Formula sheet

Data Center Power and Cooling Formula Sheet

Data center formulas for IT load, three-phase power, UPS autonomy, PUE, cooling load, airflow, liquid cooling flow, pumping power, heat exchangers, and validation.

Branch: Energy Engineering
Content: Formula sheet
Updated: Jun 20, 2026
Revision: v1.1.0 · reviewed

This formula sheet collects first-pass calculations for data center power and cooling engineering. Use it for screening, design review, commissioning checks, and troubleshooting. Detailed design still requires equipment data, electrical studies, heat-transfer models, controls review, safety review, code compliance, and measured operating data.

State the boundary before using any metric. A rack, data hall, modular unit, building, or full campus will produce different power, cooling, and efficiency results.

IT Load

Total IT power:

P_{IT}=\sum_i P_i

where $P_i$ is the measured or estimated electrical power of each server, accelerator, storage, network, or other IT device.

Energy over a time interval:

E=\int_{t_1}^{t_2}P(t)\,dt

For constant power:

E=Pt

Do not substitute nameplate power for operating power unless the purpose is worst-case capacity screening. Nameplate values can overstate normal operation, but average values can understate peak and failover conditions.

Design Load Cases

Design IT load should be tied to an operating case:

P_{design}=P_{measured,peak}f_{growth}f_{coincidence}f_{reserve}

where $f_{growth}$ accounts for committed expansion, $f_{coincidence}$ accounts for the fraction of loads expected to peak together, and $f_{reserve}$ accounts for required engineering margin. These factors should not be hidden inside one generic safety factor.

Installed capacity and usable capacity should be separated:

P_{usable}=\min(P_{electrical},P_{cooling},P_{space},P_{network},P_{controls})

where the limiting term may be busway capacity, UPS capacity, cooling capacity, rack airflow, liquid flow, floor loading, network path, or control constraint. A room can have enough nominal megawatts but still lack usable capacity for high-density racks.

Rack Power Density

Rack power density by rack:

\displaystyle D_{rack}=\frac{P_{rack}}{N_{rack}}

where $P_{rack}$ is rack IT power and $N_{rack}$ is one rack. In practice this is usually reported as kW per rack.

Area-based power density:

\displaystyle D_A=\frac{P_{IT}}{A}

where $A$ is the data hall or white-space area included in the boundary.

Rack density is often more important than average room density. The same total IT load can be easy or difficult to cool depending on how concentrated it is.

For planning, report a rack-density distribution rather than only an average:

D_{max}=\max(D_{rack,i})

\displaystyle D_{avg}=\frac{\sum_i P_{rack,i}}{N_{racks}}

The maximum and distribution matter because local inlet temperature, coolant flow, busway taps, floor loading, and service procedure are driven by the densest racks, not the room average.

Facility Power and PUE

Power usage effectiveness:

\displaystyle PUE=\frac{P_{facility}}{P_{IT}}

where $P_{facility}$ includes the defined facility boundary and $P_{IT}$ includes IT equipment power.

Facility overhead power:

P_{overhead}=P_{facility}-P_{IT}

Overhead fraction relative to IT load:

\displaystyle f_{overhead}=\frac{P_{facility}-P_{IT}}{P_{IT}}=PUE-1

PUE is not a complete measure of engineering performance. It does not directly measure computing efficiency, water use, carbon intensity, redundancy, heat reuse, or resilience. Always state the metering boundary and load condition.

Three-Phase Electrical Power

For a balanced three-phase AC load:

P=\sqrt{3}V_L I_L PF

where $P$ is real power, $V_L$ is line-to-line voltage, $I_L$ is line current, and $PF$ is power factor.

Apparent power:

S=\sqrt{3}V_L I_L

Reactive power:

Q=\sqrt{S^2-P^2}

These formulas support early checks, but final electrical design also requires short-circuit analysis, protection coordination, voltage drop, grounding, harmonic review, thermal ratings, arc-flash review, and local code compliance.

UPS Autonomy

Approximate UPS autonomy from usable stored energy:

\displaystyle t_{UPS}=\frac{E_{usable}\eta_{UPS}}{P_{load}}

where $E_{usable}$ is usable stored energy, $\eta_{UPS}$ is discharge and conversion efficiency, and $P_{load}$ is supported load.

Required usable energy:

\displaystyle E_{usable}=\frac{P_{load}t_{required}}{\eta_{UPS}}

Battery aging, temperature, discharge rate, minimum state of charge, inverter limits, maintenance bypass, and redundancy requirements can reduce usable autonomy. UPS runtime should be verified with manufacturer data and commissioning tests.

Cooling Load

Nearly all IT electrical power becomes heat. A simplified cooling-load balance is:

\dot{Q}_{cooling}\approx P_{IT}+P_{room,aux}+P_{losses}-\dot{Q}_{recovered}

where $P_{room,aux}$ includes auxiliary electrical loads that reject heat into the cooled boundary, $P_{losses}$ includes power-conversion and distribution losses inside the cooling boundary, and $\dot{Q}_{recovered}$ is useful heat removed from that boundary for reuse.

Unit conversion:

1\ \text{kW}=3412\ \text{Btu/h}

Use consistent boundaries. Transformer losses outside the data hall may still affect facility energy, but they do not necessarily add heat to the data hall cooling load.

Air Cooling Flow

Sensible heat removal by air:

\dot{Q}=\rho C_p \dot{V}\Delta T

where $\rho$ is air density, $C_p$ is specific heat capacity, $\dot{V}$ is volumetric airflow, and $\Delta T$ is air temperature rise across the load.

Required airflow:

\displaystyle \dot{V}=\frac{\dot{Q}}{\rho C_p\Delta T}

This is a screening calculation. Actual airflow design must account for fan curves, pressure drop, rack bypass, recirculation, leakage, containment, filtration, altitude, humidity, and server inlet temperature limits.

Liquid Cooling Flow

Single-phase liquid heat removal:

\dot{Q}=\dot{m}C_p(T_{out}-T_{in})

Mass flow required:

\displaystyle \dot{m}=\frac{\dot{Q}}{C_p\Delta T}

Volumetric flow:

\displaystyle Q_v=\frac{\dot{m}}{\rho}

where $Q_v$ is volumetric flow and $\rho$ is coolant density. Use fluid properties at the expected temperature and concentration. Water-glycol mixtures, dielectric coolants, and inhibited water do not have identical properties.

Heat Flux

Heat flux:

\displaystyle q''=\frac{\dot{Q}}{A}

where $A$ is the heat-transfer area. High heat flux at a chip, cold plate, or power device can control the design even when total rack power appears manageable.

Do not compare heat flux values without stating the area basis. Chip area, package area, cold-plate contact area, and rack footprint area describe different engineering problems.

Pumping Power

Hydraulic pumping power:

P_{hyd}=\Delta p Q_v

Electrical pump input:

\displaystyle P_{pump}=\frac{\Delta p Q_v}{\eta_{pump}}

where $\Delta p$ is pressure rise, $Q_v$ is volumetric flow, and $\eta_{pump}$ is pump efficiency.

Pressure drop depends on flow rate, pipe diameter, fittings, valves, filters, manifolds, quick disconnects, cold plates, and coolant viscosity. Higher flow can improve temperature rise but increase pumping power and balancing difficulty.

Heat Exchanger Balance

For a liquid-to-liquid heat exchanger with negligible external loss:

\dot{Q}_{hot}+\dot{Q}_{cold}=0

Magnitude of heat transfer:

\dot{Q}=\dot{m}_h C_{p,h}(T_{h,in}-T_{h,out})

\dot{Q}=\dot{m}_c C_{p,c}(T_{c,out}-T_{c,in})

The hot-side and cold-side heat rates should match within measurement uncertainty and heat-loss assumptions. A mismatch can indicate sensor error, incorrect flow data, fouling, bypass, heat loss, or incorrect fluid properties.

LMTD Check

Overall heat-transfer equation:

\dot{Q}=UA\Delta T_{lm}

Log-mean temperature difference:

\displaystyle \Delta T_{lm}=\frac{\Delta T_1-\Delta T_2}{\ln(\Delta T_1/\Delta T_2)}

Use the correct terminal temperature differences for the exchanger arrangement. Multipass and crossflow arrangements may require correction factors. Fouling and approach temperature should be included in design review.

Cooling Plant COP

Cooling coefficient of performance:

\displaystyle COP=\frac{\dot{Q}_{cooling}}{P_{input}}

where $\dot{Q}_{cooling}$ is useful heat removed and $P_{input}$ is the electrical input for the defined boundary.

If the boundary includes pumps, fans, controls, water treatment, or auxiliary loads, the reported COP will differ from a chiller-only COP. State the boundary whenever COP is reported.

Temperature Margin

Component temperature margin:

M_T=T_{limit}-T_{measured}

Thermal alarms should not be based only on central supply temperature. Local inlet temperature, coolant flow, cold-plate performance, server telemetry, and junction-temperature estimates may reveal risk earlier.

Measurement Reconciliation

Electrical-to-thermal reconciliation:

P_{IT}\approx \dot{Q}_{air}+\dot{Q}_{liquid}+\dot{Q}_{stored}+\dot{Q}_{loss}

For steady operation with small storage and losses:

P_{IT}\approx \dot{Q}_{air}+\dot{Q}_{liquid}

This comparison is useful during commissioning. If measured IT power and measured heat removal disagree strongly, review boundaries, flow meters, temperature sensors, air mixing, calibration, and transient storage in equipment and coolant.

Worked Air-Liquid Reconciliation

Suppose a data hall has:

P_{IT}=1800\ \text{kW}

Measured liquid heat removal is:

\dot{Q}_{liquid}=980\ \text{kW}

Measured air-side heat removal is:

\dot{Q}_{air}=760\ \text{kW}

The measured heat removal is:

\dot{Q}_{measured}=980+760=1740\ \text{kW}

The mismatch relative to IT power is:

\displaystyle \epsilon=\frac{1800-1740}{1800}=0.033

A 3.3 percent mismatch may be acceptable if it is within measurement uncertainty, thermal storage, and boundary assumptions. A larger mismatch should trigger review of airflow estimates, coolant flow meters, temperature sensors, auxiliary loads, liquid-captured fraction, and whether all heat paths are inside the selected boundary.

Validation Notes

Useful quantitative checks include:

IT power versus measured cooling load.
Rack power density versus cooling architecture.
UPS autonomy versus supported load and battery condition.
Airflow or liquid flow versus measured temperature rise.
Pumping power versus pressure drop and flow.
Heat-exchanger hot-side and cold-side balance.
PUE trend versus IT load and operating mode.
Temperature margin during normal, peak, and degraded operation.
Protected reserve and staged restart assumptions during utility or cooling disturbance.
Air-liquid heat split during representative workload changes.

Every formula in this sheet is boundary-dependent. The calculation is useful only when the measured quantities refer to the same physical system, time interval, and operating state.

REF

Disciplines