Vertiv Liebert CRV CRD25 In-Row Cooling Unit

The Vertiv Liebert CRV CRD25 is a 25 kW in-row cooling unit purpose-built to address the extreme thermal densities generated by modern AI accelerator clusters, high-performance computing nodes, and hyperscale edge deployments. Positioned directly adjacent to heat-generating racks, the CRD25 delivers targeted, close-coupled cooling that dramatically reduces the distance hot air must travel before being captured and recirculated, enabling precise control over rack inlet temperatures even under sustained full-load GPU and CPU operation. The unit integrates variable-speed EC (electronically commutated) fan arrays that modulate airflow continuously in response to real-time thermal demand, minimizing energy consumption while maintaining critical temperature setpoints without manual intervention.

The CRD25 employs a direct-expansion (DX) refrigeration circuit with an advanced electronic expansion valve and microprocessor-based controls, providing granular capacity modulation from partial load to full rated output. Its compact in-row form factor occupies standard 600 mm or 700 mm deep aisle footprints and fits within 19-inch rack row alignments, making it compatible with existing data center infrastructure without requiring raised floor plenum access or extensive civil modification. Intelligent sensors distributed across supply and return air paths continuously feed data to the onboard Liebert iCOM control platform, enabling coordinated multi-unit operation, adaptive setpoint management, and seamless integration with DCIM and BMS environments via Modbus RTU, SNMP, and BACnet protocols.

Launched in Q4 2025, the CRD25 directly targets the rapidly evolving landscape of AI-accelerated infrastructure where rack power densities routinely exceed 30–50 kW per rack. Its variable capacity architecture allows facilities teams to right-size cooling output dynamically as workloads shift between inference and training cycles, preventing the energy waste associated with fixed-capacity legacy units. The unit supports hot-aisle or cold-aisle containment configurations and can be deployed in N+1 or 2N redundancy arrangements to meet carrier-grade and Tier III/IV uptime requirements. Remote firmware updates, predictive maintenance alerts, and integration with Vertiv's Environet Alert platform round out a system designed for lights-out, automated data center operations.

Ideal for

• Cooling GPU-dense AI training clusters where individual racks sustain 30–50 kW continuous thermal loads from NVIDIA or AMD accelerator nodes
• Supplemental in-row cooling in retrofitted legacy data centers transitioning to high-density compute without full raised-floor overhaul
• Edge colocation facilities requiring close-coupled precision cooling in space-constrained environments with limited overhead plenum capacity
• HPC research centers running sustained simulation and modeling workloads demanding stable sub-25°C inlet temperatures to prevent thermal throttling
• Financial services trading floors and latency-sensitive infrastructure where consistent thermal conditions are mandatory for deterministic processor performance
• Hyperscale enterprise private cloud expansions scaling AI inference capacity rapidly within existing data hall footprints using containment architectures

Technical specifications

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