The EU Commission and their technical consultants have proposed minimum performance standard (MPS) values for PUE, water usage effectiveness (WUE) and renewable energy factor (REF) to drive improvements in data center energy and sustainability performance. The proposed values, however, are based on an analysis of an unrepresentative sample of the current EU data center population. The limited nature of the data set risks creating data center space shortages as existing data centers will need to be upgraded or replaced. The number and square footage of affected facilities risk injecting chaos into the data center market as the facilities are taken offline to meet the 2030 thresholds.

At a minimum, the Commission should implement the PUE and WUE thresholds in two steps: a less stringent threshold in 2030 and the proposed final threshold implemented in 2034. A tiered implementation will drive improvement in performance while reducing the stress on the market. The final timeline should be based on a detailed analysis of the quantity of space that will have to be replaced or upgraded in conjunction with the already ambitious data center growth plans to support AI implementation.

The Commission should not set an MPS value for the REF. Forcing operators to purchase Guarantees of Origin (GOs) to offset fossil-based energy consumption misdirects resources that should be applied to increasing the quantity of carbon-free energy consumed by the data center.

Analysis of 2023 EU database values

In the third consultation meeting (June 18, 2025) to review the EU database values and initial MPS values proposal, the technical consultants presented an analysis of the data submitted to the EU database for the 2023 operating year. According to the consultants, 740 data centers submitted data. However, using the percentages of reporting facilities by country and the data center inventory provided on the source website, the correct number of reported data centers appears to be 800 (See Table 1). The potentially larger number of data centers that submitted incomplete data does not materially affect the analysis but does reflect the current struggles associated with the data reporting process.

Table 1 Data centers reporting to the EU database for 2023, by country

Uptime Intelligence expressed concerns to the Commission regarding the dataset’s representativeness, citing the low number of submitting data centers, the number of incomplete submittals and the geographic bias toward data centers in France, Germany and the Netherlands, which may affect its suitability for proposing MPS values.

• At least 1,310 data centers (assuming 800 data centers reported) did not submit data; only 35% to 38% (by facility count) of the EU facilities listed on the source website submitted data.
• The source database reports colocation and hyperscale data centers. Enterprise data centers appear to be excluded, indicating that 2,210 data centers do not represent the complete EU data center count.
• Valid total and/or IT energy consumption values were not submitted for between 19% and 25% of the reporting data centers.
• France, Germany and the Netherlands account for 75% of the reporting data centers. On average, the remaining 24 member states reported 16% of their colocation and cloud data centers.
• The data center count does not account for the percentage of space that has reported data. Uptime’s analysis of the reported data published by the Netherlands indicates that 55% of the total reported space did not report total energy consumption and 70% of the reported IT space did not report IT consumption. This suggests that although 600 data centers reported total and IT energy consumption to the EU database, the percentage of the total data center space they represent may be significantly less than the percentage of data centers that have reported. By applying the Netherlands response data, Uptime estimates that both total and IT energy use has been reported for only 20% to30% of EU data center space.

Uptime is concerned that the sample of approximately 600 data centers used to set the MPS values for PUE, WUE and REF is biased to more developed data center markets, does not capture enterprise data center population, and represents only 20% to 30% of the operating data center space in the EU.

PUE MPS value

A total of 600 data centers submitted annual total and IT energy consumption data, enabling a PUE calculation. Based on the analysis above, the sample is too small to provide:

• A representative distribution of PUEs across the whole EU market.
• The information needed to undertake the analysis of the cost and reasonable timeline needed to transition non-compliant data centers to operate at an average PUE value of less than 1.5.

Half (50%) of the existing data centers have a PUE over 1.5 and approximately 30% have a PUE above 1.65 (see slide 25 of EY_EUDCEAR Workshop 3 presentation, attached). The median PUE value for enterprise and colocation data centers is 1.51 and 1.48, respectively. These two values indicate that roughly 225 out of 450 reported data centers will require operational improvements, upgrades or replacements to meet a PUE threshold of 1.5. Assuming the reported values represent 30% of the operating data centers, at least 650 data centers across the EU will have to install new equipment and controls or procure new space to meet the compliance thresholds.

Given the limited understanding of the economic cost and the achievable data center upgrade and conversion timeline, setting a PUE threshold of 1.5 for operating data centers by 2030 is likely to be overly ambitious. Uptime recommended that a more prudent timeline may be to set the 2030 threshold at 1.65 and move the 1.5 threshold to 2034. Setting the PUE threshold at 1.65 would force the poorest-performing data centers to either rebuild or shut down, while allowing a more gradual upgrade and replacement schedule over an additional four years.

Setting a design PUE of <1.3, with the achievement of an operational PUE of 1.4 within 3 years for all data centers commissioned in 2027 and beyond, should be a realistic target to reduce cooling energy consumption.

WUE MPS value

A total of 413 of the 600 data centers submitted annual water consumption data, enabling a WUE calculation. However, in slide 10 of EY_EUDCEAR Workshop 3 presentation (attached), the Commission’s technical consultant states that, “e.g., although water reporting was inaccurate, the data entry can still offer valuable insights on REF, ERF, and PUE”. Uptime commented to the Commission and its consultants that if the data is inaccurate, it should not be used to set an MPS value.

Furthermore, the current dataset shows that 70% to 75% of the reporting data centers can meet the 0.4 WUE threshold. However, approximately 80% of the water consumption is the responsibility of the remaining 25% to 30% of the data centers. Furthermore, 70% of the water use comes from the 15% of the data centers with a WUE greater than 1.0.

Uptime recommended that the Commission set an initial WUE threshold for data centers in operation before 2027 at 1.0 liters/kWh in 2030. The higher initial MPS requires data centers consuming the majority of the water (as identified in the current EU dataset) to improve their water use efficiency. It also allows time to build a more representative database, establish a requirement to report the type of installed cooling system and identify which data centers are in water-rich, water-scarce or water-stressed locations.

Water use is a function of the data center location (see Water is local: generalities do not apply) . The Commission needs to collect adequate data to determine the appropriate WUE threshold for data centers in water-rich regions.

Uptime also recommended that the Commission clarify the boundary for measuring the water consumption value in the WUE calculation. Data centers use water for evaporative cooling, maintenance and domestic consumption. The Commission needs to clarify, through a frequently asked questions (FAQ) document or the definition of water use in the delegated regulation, whether water use from maintenance functions and/or domestic use should be included with evaporative and blowdown losses in the WUE calculation.

Uptime agreed that new data centers facilities commissioned in 2027 and beyond should be required to achieve an average WUE of 0.4 liters/kWh. Evaporative cooling systems are available that maximize free cooling and use efficient media and spray systems to facilitate heat transfer and reduce water use. There are also economically and operationally competitive dry cooler options.

REF calculation

The data required for calculating the REF was supplied by 615 data centers, with between 70% and 75% reporting 100% consumption of renewable energy by means of acquisition. This is calculated by combining the percentage of renewable electrons consumed and the application of GOs) to offset remaining non-renewable electrons. This analysis is also limited because the dataset only represents 20% to 30% of EU data center space.

Uptime has recommended that an MPS threshold should not be proposed for the REF. Buying GOs to achieve a REF value of 100% misallocates resources and gives a distorted view of the carbon intensity of the data center’s energy consumption. Data center operators should focus their energy investments and sourcing efforts on increasing the consumption of carbon-free energy from all sources, not just renewable energy.

Instead, the Commission should calculate, assess and report (at member state level) a location-based REF. This value would be calculated by dividing the quantity of renewable or carbon-free energy consumed by the total data center energy consumption. A location-based REF would report the percentage of megawatt-hours of consumed energy generated by renewable or carbon-free energy sources.

The Commission should clarify that carbon-free energy sources, such as nuclear power, should count as renewable energy in the calculation of the REF value. Some operators are purchasing GOs to offset the consumption of nuclear generated energy to satisfy greenhouse gas inventory certification requirements. This is a misuse of resources and double counting of carbon-free energy.

Uptime proposed a methodology to set a location-based REF MPS value (see Uptime Intelligence comments to the third consultation MPS proposals, attached), but the approach may not be practical at this time.

Note: The regulatory analysis provided in this Update is the opinion of Uptime Intelligence. Data center operators should validate the interpretations with their legal staff and any relevant regulatory authorities.