A liquid-cooled data center container packages high-density computing infrastructure and selected support systems in a modular enclosure. The container is one part of the design: IT load, coolant distribution, heat rejection, electrical service, controls, fire protection, and site conditions must be engineered together. This guide identifies the decisions to settle before requesting a proposal or freezing a design.
What belongs inside the project boundary?
List what the container supplier will deliver and what remains with the site contractor. The package may include the enclosure, racks, cold-plate distribution, coolant distribution units (CDUs), manifolds, dry coolers or cooling towers, power distribution, monitoring, fire protection, and environmental controls. Specify whether servers, network equipment, UPS and batteries, external piping, transformers, and installation are included. Compare proposals only after their equipment schedules and exclusions use the same boundary.
How does heat leave the servers?
In a common direct-to-chip design, cold plates capture heat at CPUs and GPUs. A secondary coolant loop carries it to a CDU, where it transfers to a primary or facility loop. Outdoor heat rejection equipment then releases that heat. The exact arrangement depends on server requirements, climate, water constraints, and the selected equipment.
Document the approved coolant, water quality, supply and return temperatures, flow, pressure limits, connection standards, filtration, and leak detection. Confirm these values with the server and CDU manufacturers. Read how cold plates, CDUs and dry coolers work for the equipment interfaces.
What load should the cooling system be sized for?
Use the approved server configuration and expected operating load. Record the number of racks, typical and design power per rack, the portion of heat captured by liquid, heat remaining in the room, auxiliary loads, and planned growth. Apply project-specific margin and redundancy only after these inputs are agreed.
Proposal figures are estimates until the operating profile, design weather, measurement boundary, and test method are documented. A PUE target is not a measured result. Use the GB300 project design checklist to reconcile rack count, power, and cooling assumptions.
How are power and UPS scope coordinated?
Confirm voltage, frequency, service capacity, feeders, rack distribution, protection, grounding, and monitoring. UPS is a separate scope decision: identify protected loads, required runtime, batteries, bypass, maintenance access, and whether equipment sits inside or outside the container. A proposal identified as without UPS must state the exclusion in its equipment schedule and site interfaces.
Which site conditions change the design?
Provide the ambient temperature and humidity range, altitude, wind and corrosion exposure, local codes, seismic requirements, noise limits, water availability, drainage, access route, lifting plan, and foundation details. Dry coolers, evaporative assist, closed towers, and hybrid systems have different water, energy, maintenance, and peak-weather implications. Compare cabinet-based and containerized architectures against the actual site constraints.
How should redundancy and controls be specified?
State the availability target and which components require N+1 or another redundancy strategy. Clarify coverage for pumps, CDUs, heat rejection, power paths, controls, and communications. Define sensor points, leak response, alarms, remote monitoring, safe shutdown actions, and integration test responsibility. Component redundancy must be checked against the capacity available after one component fails.
What should the design review include?
- Server bill of materials, rack count, rack dimensions, and approved power envelope.
- Coolant, temperatures, flow, pressure, connection standards, and water quality.
- Heat rejection design conditions, water constraints, and redundancy.
- Electrical service, distribution, UPS boundary, and autonomy requirements.
- Monitoring, leak detection, fire protection, and commissioning scope.
- Site access, structure, acoustics, environmental limits, and code requirements.
- Shipping, installation, acceptance tests, documentation, and service expectations.
For a delivery format comparison, see containerized vs. room-based liquid cooling.
Frequently asked questions
What does a liquid-cooled data center container include?
It may include IT racks, liquid cooling distribution, CDUs, heat rejection equipment, power distribution, monitoring, and environmental or safety systems. The approved bill of materials defines the actual scope.
Does a liquid-cooled container always need a cooling tower?
No. Heat rejection equipment is selected for the required capacity, climate, water availability, operating costs, and maintenance model. A dry cooler or another suitable system may be used.
Is PUE fixed for a container design?
No. PUE depends on IT load, climate, operating conditions, included facility loads, and measurement boundaries. A credible value needs a stated design basis and measurement method.
What should be sent for a preliminary review?
Send the server configuration, rack count and load, site climate, electrical service, water constraints, redundancy target, location, desired delivery scope, and project schedule.
Next step
Share the approved IT load and site inputs through the liquid cooling solutions page for a project-specific configuration review.