Nautilus EcoCore Cooling Unit Earns NVIDIA Validation for AI Factory Duty

Nautilus Data Technologies, Inc. has announced that its EcoCore Facility Cooling Distribution Unit (FCD) has been recognized as NVIDIA-Validated AI Factory Infrastructure — a designation that puts the liquid cooling system on NVIDIA's DSX Infrastructure Marketplace with two notable distinctions: the highest-rated capacity among liquid-to-liquid cooling distribution units listed there, and the first Large-Scale CDU validation.

For data center operators weighing cooling options for dense AI deployments, that's a meaningful signal. Validation against NVIDIA's requirements removes some of the guesswork from pairing facility infrastructure with AI factory workloads.

Capacity and Thermal Performance

The numbers behind the validation are substantial. The EcoCore FCD demonstrated cooling capacity exceeding 4.1 MW in N-mode dual-pump operation, and 3.2 MW in N+1 mode — the latter figure being the one that matters for facilities that can't afford to drop capacity when a pump goes down.

The unit also demonstrated a 2°C approach temperature difference (ATD). A tight approach temperature translates directly into heat transfer efficiency, and 2°C is an aggressive figure for a facility-level cooling distribution unit.

Water Flexibility and Serviceability

One of the more practical differentiators: the EcoCore FCD is water-source agnostic. It can run on process water, chilled water, river water, seawater, or wastewater — giving operators latitude in how they source cooling water depending on site geography and available infrastructure.

Serviceability follows the same philosophy. The unit is designed to be maintained and upgraded in place, and it can integrate with Nautilus's degassing technology, the Air Detection Removal System.

Track Record and Availability

The EcoCore FCD carries more than 650,000 unit-hours of production operation across three product generations, backed by Nautilus's 13 years of designing and delivering production environment cooling technologies. Delivery times are a proven 15 weeks from order — a figure worth noting in a market where infrastructure lead times often dictate deployment schedules.

The unit is designed to support both today's AI workloads and the next generation of AI computing, positioning it as infrastructure intended to outlast a single hardware refresh cycle.