Leave Your Message
Explore our premium 0U rack PDUs tailored for the demanding thermal and electrical footprints of modern AI server racks.
The global surge in generative artificial intelligence (AI), large language models (LLMs), and high-performance computing (HPC) has fundamentally transformed the data center landscape. Modern AI clusters running NVIDIA Blackwell, Hopper, or AMD Instinct accelerators demand unprecedented levels of electrical power. Rack densities have skyrocketed from a traditional 10kW–15kW up to 100kW or even 150kW per rack.
At these extreme densities, conventional air-cooling systems reach their physical limits. Liquid cooling—primarily Direct-to-Chip (D2C) cold plate cooling and Immersion cooling—has transitioned from an experimental technology to an industrial mandate. However, delivering hundreds of kilowatts of clean, balanced, three-phase power directly to these high-density servers within the tight confines of a liquid-cooled rack requires a specialized power distribution strategy. This is where Vertical Rack PDUs (0U PDUs) become indispensable.
"Liquid cooling handles the heat, but the Vertical Rack PDU manages the heartbeat. Without high-density vertical power distribution, scaling AI clusters to 100kW+ per rack is physically and thermally impossible."
In a high-density AI rack, every millimeter of space is contested. Horizontal rack space (U space) is reserved entirely for heavy GPU servers, switches, and liquid manifold connections. Vertical Rack PDUs mount in the zero-U space at the rear of the cabinet, ensuring that airflow is not obstructed and liquid manifolds remain easily accessible.
Furthermore, liquid-cooled environments present unique installation and operating challenges. The presence of coolant hoses, quick-disconnect valves, and manifold pipes means the PDU must feature an ultra-slim profile. It must also withstand elevated ambient temperatures, as liquid-cooled racks often operate at higher internal temperatures to maximize cooling loop efficiency. Our specialized vertical PDUs are engineered to run reliably at up to 60°C ambient temperatures while maintaining maximum power efficiency.
Three core pillars that define our Vertical Rack PDU technology for liquid-cooled server environments.
Integrating hybrid C39 outlets that combine C13 and C19 functionality into a single receptacle. This provides maximum provisioning flexibility for diverse server configurations without needing to swap PDUs.
Real-time, outlet-level power monitoring and switching. Capture precise metrics for Current, Voltage, Power Factor, and Active Power (kWh) with +/-1% billing-grade accuracy to optimize AI cluster PUE.
Designed with high-grade industrial components and robust sheet metal enclosures to operate seamlessly in high-temperature environments adjacent to liquid-cooled server manifolds.
Scenario A: Direct-to-Chip (D2C) Cold Plate Cabinets
In a D2C setup, liquid coolant is delivered directly to the GPU and CPU cold plates. A network of hoses runs along the rear of the rack. The Vertical Rack PDU must be positioned such that it does not block the installation or maintenance of these hoses. By utilizing an ultra-slim 0U form factor, our PDUs sit flush against the cabinet frame, providing ample clearance for coolant loops and preventing accidental physical damage during hot-swapping of servers.
Scenario B: Rear Door Heat Exchanger (RDHx) Integration
RDHx systems replace the standard rear door of the rack with a liquid-cooled coil. This configuration creates a localized micro-climate of warm exhaust air before it is cooled by the door. The Vertical PDU, mounted right at the rear edge of the rack, is exposed to this warm air zone. Our PDUs are built with high-temperature rated internal wiring and circuit breakers to prevent thermal nuisance tripping under these demanding conditions.
To support 100kW+ racks, datacenters are shifting from single-phase power to high-voltage 3-phase power (such as 415V or 480V input). 3-phase power reduces the required cable thickness and minimizes line losses. Our Vertical PDUs feature balanced phase layouts, distributing the load evenly across all three phases to prevent voltage imbalances that could degrade the lifespan of sensitive AI hardware.
Additionally, software integration via SNMP, Modbus, or Redfish APIs allows automated datacenter infrastructure management (DCIM) tools to communicate directly with the PDU. During peak AI training sessions, if a rack approaches its thermal threshold, the DCIM can dynamically adjust server workloads or trigger automated cooling loop adjustments based on the real-time power metrics reported by the smart PDU.
Welcome To Learn About Our Company
The company operates three specialized business departments to support diversified development.
Take a virtual tour inside our state-of-the-art production lines and testing facilities.
Select from our comprehensive list of intelligent and modular power distribution units to find the perfect fit for your datacenter deployment.
