Co-created and aggregated by the Sustainable Digital Infrastructure Alliance e.V. and the following collaborators, each of which is a signatory of the document: Jens Gröger, Marc Wilkens, Gaia Balzarini, Giuseppe Leto, Quentin Laurens, Raphael Daniel, Prof. Patricia Lago, Max Schulze.

4.1 Answers (Definition of a Data Centre)

Which definition is the clearest and most appropriate for the EED objective?

Is the clarification in Option 3 needed to prevent possible loopholes in Option 1?

There is no Option 3; however, we assume you refer to this suggestion:

We are in favour of explicitly including ICT equipment itself in the definition.

We believe it's important to use the word "ICT equipment," rather than the word "IT equipment" to include telecommunications and software components.

ICT has three components:

4.2 Answers (classification of a data centre)

We recommend Option 1 without the Complementary classification based on installed IT power, but we would rather suggest to rate not on installed IT power but the total power capacity or design IT power capacity (otherwise a half-empty data centre might slip into a lower category and distort reporting).

With the ultimate goal of "establishing a common EU scheme for the reporting of the sustainability of data centres within EU," there is no need of classifying data centres based on their "installed IT power" consumption. The EN 50600 definition disregards it; better to be consistent with it and avoid deviations that could lead to misinterpretations. We propose to adopt the total energy consumption as the relevant metric, rather than the IT power consumption, following the same approach used for the buildings (with thresholds measured in TJ, according to the rest of the EED).

4.3 Answers (reporting obligations)

Is the clarification about colocation data centres really needed to ensure the consistency of the scope?

We support the additional distinction between co-location and other data centres as it creates clarity for reporting obligations.

Do you agree with the installed IT power demand threshold? Any alternative proposal?

We believe it can be dangerous to focus only on installed IT power capacity, as it would potentially leave out empty or half-empty facilities. We believe any rating or threshold should be based on the total energy consumption.

The whole EED proposal is based on the total energy consumption, we should adopt the same approach for data centres too (instead of basing it on IT power consumption). We therefore propose that the minimum energy consumption for a data centre to report data to the EU is 4.7TJ/year (with the assumption of IT Load=100kW and PUE=1.5 and 8,760 running hours).

What are your views on the list of elements to be within the reporting scheme?

• We would suggest to be specific on the name/ownership of the facility, and emphasize the legal entity name that is owning the facility.

• Further, the data centre facility should report the total available heat for recovery with a price list for the heat per MWh at different availability and temperature levels. We have written about that extensively here (with a table).

On the list of elements on the reporting scheme, it's critical to include the utilisation rate of all IT equipment in relation to mentioned compute/memory/storage/network capacity provided by the data centre. This could be provided as a monthly average to simplify reporting, e.g.

Compute utilisation:

• 3,000 CPU cycles out of 300,000 CPU cycles available utilised (1%)

• 1,000 GB out of 5,000 GB available utilised (20%)

Storage utilisation: • 120 TB out of 1,200 TB utilised (10%)

Network utilisation:

• 120 Mbit out of 1,200 Mbit utilised (10%)

• 100 GB inbound traffic

• 100 GB outbound traffic

The "Building and infrastructure" section could be considered confidential, due to the secrecy of the design practices adopted by the data centre owner or requested by the tenant(s). Therefore, we would suggest adopting a hybrid approach where the energy- & KPI-related data are public and all the others are treated as confidential.

Voluntary Elements and Implementation Obstacles

Any obstacle to the implementation of the energy efficiency strategies (e.g., the Heat Network operator is not ready to reuse heat generated by a data centre) should be reported and discussed with the local or national authorities.

Technical Obstacles

There is a security gap (networking) between the IT monitoring systems and the facility monitoring systems (IT & OT). An interface between the two worlds is needed, e.g., similar to the smart meter in the utility sector (bridging the customer energy system to the power grid with transparency for both consumer and utility).

A proposal for such a tool has been developed as part of the Horizon 2020 project 'ECO-Qube' and a summary can be found here. Any data centre infrastructure management system or IT monitoring system should be capable of taking this role as 'data aggregator and bridge' as well.

Data Confidentiality

All of the data is relevant for the public interest and is necessary for a fact-based policy process on the EU and national level.

5.1 Answers (Reporting Obligations)

Which option would be more appropriate to pursue sustainability objectives? Which one is more realistic and easier to apply?

We believe this depends on the objective of the EU Commission—if the objective is to monitor the energy use and environmental impacts of real-estate assets in the EU, it should be the co-location companies and data centre facility owners who should do the reporting. If the aim is to monitor the energy use and environmental impact of the IT sector, then it should be the IT owners making the reports.

In any case, any data centre facility owner (landlord) should make consumption information transparent to the IT owners (tenants), as is the case in other housing & commercial buildings. This transparency enables IT owners to measure & report their energy use and environmental impact voluntarily. Currently, a lack of transparency from the side of the

Data center facility owners (co-location providers, hyperscale operators, and others) often lack transparency, making it difficult for the IT sector to assess its environmental impact.

The landlords must report on energy system utilization, including IT power usage versus design capacity. This helps the EU Commission track overprovisioning and overbuilding, which can significantly impact sustainability goals.

Contractual Requirements

The proposed reporting requirements won't need new contracts. Co-location providers may need their IT customers to report power usage, though this is typically covered by existing metering systems.

Optionally, Service Level Agreements between IT equipment customers, facility providers, and software/cloud companies could include transparent reporting requirements.

5.2 Answers (Ways to Ensure Data Quality)

We believe for consistency and quality of the reported data only a modification of option (Third party verification) makes sense.

The actual requirement in the document is: "… a third-party entity, which must be accredited according to different standards for verification and certification bodies: ISO 14001, 50001, EMAS"

These standards are not related to data centre facilities. An auditor without specific data centre facility knowledge will not be able to carry out such a complex audit properly. The certification bodies would even run the risk of losing accreditation if the auditors audit without specialised data centre knowledge.

To ensure the quality of the audits we propose the following requirements for the auditor:

• Training and experience:

● Audit experience

Requirements for auditors:

• Minimum 5 audits with 30+ days experience in DC Physical Infrastructure

• All audits completed within past 3 years

• Completed DC-specific training (EUCoC, EN 50600, ISO 30134)

Required modification: "... accredited certification bodies from the field of energy/environmental management (ISO 14001, 50001, EMAS) or the field of data centres like ISO 17165 with qualified auditors according to the described requirements."

Note: The quality of data and results depends entirely on auditor competence. Self-audits and self-declarations cannot provide the consistent data needed for this significant topic..

6.1.2 Answers (energy consumption measurements and indicators) For the PUE we prefer Option 2. With PUE level 2, good results can be achieved with a

reasonable amount of effort. On the basis of the PUE level 2, one can make good decisions

to improve the energy efficiency of the data center.

ISO 30134-2 Power Usage Effectiveness (PUE) is actually under revision.

As PUE is merely an additional KPI along with the Total Energy Consumption, impact from

technologies such as heat recovery, etc. on PUE can be mitigated through clear transparency

on actual power consumption of IT equipment and data center facility equipment.

6.1.3 Energy Consumption at Idle/Low Utilization

IT technology has a high refresh rate: infrastructure changes every 3-5 years, with software cycles even shorter. Due to this rapid evolution, energy consumption measurements at idle or full load may not provide reliable long-term data for the proposed reporting frequency. Benchmarking across different time periods becomes challenging as both technology and applications evolve.

A better metric would be the capacity utilization of IT resources (e.g., disk space, memory, CPUs).

Impact on Data Center Optimization

Server idle mode is a major source of energy waste, with approximately 50% of energy consumed during idle periods without performing useful computations.

Implementing KPIs that measure these energy losses is crucial. The SIC or DCIC indicators (rather than E_ICT_equipment_total) would effectively highlight these inefficiencies. These metrics can be improved through:

These KPIs should be incorporated into data centre standardization (EN 50600) beyond the EED requirements.

6.2.1 Additional Indicators to Consider

Beyond SIC and DCIC energy metrics, we recommend tracking actual CPU utilization. This metric addresses resource consumption efficiency, particularly for computing capacity. We propose ITEU_SV as a metric (average CPU utilization of all servers over 1 year).

For more detailed analysis, we suggest implementing weekly CPU utilization curves over a one-year period. These curves track usage from Monday 0:00-1:00 through Sunday 23:00-24:00 (168 data points), helping identify load optimization opportunities.

6.3 Ancillary Metrics

6.3.1 Water Usage Effectiveness (WUE)

We endorse Option 1 as the optimal KPI for DC water usage. As the ISO-editor of WUE, discussions with DC experts confirm Option 1's viability for operators (Marc Wilkens). EN 50600-4-9 WUE already incorporates water quality distinctions (industrial, non-industrial, rainwater) and emphasizes water reuse. Level 3 WUE includes land consumption and water scarcity metrics. Multiple DCs currently track WUE successfully.

6.3.2 Energy Reuse Factor (ERF)

As a member of WG1 SC39 responsible for ISO 30134-series, my experience with data centre experts confirms that all major waste heat utilization scenarios are effectively covered (see EN 50600-4-6, Annex A examples) (Marc

While it's technically possible to map temperature levels more precisely using the first law of thermodynamics (exergy and anergy), this approach isn't practical for real-world applications.

Therefore, we recommend Option 1, along with additional reporting of temperature levels (specifically inlet and outlet temperatures at the heat exchanger where it connects to district heating).

6.3.3 Renewable Energy Factor (REF)

In order to use the REF meaningfully, an EU-wide uniform and generally recognized definition of renewable energy is needed. Perhaps this can be achieved with Option.

6.3.4 Carbon Usage Effectiveness (CUE)

My experience as the ISO-editor of the CUE from discussions with data centre experts has shown that CUE is a good complement to the REF (Marc Wilkens). The discussions have also shown that category 1 does not cover significant emissions from data centres.

Refrigerants (e.g., R134a) and isolation gas of main switchboards (e.g., SF6) have a large CO₂ equivalent. These equivalents must be taken into account. Therefore, we believe CUE category 2 is the best KPI.

Note: The use of different refrigerants is always related to energy consumption, because the efficiency of the cooling system differs.

6.3.5 Additional Indicators

Another important indicator, which is also required in the German draft of the Energy Efficiency Act, is the share of nuclear-generated electricity in the total amount of electricity.

Members of the self-regulatory initiative "Climate Neutral Data Centre Pact" have the option of achieving their climate neutrality through "carbon-free energy," which can also consist of nuclear energy. In order to make transparent whether climate neutrality is achieved through renewable energy or nuclear power, reporting the nuclear energy factor ("NEF") is therefore an important distinguishing characteristic.

Further, as highlighted by Patricia Lago during the discussions and hereby repeated in writing, an efficiency indicator should be established to summarize how well utilized the IT capacity of a data centre is by software. This could be accomplished by including prominently a metric like "Software Usage Efficiency" which provides the foundation for data centres to be responsible and accountable for how effective they are in managing the assets (software + data) of their clients; and (2) treat software and data management (i.e., the content managed by data centres) as first-class citizens and not, as it is now, as an 'after-thought,' mentioned in the EED and there as "application" or as "infra management software" with no real associated action.