Sizing Backup Generators for Liquid-Cooled Data Centers

Updated 9 November 2026 · Standard formula, shown below

This guide helps you estimate the standby generator capacity needed to support a liquid-cooled data center during a utility outage. Gather measured or designed electrical loads, cooling-system requirements, and equipment starting data first; the result is a preliminary sizing basis, not a substitute for a coordinated electrical and mechanical design.

What You’ll Need

A current one-line electrical diagram and load schedule, including IT equipment, UPS losses, cooling, pumps, controls, lighting, and other essential loads.
Cooling-system design documents: heat-rejection method, pump and fan ratings, control sequence, redundancy, and minimum operating conditions.
Equipment nameplates or manufacturer data for running kW, kVA, power factor, motor starting current, and permissible voltage and frequency dips.
UPS and power-distribution specifications, including transfer behavior, recharge settings, and any generator compatibility requirements.
Applicable codes, authority requirements, site conditions, and a qualified electrical engineer or generator specialist to validate the design.

Step-by-Step

Define what must stay online. Set the outage operating objective: full IT load, selected halls, or a staged critical-load plan. Identify required autonomy and the loads that must start immediately versus those that can wait. Include safety systems and controls needed to keep the facility stable, not just the servers.
Draw the complete generator-backed load boundary. Trace the supply path from generator through switchgear, UPS, distribution, and cooling plant. List every load on that path. Account for UPS input power rather than only its output: include conversion losses and battery charging behavior. Separate loads that are already included in an IT power figure so they are not counted twice.
Establish realistic operating load. Use measured demand where available, preferably at representative high utilization, and reconcile it with equipment design data. Record kW and power factor, then determine kVA when necessary: kVA = kW / power factor. Do not treat a nameplate maximum as a normal demand or assume a low average load will persist during an outage.
Calculate cooling and heat-rejection demand. Liquid cooling removes heat from IT equipment, but it does not eliminate the electrical load required to move and reject that heat. Include coolant distribution unit (CDU) pumps, facility-water pumps, dry coolers or cooling-tower fans, chillers if used, valves, controls, and any heat-exchanger auxiliaries. Use the specified emergency operating mode and account for seasonal or design-ambient conditions. Confirm which components are essential for each supported IT load.
Build a load table by operating state. For each generator-backed item, document running kW, kVA, power factor, starting method, start current or transient kVA, and start priority. Create scenarios such as steady operation, initial recovery after transfer, and the largest credible equipment failure. Include redundant equipment only when the control sequence could run it at the same time; do not simply add every installed standby component without checking operation.
Check motor starts and load steps separately from steady-state capacity. Pumps, fans, and compressors can impose brief high current and cause voltage or frequency dips. Identify the largest credible start and whether motors start across the line, through a soft starter, or with a variable-frequency drive. Review the generator manufacturer’s transient performance data and the driven equipment’s limits. A generator that meets total running kW may still fail a motor-starting or UPS step-load requirement.
Choose a safe sequencing strategy. Specify the order and delay for restoring loads after transfer. A typical strategy may bring controls and essential cooling online first, then add UPS or IT blocks and remaining cooling equipment in controlled steps. Coordinate the sequence with the UPS: abrupt load pickup, battery recharge, or nonlinear rectifier loading can affect generator stability. Confirm the sequence works for both normal operation and a failed component.
Account for site and rating conditions. Select the applicable standby rating for the intended duty and operating hours; do not confuse standby, prime, and continuous ratings. Apply manufacturer-supported derating for altitude, ambient temperature, ventilation, fuel characteristics, and other site conditions. Check both kW and kVA limits, generator power factor capability, and any harmonic or nonlinear-load requirements. Avoid applying generic derating percentages in place of the supplier’s data.
Evaluate redundancy and fuel arrangements. Decide whether the design must meet N, N+1, or another resilience objective, and assess capacity with the largest planned unit unavailable where required. Confirm fuel supply, storage, replenishment assumptions, and starting reliability against the facility’s outage plan and applicable requirements. Multiple smaller sets can improve maintainability but add controls, paralleling, and coordination needs.
Validate the design with integrated testing. Have qualified professionals review calculations, protection coordination, grounding, exhaust, ventilation, noise, and transfer equipment. Test the actual generator, switchgear, UPS, pumps, and controls as an integrated system under representative load steps. Verify voltage and frequency recovery, alarms, cooling flow and temperatures, and failure modes. Update the load schedule and settings after commissioning or major IT and cooling changes.

Common Mistakes

Sizing from IT nameplate power alone while omitting pumps, chillers, heat-rejection equipment, UPS losses, and controls.
Adding kW values and assuming the result is sufficient without checking kVA, power factor, harmonics, and transient performance.
Assuming liquid cooling means cooling power is negligible, or counting every installed redundant component as if all must run simultaneously.
Ignoring motor-starting current, UPS recharge, load-step size, or the recovery sequence after a transfer.
Using a generator’s headline rating without checking duty rating and site derating, or relying on a generic oversizing factor.
Failing to retest after IT density, cooling topology, firmware, or control-sequence changes.

Tips

Keep separate totals for steady kW, steady kVA, largest start, and largest load step. They answer different sizing questions.
Use operating measurements to challenge assumptions, but retain a documented design allowance for known growth and uncertainty rather than an arbitrary blanket margin.
Model the cooling system at the actual emergency mode, including the consequences of losing one pump, CDU, or heat-rejection unit.
Coordinate generator, UPS, drive, and switchgear suppliers early; their acceptable voltage and frequency windows must overlap.
Maintain a controlled load schedule with source, date, assumptions, and owner so future changes trigger a review.

Troubleshooting

Generator starts but voltage or frequency sags: Check whether a large motor or UPS block starts too quickly, whether load steps exceed the set’s transient capability, and whether fuel, air, or engine settings are correct. Review event records with the generator supplier; consider staged starts or revised equipment starting methods.

UPS transfers to battery or rejects generator power: Compare generator voltage and frequency behavior with the UPS acceptance window. Check waveform quality, grounding, neutral arrangement, synchronization, and recharge settings with the respective suppliers. Do not widen protection settings without an approved engineering review.

IT load is stable but coolant temperatures rise: Verify pump rotation, flow, valve positions, CDU operation, heat-exchanger availability, and the heat-rejection system’s outdoor operating limits. Confirm controls receive power and that the programmed sequence does not delay cooling too long.

Measured load differs from the calculation: Check for double-counted or omitted loads, changed power factor, battery charging, seasonal chiller demand, and equipment operating in an unexpected mode. Update the load table, then repeat an integrated test before treating the revised capacity as adequate.

Frequently Asked Questions

Can I size the generator using IT load alone?
No. Include the UPS input demand, cooling and heat rejection, controls, and every other required load on the generator-backed path.

Does liquid cooling reduce generator size?
It may change the cooling load, but it does not remove it. Pumps, CDUs, chillers where present, and heat-rejection equipment still need power.

Should generator capacity be specified in kW or kVA?
Check both. The engine is generally constrained by real power in kW, while the alternator also has a kVA limit; power factor and transient loads affect the selection.

How much extra capacity should I add?
There is no universal percentage. Base capacity on documented growth, site derating, redundancy criteria, load steps, and supplier transient calculations.

Do I need to test with the real data-center load?
Use an approved commissioning plan that verifies representative steady load, load steps, transfer behavior, and cooling response. Testing must be conducted safely by qualified personnel.

How often should sizing be reviewed?
Review it whenever IT capacity, UPS settings, cooling equipment, redundancy, or operating sequences change, and include it in periodic resilience reviews.

Method: the standard formula for this calculation, rounded as each result states. See how the calculators work or report an error.