Calculating Waste Heat: A Guide for Energy Managers

The basic formula

A data centre's annual usable waste heat can be estimated using a relatively simple formula:

Waste heat (MWh/year) = IT load (MW) x PUE x 8,760 h x η_capture

Here, η_capture is the recovery rate: the proportion of technically capturable waste heat relative to the total heat output. The four variables must be determined independently, and each carries its own margin of error. The product of these uncertainties makes a conservative calibration of all input values essential.

Recovery rate by cooling type

The recovery rate η_capture is the most variable factor in the calculation. It depends directly on the installed cooling infrastructure.

Hyperscale facilities with modern liquid cooling (Direct Liquid Cooling or Rear-Door Heat Exchangers) achieve η_capture values of 70 to 90%. Colocation data centres with mixed infrastructure range from 50 to 70%. Legacy facilities with exclusively air-based cooling achieve 30 to 50%. Edge computing sites, often operated in minimally optimised spaces, reach 20 to 40%.

For initial feasibility calculations, the lower bound of the respective range is recommended. Anyone who calculates with 70% and actually achieves 55% has a funding gap in their business case.

Temperature profiles and their implications

The usable heat volume is only one dimension. Temperature determines which off-takers are technically viable and what additional investment is required.

Liquid-cooled facilities deliver supply temperatures between 60 and 80 degrees Celsius. This heat can be fed directly into district heating networks operating in that range, without a heat pump. Rear-door systems deliver 40 to 60 degrees. Air-cooled facilities release heat at 25 to 40 degrees Celsius.

Below 55 degrees, most district heating networks require a heat pump uplift. This significantly increases investment costs and reduces the net energy efficiency of the overall system. Operators of air-cooled facilities planning waste heat utilisation should include the heat pump investment in the business case from the outset.

Calculation example 1: Mid-sized colocation data centre

Assume a colocation data centre with 10 MW IT load, a PUE of 1.4, and air-based cooling with a recovery rate of 70%:

Waste heat = 10 MW x 1.4 x 8,760 h x 0.70 = 86,184 MWh (gross waste heat)

The usable waste heat after deducting internal losses amounts to approximately 61,320 MWh per year. This corresponds to the heating demand of roughly 2,400 single-family homes based on the German average.

Since waste heat from air cooling is produced at below 40 degrees, feeding it into a conventional district heating network would require a heat pump. Taking into account a heat pump with a COP of 3.5 and its associated electricity consumption, the net energy saving is reduced to approximately 44,000 MWh/year in primary energy equivalent.

Calculation example 2: Hyperscale with liquid cooling

A hyperscale data centre with 20 MW IT load, PUE 1.15, and Direct Liquid Cooling (η_capture = 0.92):

Waste heat = 20 MW x 1.15 x 8,760 h x 0.92 = 185,299 MWh (gross waste heat)

After deducting system-internal losses, the usable waste heat amounts to approximately 161,184 MWh per year. This corresponds to the heating demand of roughly 8,000 single-family homes.

Since supply temperatures of 65 to 75 degrees Celsius are achieved, direct grid feed-in without a heat pump is possible, provided the district heating network is designed accordingly. This makes this facility type the most economically attractive candidate for waste heat utilisation projects.

Economic minimum requirements

Not every waste heat source is economically viable to develop. Based on realised projects in Germany and Scandinavia, three threshold values can be derived below which projects are rarely profitable.

First: at least 2,000 MWh/year of usable waste heat. Smaller volumes typically do not justify the investment and operating costs. Second: a supply temperature above 45 degrees Celsius. Below that, heat pumps are mandatory, which increases costs and complexity. Third: off-takers within 3 kilometres. Each additional kilometre of pipeline costs EUR 300,000 to EUR 1.5 million and significantly extends the payback period.

Common errors in the calculation

In practice, three errors occur particularly frequently in waste heat calculations.

First error: PUE unknown. Many operators do not know their current PUE with the required accuracy. In this case, a conservative default value of 1.5 to 1.6 should be used, not the manufacturer-stated value of the cooling units.

Second error: peak load instead of average load. The IT load in the formula should be the average annual load, not the installed peak capacity. Many data centres consistently operate at 40 to 60% of their installed capacity. Calculating with the rated capacity significantly overestimates the waste heat potential.

Third error: overestimating seasonal variation. Intuitively, it seems plausible that data centres generate more heat in summer than in winter. In fact, the seasonal variation of IT load is small in most facilities. The greater seasonal fluctuation lies on the cooling side, not the IT side.

Questions about implementation? Ardor helps operators with automated data capture, calculation, and BAFA-compliant preparation of their waste heat data. contact@ardor.institute