Starlight ToolsData Center White Space Capacity Planner: Racks, Power and Cooling
Find how many racks a data hall can support before usable floor area, PUE-adjusted electrical capacity, or sensible cooling becomes the binding constraint. Compare that result with a target deployment and see the capacity needed to close a shortfall.
How to use the white-space capacity planner
- Keep one boundary: area, facility power, PUE, cooling, and other heat must describe the same hall, suite, or allocation.
- Set a realistic rack target: use the total installed racks and a diversified design peak in kW/rack. Split conventional and high-density zones when one average would hide local constraints.
- Use gross planning area: include a rack's share of aisles, containment, egress, columns, staging, and layout losses rather than cabinet footprint alone.
- State the redundancy condition: enter only electrical and sensible cooling capacity available in the normal or failure state being tested.
- Preserve headroom explicitly: separate space, electrical, and cooling reserves expose which policy is consuming capacity.
- Review every ceiling: the smallest whole-rack result controls the aggregate plan, but a close second constraint can become limiting after rounding or design changes.
- Validate locally: check PDU branches, A/B failover, cable paths, airflow, liquid distribution, floor loading, fire protection, egress, controls, and restart behavior before deployment.
Formulas and boundary assumptions
Usable planning area: white-space area × (1 − space reserve).
Space rack ceiling: floor(usable planning area ÷ gross area per rack).
Usable facility power: facility power allocation × (1 − electrical reserve).
IT power budget: usable facility power ÷ PUE.
Power rack ceiling: floor(IT power budget ÷ average design kW/rack).
Usable sensible cooling: installed sensible cooling × (1 − cooling reserve).
Cooling IT budget: max(0, usable sensible cooling − other in-room heat).
Cooling rack ceiling: floor(cooling IT budget ÷ average design kW/rack).
Deployable rack capacity: min(space racks, power racks, cooling racks).
The model treats steady-state IT electrical input dissipated inside the selected boundary as sensible heat. PUE is used only to translate whole-facility electrical capacity into an IT power budget; it is not multiplied into white-space heat. One refrigeration ton equals 3.516853 kW. All rack ceilings round down to whole racks.
Worked example: 120 racks at 6 kW per rack
The 10% space reserve leaves
420 × 90% = 378 m².floor(378 ÷ 3) = 126 racks.1,200 × 80% ÷ 1.40 = 685.7 kW IT.floor(685.7 ÷ 6) = 114 racks.900 × 80% − 40 = 680 kW IT heat.floor(680 ÷ 6) = 113 racks.min(126, 114, 113) = 113 racks.The 120-rack target is short by 7 racks.
At the selected policies, the target requires 1,260 kW total facility power and 950 kW installed sensible cooling. The entered 420 m² area already exceeds the 400 m² requirement, so adding floor area alone would not resolve the shortfall.
What the aggregate result does not prove
- Distribution capacity: aggregate kW can pass while switchgear, UPS modules, transformers, PDUs, RPPs, busways, whips, breakers, or A/B paths fail locally.
- Cooling delivery: aggregate sensible capacity does not prove airflow reaches every inlet, liquid cooling removes its stated share, controls coordinate, or capacity survives a unit or pump failure.
- Layout feasibility: average area per rack cannot validate actual aisles, door swings, columns, containment, staging, egress, accessibility, or fire-code requirements.
- Structural suitability: rack count does not check concentrated floor loads, rolling loads, anchorage, seismic restraint, or overhead services.
- Peak and future behavior: average rack kW can miss GPU boost events, synchronized workloads, power-capping behavior, commissioning loads, or phased growth.
- PUE stability: a planning PUE is boundary-, load-, climate-, and time-dependent. Use a value appropriate to the modeled condition, not an unrelated annual marketing figure.
Engineering limit: use this browser tool for screening and scenario comparison only. Final capacity must be based on drawings, equipment schedules and performance curves, metering, failure-mode analysis, applicable codes, manufacturer requirements, and qualified electrical, mechanical, structural, and fire-protection review.
Frequently asked questions
How is maximum rack capacity calculated?
The planner calculates independent whole-rack ceilings for floor area, IT power, and sensible cooling. The smallest ceiling is the modeled deployable capacity because every deployed rack must fit within all three budgets.
What area per rack should I use?
Use a project-specific gross planning allowance: cabinet footprint plus its share of hot and cold aisles, containment, egress, columns, staging, and other layout losses. Cabinet footprint alone overstates capacity.
Why is PUE applied to facility power but not cooling?
PUE is total facility energy or power divided by IT equipment energy or power. It translates a whole-facility allocation into an IT budget. Cooling is independently compared with heat released inside the chosen boundary, so multiplying rack heat by PUE would double count unrelated facility overhead.
Should redundant cooling or power modules be included?
Only include capacity available in the operating or failure state you are planning. For an N+1 failure check, remove the unavailable unit's usable capacity first, then enter the remainder. Repeat the calculation for other credible states.
Does 1 kW of IT load equal 1 kW of heat?
For a steady-state white-space screen, approximately yes when that electrical energy is dissipated inside the cooling boundary. Add in-room lighting, people, and distribution losses separately, and subtract heat removed directly by a verified liquid path when modeling residual air cooling.
What happens when non-IT room heat exceeds usable cooling?
The cooling IT budget becomes zero, so the cooling rack ceiling is zero. That explicit result means the entered cooling system cannot preserve the chosen reserve even before rack IT heat is added.
Why can two constraints have the same rack ceiling?
Each raw capacity is divided by average rack load and rounded down. Different budgets can therefore support the same whole-rack count. Treat every tied constraint as limiting and compare their utilization percentages.
Are my inputs sent anywhere?
No. The planner calculation, copy function, and CSV download run locally in your browser. Calculator inputs are not submitted to a backend.
Method sources
- U.S. EPA ENERGY STAR: Reduce Energy Losses from Power Distribution Units — defines PUE as total data center facility energy divided by energy delivered to IT equipment.
- The Green Grid glossary: Power Usage Effectiveness — PUE boundary definition and infrastructure-overhead context.
- NIST Guide to the SI, Appendix B.9 — heat-flow conversion for a refrigeration ton: 3.516853 kW.
- U.S. Department of Energy: Best Practices Guide for Energy-Efficient Data Center Design — data-center design and energy-efficiency context.
- ASHRAE TC 9.9 data center resources — thermal guidelines and data-processing-environment design context.
Sources and calculation method checked 28 July 2026. The rack-area allowance, reserve percentages, average rack load, PUE, and available failure-state capacities remain project inputs rather than universal defaults.