Vertiv CoolPhase Liquid Cooling CDU 1.35MW

The Vertiv CoolPhase Liquid Cooling CDU 1.35MW is a high-capacity coolant distribution unit designed to serve as the thermal backbone of next-generation AI data centers and HPC installations. Utilizing a liquid-to-liquid heat exchange architecture, the unit accepts facility-side chilled water and conditions it to precise supply temperatures required by direct-to-chip cold plates fitted to accelerators such as NVIDIA H-series and AMD Instinct GPUs. The system is engineered to handle aggregate thermal loads of up to 1.35 megawatts within a single frame, enabling data center operators to densify GPU racks far beyond what conventional air cooling permits — supporting rack densities commonly exceeding 100 kW per rack when deployed in conjunction with compatible manifold distribution systems.

At its core, the CoolPhase CDU 1.35MW incorporates variable-speed primary and secondary pump assemblies with redundant configurations, intelligent flow balancing across multiple rack-level distribution manifolds, and a built-in controls platform that communicates over SNMP, Modbus TCP, and BACnet for full integration into building management and DCIM ecosystems. The precision control system maintains coolant supply temperature within ±0.5 °C of setpoint, preventing condensation on facility infrastructure while delivering optimal thermal performance to sensitive silicon. The unit is offered in an ISO-standard 19-inch-compatible floor-standing enclosure with side-accessible service panels, reducing maintenance downtime in live production environments.

Launched in Q1 2026, the CoolPhase CDU 1.35MW targets hyperscale colocation operators, national AI research consortiums, financial services HPC clusters, and enterprise organizations deploying large-scale GPU fabrics for generative AI training and inference. Its modular internal architecture allows future capacity expansion and component replacement without full system downtime, providing a forward-looking thermal infrastructure investment aligned with the continued scaling trajectory of AI accelerator thermal design power.

Ideal for

• Large-scale generative AI training clusters deploying hundreds of direct-liquid-cooled GPU nodes requiring centralized, high-capacity coolant distribution at megawatt scale
• Hyperscale colocation facilities retrofitting existing data hall rows with direct-to-chip cooling to accommodate tenants running GPU-dense workloads beyond 50 kW per rack
• National laboratory and university supercomputing centers building petascale HPC clusters where air cooling infrastructure cannot meet aggregate thermal load requirements
• Financial services firms running real-time risk modeling and quantitative simulation workloads on GPU clusters that demand deterministic, low-latency thermal management
• Sovereign AI infrastructure projects requiring high-reliability, redundant liquid cooling distribution with deep integration into building automation and DCIM platforms
• Edge AI inference hubs in compact, high-density facilities where a single CDU must serve multiple GPU server rows with minimal physical footprint

Technical specifications

Available from Omnixon Global. Submit an RFQ and our team will confirm configuration and availability for your order.