Specifications
Product Type
Coolant Distribution Unit (CDU) portfolio
Max Cooling Output
1 MW (CDU In-Row)
In-Rack CDU
150 kW, compact 4U form factor
Rear-Door Solution
70 kW Rear Door HEx with Coolant Circulation Unit
Portfolio Range
70 kW to 1 MW
Architecture
Liquid-to-liquid (L2L) and liquid-to-air (L2A)
Per-Drawer Rating
~250 kW heat removal
Row Scaling
250 / 500 / 750 / 1 MW per row
Redundancy
N+1 spare module for service
Rack Standard
OCP Open Rack V3 / 21-inch rack designs
Cooling Method
Direct-to-chip (DLC), water-based
Coolant Path
Technology water loop + facility water secondary
Manifolds
Bespoke distributor channels for flexible fluid routing
Serviceability
Hot-serviceable sensors, controller and pumps
Application
Hyperscale GPU clusters and AI factories
Overview
Rittal's direct liquid cooling portfolio spans three product shapes built for the same problem — pulling heat off dense AI silicon at the chip rather than the room. The entry point is a 70 kW rear-door heat exchanger with a Coolant Circulation Unit for facilities without a water connection; the mid point is a 150 kW liquid-to-liquid CDU packed into a compact 4U in-rack chassis; the flagship is a 1 MW in-row CDU.
The megawatt-class unit was engineered around Open Rack V3 and Open Compute Project principles, with modularisation borrowed from server handling. Each rack-format drawer is rated at roughly 250 kW of heat removal, so operators deploy 250, 500, 750 kW or a full megawatt per row with an N+1 spare module available for service without taking the row down.
Because it is designed on OCP ORV3 form factors and GPU sleds, the unit drops into existing OCP racks. Technology water circulates through supply and return manifolds to processor-level cold plates, while facility water carries the heat away on the secondary side.
Key Benefits
Megawatt class in an OCP rack — 1 MW per row built from 250 kW modular drawers. Three architectures, one vendor — liquid-to-air rear door (70 kW), 4U in-rack (150 kW) and in-row (1 MW) match the facility's water availability. Modular N+1 spare drawer lets maintenance happen without capacity loss. Hot-serviceable sensors, controller and pumps maintain constant cooling during service. OCP ORV3 and 21-inch rack alignment means no custom racking project.
Applications
Hyperscale GPU clusters built on OCP Open Rack, AI factory deployments above 100 kW per rack, colocation halls with facility water available, and mixed-density data centers where only part of the floor can move to liquid cooling.
Request a Quote — RITTAL CDU IN-ROW — 1 MW LIQUID-TO-LIQUID COOLANT DISTRIBUTION UNIT
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →