Liquid Cooling Without Facility Water: Where Does the Heat Go?

Follow the heat beyond the components

In a cold-plate arrangement, selected components transfer heat into circulating coolant. That describes how heat leaves those components. The coolant must also give that heat to something else so the cooling path can continue. The CDU's heat exchanger provides that next transfer.

The two arrangements considered here differ at that exchanger. Liquid-to-air, or L2A, hands the captured heat to air. Liquid-to-liquid, or L2L, hands it to a separate liquid circuit associated with the facility. Vertiv's public CDU explanation distinguishes an air-side heat-exchange coil from an exchanger connected to the facility-water system.

This makes the heat destination a separate buying decision from choosing a liquid-cooled server. A proposal can specify cold plates correctly while leaving unanswered which facility system must accept their heat.

References: [V1]

Liquid-to-air hands the captured heat to the room

For a CDU that rejects into the data hall, heated coolant reaches an exchanger where air receives its heat. Fans move air through the exchanger in the named CoolIT AHx240 example. CoolIT describes the captured server heat entering the data-center air without a facility-water connection. This is a concrete example of the mechanism, rather than a selected product for the hypothetical load below.

Eaton makes the remaining dependency explicit: its L2A description says these units use data-center air conditioning to dissipate the rack or row heat. The absence of a water connection at the CDU therefore does not make the arrangement independent of facility cooling. The room's cooling system must carry the transferred heat onward.

This can still provide a useful component cooling path. Transferring heat from a chip into liquid is a different task from removing heat from the room. L2A puts an additional exchanger between those tasks; it does not make them the same task or establish that the existing room has capacity for an added load.

References: [C1][E1]

Liquid-to-liquid hands the captured heat to facility water

In the L2L arrangement described by Eaton, coolant carries heat from the IT equipment to an exchanger inside the CDU, which transfers it to facility water. CoolIT's CDU overview explains that its internal liquid-to-liquid exchanger separates the technology cooling system from the facility water. Heat crosses the exchanger; the two fluid circuits remain separate.

The captured component heat consequently reaches the designated water interface instead of being discharged into room air by an L2A exchanger. That distinction concerns the captured heat path. Any components or other equipment still rejecting heat directly to air retain their own room-cooling requirement.

The facility-water interface is another transfer boundary, not proof that heat has already left the building. The facility system beyond it must accept the heat and carry it onward to its actual rejection or use destination. A CDU-only offer does not establish that downstream system's scope or capacity.

References: [E1][C2]

Trace 18 kW through the CDU and keep 2 kW separate

Consider an original hypothetical example, independent of the named products above. Stipulate that selected components send 18 kW into a technology cooling loop and another 2 kW directly into room air. These are supplied assumptions, not a measured heat split or values inferred from a GPU model.

In the hypothetical L2A arrangement, the CDU transfers all of that stipulated 18 kW into the same room air. The room therefore receives 18 kW through the exchanger plus 2 kW directly: a 20 kW contribution from the specified components. A planning note that lists only the direct 2 kW as the room's component contribution has omitted the 18 kW arriving through the CDU.

In the hypothetical L2L arrangement, the stipulated 18 kW reaches the designated facility-water interface and the direct 2 kW still reaches room air. The same 20 kW component contribution now follows two receiving paths. The 18 kW and 2 kW are parts of that total; adding them to the total again would count the component heat twice.

Actual CDU auxiliary heat and other equipment loads are additional unknowns. No auxiliary-loss value is invented or assumed to be zero. This example illustrates heat paths only: it is not a product rating, facility-load calculation or assessment of available cooling. It shows why moving heat into liquid does not, by itself, tell the buyer which facility system ultimately receives it.

Buy the receiving interface you actually need

Start the comparison with the intended result. If the objective is to serve liquid-cooled components without a facility-water connection at the CDU, an L2A arrangement offers that type of interface. Its room-air destination still needs to be supported. If the objective is to move the captured heat away from room air into a facility-water system, an L2A unit discharging into that room does not accomplish that particular objective.

Conversely, water being present in the building does not establish a usable L2L arrangement for the offered equipment. The supplier's actual unit and the customer's appointed thermal and facility parties must establish the applicable operating conditions and receiving system. Product capability and the site's available capacity are separate inputs.

Compare what each offer includes beyond the server: the identified CDU, its receiving medium and destination, the supporting facility scope and the parties returning the applicable equipment and facility requirements. Two offers saying liquid cooling included can stop at different boundaries. A self-contained liquid circuit can still depend on room cooling supplied outside that offer.

An L2A proposal described as a bridge to later facility water also needs two distinct configurations identified. A future L2L heat path changes the receiving interface; the bridge label does not establish that the same CDU can be converted or that the later facility package is included. Keep the proposed future arrangement visible without treating it as an available upgrade.

Once the heat destination is clear, use the related CDU capacity guide for the separate question of reading a rating under its stated conditions. A larger headline kW figure cannot substitute for the receiving interface the project needs.

References: [E1][C2][V1]

Connect the decision to the physical IT scope

The customer's appointed qualified cooling, OEM and facility parties determine the actual cooling selection, design, implementation and acceptance. Leviathan does not supply a cooling approval or perform, coordinate or subcontract liquid-side work. The cooling responsibility guide explains that boundary; the heat-path comparison here supplies no field procedure.

For the surrounding physical IT package, Leviathan's rack-and-stack service can discuss agreed equipment mounting, IT cabling, labels, physical checks and handoff records under the released scope. The service's installation record addresses those tasks rather than cooling performance.

Use the linked rack-and-stack project inquiry to share the exact IT equipment, proposed rack locations, available installation and cable documents, requested physical tasks and customer-defined cooling prerequisites. Identify the appointed cooling and facility contacts and the intended heat destination. Our team can then discuss the physical scope, requirements and availability.

References: [L1][L2]