Home / Articles / 26 October 2026
Theme of the month: Data centresPUE and Article 12 reporting: why metering is designed in, not added on
Since 2024, data centres with over 500 kW of installed IT power report their energy indicators annually to a European database. What exactly is measured, where the metering points go, and why adding them afterwards is expensive.
Metering in a data centre has long stopped being a management convenience. Article 12 of Directive (EU) 2023/1791 on energy efficiency, elaborated by Delegated Regulation (EU) 2024/1364, requires operators of data centres with an installed IT power demand of 500 kW and above to report energy performance indicators annually to a common European database. The report is due by 15 May for the preceding calendar year.
The 500 kW threshold is not high. That is a hall of some fifty racks at moderate density, or fifteen at high density. A facility below the threshold today may find itself above it after an expansion — and then the missing metering architecture becomes the problem.
The four indicators
- PUE (Power Usage Effectiveness) — the ratio of total facility consumption to IT equipment consumption. Defined in EN 50600-4-2. A value of 1.0 is the theoretical ideal; real values run between 1.2 and 2.0 depending on climate and design.
- WUE (Water Usage Effectiveness) — litres of water per kilowatt-hour of IT energy. Significant with evaporative cooling and under growing scrutiny.
- ERF (Energy Reuse Factor) — how much of the waste heat is reused outside the facility.
- REF (Renewable Energy Factor) — the share of energy from renewable sources.
Of the four, PUE is the most quoted and the most abused. Calculated over one cold night hour it looks excellent. Calculated annually, at partial load and through summer temperatures, it tells the truth. The regulation requires annual values precisely — which is why metering has to be continuous, not a one-off exercise.
Where the metering points go
The denominator of PUE is IT equipment consumption. For it to be measurable, it must be separated from everything else at a level where the separation is unambiguous:
- At the facility incomer — total consumption. Utility metering is usually there, but it often covers neighbouring loads with nothing to do with the hall.
- At the UPS output — the most commonly used point for IT equipment. UPS losses must also be accounted for; at partial load they are not small.
- Per feeder to each rack — via metered power distribution units. Gives a true picture and allows cost allocation per customer in colocation.
- Per circuit in the distribution boards — separately for cooling, lighting and auxiliary loads. Without this, PUE is obtained by subtraction and every error lands in the denominator.
- A water meter for cooling — separate from domestic use, or WUE cannot be calculated.
The most common error is metering only the two end points — facility incomer and UPS output. PUE comes out, but when it deteriorates there is no way to tell where from. Per-circuit metering costs little at construction and is practically impossible to add in a working hall without interruptions — exactly what the whole facility was built to avoid.
What the smaller facility should do
A 100 kW server room is not subject to reporting. But metering it brings the same benefit every improvement starts with: it shows where the energy goes. A typical breakdown in an unoptimised hall is around 50 % IT equipment, 35 % cooling, 10 % power-chain losses, 5 % the rest. Every measure from the previous articles — aisle containment, raising supply temperature, replacing a UPS — affects a specific line, and that only becomes visible once the line is measured.
The practical minimum for a small facility: one meter at the incomer, one at the UPS output, one on the cooling supply. Three instruments wired to the monitoring system give PUE with adequate accuracy and show its seasonal curve.
And something that will weigh more from here
Waste heat reuse — the ERF indicator — enters the regulation with low values, because few facilities pass their heat anywhere. The direction is clear, though: in several member states, requirements to supply heat to district heating networks are already part of the permitting regime for new data centres.
For the designer this means one concrete decision taken early: the return temperature of the cooling medium. An air-cooled hall returns heat at too low a temperature to be useful. A hall with direct-to-chip liquid cooling returns at 45–60 °C — a temperature something can be done with. The cooling decision determines whether waste heat becomes an asset or just warm air above the roof.
That concludes the data centre series. Next month: BMS and Smart home systems.