US Liquid Cooling editorial · Reviewed · 3 min read
A photograph makes the difference look simple: one server has hoses, and another sits in a tank. For an operator, the more useful question is what changes on the day a server needs service.
Direct-to-chip cooling and immersion cooling both use liquid to carry heat away from computing equipment. They place that liquid in different relationships with the hardware, which affects much more than the cooling diagram.
Where the liquid goes
In direct-to-chip cooling, cold plates transfer heat from selected components into a contained coolant circuit. Other components may still need airflow, depending on the server design and how much of its heat is captured by liquid.
Immersion systems place suitable hardware in an electrically insulating fluid. Single-phase designs keep that fluid liquid during normal operation; two-phase designs use boiling and condensation as part of the heat-transfer process. Those architectures require their own supported equipment and operating procedures.
Vertiv's technical material describes both approaches. Neither description is a reason to assume that an ordinary air-cooled server can be converted without manufacturer approval.
Start with supported hardware
Ask the server supplier for a supported configuration and the cooling supplier for documented compatibility. For a cold-plate system, that includes the server circuit, connections, approved coolant, and operating envelope. For immersion, include the complete system and fluid compatibility of the equipment that will be exposed.
An open question should stay visible in the procurement record. A vague statement that equipment is immersion-ready is less useful than a written description of what has been qualified, by whom, and under which conditions.
Walk through one repair
For direct-to-chip, ask how technicians isolate and disconnect a server, manage any fluid at the connections, and confirm normal operation after reconnection. Determine who is responsible for leaks at the boundary between IT and facility equipment.
For immersion, ask the supplier to demonstrate removal, handling, fluid management, and return to service. Include access space, tools, staff training, and the storage or disposal procedures for the chosen fluid.
These questions do not prove that either approach is difficult. They show whether a proposal includes the work your team will actually perform.
Follow the heat beyond the rack or tank
Both approaches need a complete heat-rejection plan. Identify how heat moves from the IT circuit to the facility and then outdoors or to a useful heat consumer. A tank or cold plate is only one part of that route.
Ask what air cooling remains necessary, what facility connections are required, and how performance changes across expected outdoor conditions. Compare proposals at the same workload and availability target. A capacity figure without its operating conditions is an incomplete comparison.
Request a realistic cost boundary
Include supported servers, cooling equipment, building work, electrical supply, controls, training, consumables, maintenance, and downtime assumptions. Require suppliers to state exclusions. A lower equipment quotation may simply cover less of the project.
For an existing site, a useful next step is a scoped pilot with agreed acceptance criteria. Measure the intended workload and test the service workflow, rather than selecting a technology solely from a demonstration of peak cooling capacity.
The strongest choice is the one your hardware, facility, and operating team can support together. Begin with those constraints and let the cooling architecture follow.
Sources & further reading
- Vertiv: Liquid cooling options for data centers ↗
- Vertiv: Understanding direct-to-chip cooling ↗
- Vertiv: Deploying liquid cooling white paper ↗
Explore the company directory → · All liquid-cooling guides →