NVIDIA Vera Rubin is the first fully liquid-cooled AI server

Data centers managing AI workloads face rising electricity bills and water scarcity, but NVIDIA's new Vera Rubin platform offers a technical solution that cuts power costs while eliminating water usage. The system replaces traditional air cooling and evaporative towers with a fully liquid-cooled design that submerges all components in a closed loop. This shift matters to operators because it directly addresses two of the biggest operational expenses in modern AI infrastructure.

NVIDIA unveils its first fully liquid-cooled AI server platform

NVIDIA introduced Vera Rubin as the world's first fully liquid-cooled AI server infrastructure platform. The architecture removes all fans from the chassis, relying instead on a coolant mixture to manage heat. This design allows the server to maintain stable chip temperatures within certified limits even under full load. The system ensures no performance degradation occurs despite the high operating temperatures.

Vera Rubin Specifications

• Cooling: Fully liquid-cooled, no fans
• Coolant: 75% pure water, 25% propylene glycol
• Inlet Temperature: 45 degrees Celsius
• Compute Density: 3x higher than traditional racks
• Rack Space: Reduced from 6 to 2 units

The cooling system uses a specific mixture of 75% pure water and 25% propylene glycol. Coolant enters the system at 45 degrees Celsius and exits at 55 degrees Celsius. This temperature range is significantly higher than traditional air-cooled systems, which typically require much lower inlet temperatures. The higher inlet temperature allows the facility to reject heat to the atmosphere more efficiently without using water.

The platform delivers a threefold increase in compute density, reducing the rack space required from six units to just two. This density gain allows data centers to pack more processing power into the same physical footprint. NVIDIA states that each 1 degree Celsius increase in coolant temperature reduces electricity costs by 4%. The system achieves near-zero water consumption, avoiding the millions of gallons used by traditional cooling towers.

Vera Rubin represents a structural shift in how AI hardware manages thermal and power constraints. The platform combines high-density computing with a cooling method that drastically lowers operational costs. Data centers adopting this technology can reduce their power bills while eliminating water usage entirely.