When Retrofit Makes Sense—and When It Does Not
A retrofit can accelerate deployment when the building has sufficient utility power, structural capacity, heat-rejection potential, service access, and a practical piping route. A new build can be preferable when multiple constraints must be replaced at once or when the long-term rack-density roadmap exceeds the existing site's electrical and mechanical envelope.
Hybrid cooling is often the transition path: cold plates remove the highest-density processor heat while the existing air system manages residual server and room loads. This can reduce the size of immediate facility changes, but it requires clear responsibility between liquid and air systems under every operating mode.
Start With a Site Capacity Audit
| System | Questions to answer |
|---|---|
| Electrical | Available utility, transformer, switchgear, UPS, battery, busway, rack feed, fault current, and redundancy |
| Structure and layout | Floor loading, rack footprint, CDU placement, piping corridors, containment, service and removal paths |
| Heat rejection | Existing chiller or condenser capacity, water temperatures, dry-cooler feasibility, seasonal limits, redundancy |
| Operations | Maintenance windows, leak response, isolation, spares, monitoring, staffing, and training |
| Controls | BMS/DCIM integration, alarms, sensor standards, fail-safe behavior, cybersecurity, and trend data |
| Water and chemistry | Loop materials, treatment, filtration, sampling, fill, drain, flushing, and discharge requirements |
Common Retrofit Architectures
Rack CDU With Existing Facility Water
A CDU near or inside the rack isolates the technology loop from facility water. This can simplify server-side water quality and pressure control, but consumes rack or aisle space and adds localized service work.
Row CDU Serving Multiple Racks
A row-level CDU can consolidate pumps, filters, and controls. It requires carefully balanced distribution, isolation valves, hose or hard-pipe routing, and a failure-domain strategy.
Hybrid Air and Direct Liquid Cooling
Cold plates capture processor heat while room air handles residual components. Confirm the liquid heat-capture ratio and remaining air load; the existing room system may still need modification.
Dedicated New Technology Cooling Loop
A new loop can support higher temperatures and cleaner separation from legacy systems. It involves more piping and plant work but can provide a clearer expansion path.
Integration Risks to Resolve Early
- Temperature mismatch: legacy facility water may be too cold, too warm, or unstable for the proposed CDU mode.
- Pressure mismatch: static and transient pressures can exceed rack-side component limits without proper separation and control.
- Material compatibility: mixed metals, coolant chemistry, oxygen ingress, and legacy contamination can create corrosion.
- Condensation: cold surfaces require dew-point control, insulation, and alarm strategy where temperatures fall below ambient dew point.
- Hydraulic imbalance: long branches, elevation, valves, filters, and connectors affect available flow.
- Service conflict: hoses or headers can block rack removal, cable paths, egress, or maintenance.
- Failure domains: a shared CDU or header can concentrate impact unless isolation and redundancy are designed.
Connection and sealing choices are central to retrofit serviceability. Review manifolds and UQD/NVQD interfaces and stainless steel corrugated hoses.
A Phased Deployment Plan
- Requirements: define IT roadmap, rack loads, liquid heat capture, availability, temperatures, flow, and ownership.
- Survey and model: audit site constraints; complete thermal, hydraulic, electrical, structural, and financial studies.
- Pilot: deploy representative racks and instrument temperatures, pressures, flow, energy, water, alarms, and service operations.
- Qualification: validate normal, peak, degraded, maintenance, and emergency states.
- Controlled expansion: add rows or zones using standard interfaces and documented change control.
- Operational review: compare measured performance against design and update maintenance intervals.
Use the air-versus-liquid cooling guide for architecture selection and the cost and energy guide for financial comparison.
Commissioning and Acceptance Tests
Commissioning should verify flushing and cleanliness, coolant concentration, pump rotation, valve position, sensor calibration, alarm routing, flow balance, pressure control, heat-exchanger performance, redundancy transfer, leak detection, drainage, and service isolation. Test at representative heat load where practical.
Failure testing should cover loss of a pump, CDU, dry-cooler fan, control network, sensor, or power source; blocked filters; rapid workload changes; and leak alarms. Record acceptance criteria, responsible parties, evidence, and open items.
Specify Liquid Cooling Components for Retrofit Service
Retrofit hardware must fit real installation constraints. Manifold port direction, hose bend radius, connector access, mounting datums, valve clearance, drain points, and tool access are as important as nominal flow.
Drawings for cold plates, manifolds, connector bodies, and valve blocks should define materials, coatings, wetted surfaces, threads, sealing grooves, cleanliness, leak rate, proof pressure, flow or pressure-drop acceptance, and traceability. Shengqi provides drawing-based review through its AI server liquid cooling component machining service.
Frequently Asked Questions
Can an existing air-cooled data center be converted to liquid cooling?
Often yes, using rack or row CDUs and a hybrid approach. Feasibility depends on power, structure, facility water, heat rejection, piping routes, controls, service access, and operations.
Does a liquid cooling retrofit eliminate room air cooling?
Not necessarily. Hybrid servers may still reject heat from memory, power supplies, networking, storage, and other components to air.
Where should the CDU be placed in a retrofit?
Rack, row, or facility placement depends on space, capacity, failure domains, piping, service access, noise, redundancy, and the available facility-water loop.
What should a liquid cooling pilot measure?
Measure temperatures, flow, pressure, pressure drop, energy, water where applicable, alarms, thermal stability, residual air load, service time, leak response, and performance during failures.