Rack Cooling Airflow Calculator
Calculate the ideal sensible-heat airflow for a server or network rack from design heat and allowed air-temperature rise. Then verify real delivery, containment and recirculation.
About this calculator
Quick Calculator inputs above give an instant result; use Advanced fields (factors, units, or decoding) when your datasheet differs from defaults.
Example: 10 kW IT load with 10 °C ΔT → sensible airflow roughly tens of CFM per kW depending on density assumptions.
Result explanation / How to use: treat values as screening only. Engineering assumption and disclaimer: confirm nameplate limits, OEM selectors, and site conditions before procurement or programming.
Hub: HVAC calculators — continue related sizing from this topic cluster.
Rack heat to airflow
Rack airflow result
Formula and boundary:
Sensible heat balance
The calculation uses q = ρcpQΔT in SI units. A 10 kW rack at 1.2 kg/m³, 1.006 kJ/kg·K and 10°C rise requires about 0.828 m³/s or 1,754 CFM before any explicit factor.
Primary reference: ASHRAE Handbook sensible air heat-transfer relationship and data-center airflow guidance.
Rack kW to CFM lookup (sensible heat, standard air)
Airflow is inversely proportional to the allowed temperature rise, so doubling the allowable ΔT halves the required flow. The table is built with air density 1.2 kg/m³ and specific heat 1.006 kJ/kg·K and is stated before any design factor.
| IT heat load | ΔT 5 °C | ΔT 10 °C | ΔT 15 °C | ΔT 20 °C |
|---|---|---|---|---|
| 2 kW | 351 | 175 | 117 | 88 |
| 5 kW | 877 | 438 | 292 | 219 |
| 10 kW | 1,754 | 877 | 585 | 439 |
| 15 kW | 2,631 | 1,315 | 877 | 658 |
| 20 kW | 3,508 | 1,754 | 1,169 | 877 |
| 30 kW | 5,262 | 2,631 | 1,754 | 1,315 |
| 50 kW | 8,770 | 4,385 | 2,923 | 2,192 |
Two practical notes on reading the grid:
- The 10 kW row at ΔT 10 °C is the worked example the calculator returns by default. Treat it as a reference point for scaling every other cell rather than a recommendation for a specific rack.
- A higher ΔT always lowers the required flow, but it raises the inlet temperature the equipment sees. The useful design question is not how low the flow can go — it is what the highest acceptable rack inlet temperature is, because that value sets ΔT for the whole row.
Sensible heat, latent heat and where the numbers diverge
This calculator solves a dry-air energy balance: heat in equals mass flow times specific heat times temperature rise. That is the right model for IT equipment, because servers convert essentially all of their electrical input into sensible heat — the air gets warmer, its moisture content does not change.
| Load | Treat as sensible heat? | Why |
|---|---|---|
| Server, storage and switch power | Yes — nearly all of it | Electrical input becomes heat at the board level; almost no energy leaves the rack by another path |
| PDU and busway losses | Yes, the loss only | Only the I²R loss is released as heat, not the throughput |
| UPS losses | Only if inside the zone | Losses released in the same conditioned space count; the rest belongs to the UPS room |
| Lighting and people | Partly latent | People and outside air carry moisture, so their load splits between sensible and latent |
| Humidification / dehumidification | No | Latent load does not raise dry-bulb temperature and must not be added to ΔT-driven airflow |
Because a dry-air balance ignores moisture entirely, the result is the minimum airflow required to remove the sensible load at the assumed ΔT. Real delivery is always less, which is why a design factor and a containment check sit between this number and a fan selection.
Air density is the quiet variable in the calculation. The default 1.2 kg/m³ suits roughly sea level at 20 °C; at 2,000 m altitude the same volumetric flow moves noticeably less mass, so the same rack needs more CFM for the same ΔT. If you are working from a measured mass flow or a fan curve stated in actual m³/h, adjust density rather than adding a second fudge factor.
Common mistakes in rack airflow sizing
- Adding nameplate power instead of real load. PSU nameplates are sized for worst-case configuration. Summing them can overstate the heat load by a factor of two or more. Use measured or monitored draw where you have it.
- Adding UPS input power to the rack load. Only the losses released in the same airflow zone count. Adding the full input re-counts the IT load a second time.
- Confusing airflow with cooling capacity. This result is a flow, not a tonnage. A chilled-water unit rated for the load can still fail to deliver the flow, and flow without a cold supply is just a fan.
- Ignoring bypass and recirculation. A rack can draw its rated CFM and still run hot if part of the stream short-circuits back to the return. That is a containment problem, and no amount of extra CFM fixes it cleanly.
- Treating the ideal value as the fan specification. Fans are selected at an operating point on a curve against system pressure. Adding filtration, blanking, long cable runs or a raised floor plenum moves that point, often well below the free-air figure.
- Applying one ΔT across the whole room. A dense rack and an empty one share a room but not an airflow problem. Size per rack group, then check whether the room-level system can support the aggregate.
Frequently Asked Questions
How do I convert kW of heat to CFM?
Multiply the sensible heat in kW by 3,412 to get BTU/h, or divide by the density and specific heat product in SI: flow equals heat divided by ρ times cp times ΔT. At 1.2 kg/m³ and 1.006 kJ/kg·K, roughly 176 CFM per kW at a 10 °C rise.
What allowance should I use for hot aisle containment leakage?
There is no universal figure, because it depends on door discipline, cable openings and floor gaps rather than the containment product. Measure bypass where it matters or apply an explicit design factor you can defend, rather than burying an allowance in the air properties.
What temperature rise should I design for?
Work backwards from the maximum rack inlet temperature the equipment allows and the supply air temperature you can actually deliver. The difference between those two numbers is the ΔT, and everything else follows from it.
Does this give me cooling tonnage?
No. It returns a volumetric airflow from a sensible-heat balance. Tons of refrigeration is a capacity figure for the cooling plant and must be sized separately, then checked against the flow the room can deliver.
How does altitude affect the airflow?
It lowers air density. At 2,000 m the same volumetric flow carries noticeably less mass, so the same heat load needs more CFM for the same temperature rise. Adjust the density input instead of adding a second factor.
How many CFM does a 10 kW rack need?
At 10°C air rise and the displayed standard-air properties, approximately 1,754 CFM before an explicit design factor.
Should I add UPS heat?
Add only losses released in the same airflow/conditioned zone; do not add full UPS input power.
Can I use °F temperature difference?
Yes. Select °F and the calculator converts the temperature difference to kelvin/°C difference.
Why can measured airflow differ?
Fan operating point, filters, floor pressure, doors, cables, leakage, bypass and recirculation all affect delivery.