Cooling, matched
to your environment.

“We have a new AI rack and facility water. Which CDU and rack connections should we specify?”

You are installing liquid-cooled servers and have a building cooling-water connection. You need to match the server loop to the facility without mixing the two fluids.

• What is the maximum liquid-side heat load and the supported server coolant temperature?
• What facility supply/return temperatures, flow and pressure are available?
• Which coolant, connector series and redundancy level does the server platform require?

65 / 130 kW rack CDU families provide a starting point for evaluation—not an automatic match to a rack’s electrical rating.

“We want to test two liquid-cooled servers without adding building-water pipework.”

You need to evaluate a small number of liquid-cooled servers in a laboratory or IT room that has air conditioning but no facility-water connection.

• What are the server model, maximum liquid-side load and permitted coolant temperatures?
• What is the worst-case room temperature and verified spare air-cooling capacity?
• What rack space, airflow clearance and noise limits apply?

8 / 12 / 18 kW rack CDU classes suit initial product discussions. Actual duty depends on room-air temperature and coolant conditions.

“We are adding higher-density servers to a working room and need a phased cooling upgrade.”

New compute equipment is creating a cooling bottleneck in an existing room. Some servers may remain air-cooled while others support direct-to-chip cooling.

• Which racks and servers remain air-cooled, and which are approved for liquid cooling?
• What water, electrical capacity, rack dimensions and maintenance clearances are available?
• What downtime is acceptable, and how will installation be phased?

Rack, rear-door and 35 / 55 kW hybrid cabinet families support different retrofit approaches. Validate the heat split before selecting a configuration.

“We need a modular AI installation at a hot, dusty site. Can the cooling maintain the server’s required inlet temperature?”

You need a packaged compute installation at a new site. Outdoor temperature, humidity, dust and utility availability will shape the cooling design.

• What are the location, design dry-bulb/wet-bulb temperatures, dust and corrosion exposure?
• What grid power, water, footprint and external cooling space are available?
• Which compute platform, availability target, local requirements and commissioning scope apply?

20 ft and 40 ft module concepts provide packaging options. Useful IT capacity depends on the compute platform, site cooling duty and electrical design.

“We have a multi-rack AI cluster planned in phases. How should we size the CDUs and distribute coolant?”

You are deploying several liquid-cooled racks and need a coordinated approach to coolant distribution, shared capacity, redundancy and future expansion.

• What are the maximum liquid-side load per rack, total load and expansion phases?
• What capacity must remain available during planned maintenance or a component failure?
• What facility duty, pressure-drop allowance, monitoring and leak-response requirements apply?

500 / 800 / 1,000 / 1,500 / 2,000 kW floor-standing CDU classes are available for evaluation at validated operating conditions.

“We want to verify temperature control, flow and load response before our AI servers are delivered.”

You need to test components, validate a coolant loop or commission installed cooling before connecting production servers.

• What load profile, temperature stability and flow/pressure range must be tested?
• What is the verified cooling-source capacity and which coolant will be used?
• What cleanliness, venting, leak-detection, alarm and fault-response tests form the acceptance criteria?

40 / 50 kW load-bank classes, a 6U / 12 kW test CDU and specialist test equipment cover different duties; they are not automatically a matched set.