DCX Liquid Cooling Systems introduced the second-generation Facility Distribution Unit, FDU V2AT2, targeting large-scale AI data centers moving to 45°C warm-water cooling. Announced from Warsaw, the system aligns with NVIDIA’s guidance for Vera Rubin rack-scale architectures, where 45°C supply water enables chillerless heat rejection for high-density AI workloads.
The FDU V2AT2 delivers up to 8.15MW of heat transfer capacity with a record 550 m³/h flow rate, consolidating multiple legacy CDUs into a single facility-scale platform. By supporting ASHRAE W45 / W+ classes and maintaining tight control to avoid condensation, the system allows operators to simplify cooling loop topology while reducing both capital and operating costs.
Unlike row-based CDUs, the new platform is designed to serve entire data halls, supporting hyperscale deployments of NVIDIA NVL72 GB200 / GB300 Blackwell and future Vera Rubin systems. DCX positions the FDU V2AT2 as a foundation for AI factories that require multi-megawatt cooling capacity, high availability, and readiness for heat reuse strategies.
- Supports 45°C warm-water cooling, enabling chillerless heat rejection in many deployments
- Up to 8.15MW heat transfer capacity at the facility level
- 550 m³/h flow rate optimized for high-density AI racks
- 2°C approach temperature heat exchanger for heat reuse and low-cost heat rejection
- Four pumps in N+1 configuration for high availability
- Integrated water quality monitoring and treatment
“As the datacenter industry transitions to AI factories, operators need cooling system that won’t be obsolete in one platform cycle,” said Maciek Szadkowski, CTO at DCX. “The FDU V2AT2 replaces multiple legacy 1.3MW CDUs and enables 45°C supply water operation, opening a clear path to NVIDIA Vera Rubin architecture and beyond while significantly reducing both CAPEX and OPEX.”
🌐 Analysis
The announcement reflects a broader industry shift toward warm-water and chillerless cooling as GPU power densities rise and operators seek efficiency gains at scale. With NVIDIA explicitly calling out 45°C supply water for Vera Rubin systems, facility-scale CDUs like DCX’s FDU V2AT2 are emerging as a critical building block alongside competing high-capacity solutions from established liquid-cooling vendors.
🌐 We’re tracking the latest developments in AI data center infrastructure and cooling. Follow our ongoing coverage at: https://convergedigest.com/category/data-centers
DCX Liquid Cooling Systems is a Warsaw-based data-center cooling company focused on high-density liquid-cooling infrastructure for AI and HPC deployments. The company’s mission is to enable scalable, energy-efficient “AI factory” data centers through warm-water and chiller-less cooling architectures that reduce power consumption and support heat reuse. DCX’s core technology centers on facility-scale and row-level cooling platforms—including advanced Coolant Distribution Units (CDUs) and Facility Distribution Units (FDUs)—engineered for very high flow rates and elevated supply temperatures, aligning with next-generation GPU platforms. Its portfolio has been positioned to support modern accelerator architectures, including NVIDIA rack-scale systems, with recent milestones highlighting second-generation FDU designs capable of multi-megawatt heat-transfer capacity and 45 °C warm-water operation. The company is led by an engineering-driven management team, with CTO Maciek Szadkowski frequently representing DCX’s system-level thermal design approach in industry briefings.