-energy mechanical chillers. The high density rack air liquid CDU supports reliable operation under ultra-high rack power loads.
Traditional cooling takes up 40% of total data center power consumption. Each 1°C increase in coolant temperature reduces cooling energy cost by about 4%. Vera Rubin achieves remarkable energy-saving effects, and qualified server rack air liquid CDU manufacturer provides compatible hardware for mass-scale project rollout.
Benefiting from 45°C high-temperature coolant design, the solution realizes nearly zero water consumption versus traditional cooling towers. Relevant hyperscale cases including chilldyne liquid cooling verify great progress in both power and water resource conservation for AI computing facilities.
Fan-free design eliminates over 85-decibel noise from conventional cooling hardware, and rack deployment density is greatly improved. Systems previously occupying six rack units can fit into two. A professional server cooling liquid supplier can supply matched mixed coolant to bring these advantages into full play.
A 50-MW hyperscale data center adopting Vera Rubin style liquid-cooled infrastructure can save over 4 million US dollars annually on energy and water costs. To realize such comprehensive benefits, operators need to deploy a complete and well-validated data center liquid cooling system for their AI compute clusters.
Every AI breakthrough depends on one thing: keeping the heat out. At EXTRCOOL, we make that our business. Our liquid cooling portfolio—CDUs, cold plates, manifolds, and piping—works right out of the box or tailored to your exact rack layout and fluid requirements. High efficiency, low environmental impact, and proven reliability at gigawatt scale. That's what global AI infrastructure runs on.
See the specs or talk to our team: https://www.extrcool.com
