PUE 1.3 by 2030: The Path to the EnEfG End Target

What PUE measures and why 1.3 is demanding

Power Usage Effectiveness (PUE) is the ratio of a data centre's total energy consumption to the energy consumed by the IT infrastructure. A PUE of 1.0 would mean that no energy is spent on cooling, lighting, UPS losses, or anything else. This is physically unattainable.

The average PUE of German data centres stands at 1.6 according to Bitkom. Modern hyperscalers such as Google report annual average values of 1.1. The range illustrates how varied the technical starting positions are. For an existing facility with a PUE of 1.8, the 1.3 target means reducing overhead consumption by 63% in relative terms. This requires systematic investment, not isolated optimisation measures.

Step 1: Hot aisle / cold aisle containment

The first and in many cases fastest step towards PUE improvement is rigorous separation of hot and cold air zones. In facilities without containment, warm exhaust air from servers mixes with the supplied cold air. This forces the cooling system to provide significantly more cooling capacity than is theoretically necessary.

Physical separation using curtains, panels, or full containment typically reduces PUE by 0.1 to 0.2 points. Investment costs range from EUR 15,000 to EUR 60,000 per server row, depending on the building situation.

This measure has an additional characteristic that distinguishes it from the rest: it simultaneously raises the available waste heat temperature level. When warm exhaust air is captured in a concentrated manner, the temperature of the heat available for external use increases.

Step 2: Free cooling and economisers

Free cooling refers to operating cooling units without active refrigeration when outside temperatures are low enough. In Germany, this is possible for between 2,000 and 4,000 hours per year, depending on location.

Modern economiser systems that combine free cooling with active cooling in a hybrid arrangement can reduce the electrical energy required for cooling by 40 to 60%. For a typical existing facility with a PUE of 1.8, this translates to a reduction to approximately 1.45.

Investment costs for economiser retrofits vary considerably depending on existing infrastructure. In facilities that already have suitable roof areas or outdoor spaces, costs are manageable. In inner-city facilities with limited outdoor space, the structural situation can significantly increase the cost of the measure.

Step 3: Variable-speed drives

Pumps, fans, and compressors in conventional data centres often run at constant speed — and therefore at constant power draw — regardless of the actual cooling demand. Variable-speed drives (VSDs) enable demand-responsive speed regulation.

The physical basis is the cube law: the power consumption of a pump or fan decreases with the cube of the speed. A reduction to 80% of rated speed reduces power consumption to approximately 51%. For systems that are continuously operated at partial load, the savings are substantial.

In practice, VSDs typically deliver savings of 0.08 to 0.12 PUE points. Combined with improved control intelligence that integrates temperature sensors with load forecasting, the potential is even greater.

Step 4: Partial liquid cooling

For facilities that fail to achieve a PUE of 1.3 despite the measures described above, a partial or full conversion to liquid cooling is the next step.

Direct Liquid Cooling (DLC), in which coolant is routed directly to CPU and GPU components, achieves PUE values of 1.05 to 1.15 for the cooled servers. For the data centre as a whole, where not all systems can be converted simultaneously, a blended PUE results that, depending on the conversion rate, can lie between 1.28 and 1.32.

Converting existing server rooms to DLC is technically demanding and requires careful planning around leak protection, compatibility with existing server generations, and integration into existing monitoring systems.

UPS losses and lighting

Two factors are frequently underestimated in PUE optimisation discussions: uninterruptible power supplies (UPS) and lighting.

Older UPS systems from the early 2000s have efficiencies of 90 to 92%. Modern transformerless UPS systems achieve 96 to 97% in normal operation and up to 99% in eco-mode. For a data centre with 10 MW IT load and an older UPS system, this difference alone accounts for approximately 700 kW, which corresponds to a PUE contribution of 0.07.

LED lighting with motion-activated sensors contributes less than 0.01 PUE points in most facilities, but it is an easy-to-implement measure with no significant operational disruption.

Timeline and investment framework

The complete technical path from a legacy PUE of 1.8 to an EnEfG-compliant PUE of 1.3 requires a period of 36 to 48 months with coordinated implementation. This encompasses planning, permits, procurement, and phased implementation without significant operational disruption.

Investment costs for the full set of measures range from EUR 200,000 to EUR 500,000 per MW of IT load, depending on the initial condition and building situation. For a 10 MW data centre, that amounts to EUR 2 to 5 million.

This investment is recoverable through energy savings alone in many cases within 5 to 8 years, without factoring in subsidies. The EnEfG obligation under Section 13(3) makes the measures non-optional in any case, but transforms the question from a business discretion into a regulatory requirement with a defined deadline.

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