Commission Staff Working Document SWD(2019) 106 - Impact Assessment accompanying Commission Regulation laying down ecodesign requirements for servers and data storage products

Overview

The document applies the significance and proportionality tests of the ecodesign framework to show why servers and data storage products should be regulated. On scale, EU-28 data centre electricity use (servers, storage and network equipment) was estimated at 78 TWh in 2015, 2.5% of final electricity consumption. Servers and storage alone used about 53 TWh (2%) in 2015 and are projected at 55 TWh in 2020 and 78 TWh in 2030; including infrastructure such as cooling, the figure is 96 TWh in 2015 (3.2%) and 121 TWh in 2030. Industry revenue in 2015 was estimated at about EUR 55 billion.

For the energy split inside data centres, a table assumes IT at 52% in 2010, 57% in 2015 and 65% from 2020, with 2020-2030 values interpolated assuming a business-as-usual PUE of 1.55 in 2030. The document also cites the EURECA project's analysis of more than 350 public administration data centres, which found average annual server utilisation of around 20%; this supports the focus on idle power. Underlying drivers include lack of functional information on energy and resources, and split incentives.

The options are business as usual (1), self-regulation (2, discarded because industry proposed none), ecodesign requirements (3), energy labelling (4, discarded because B2B products lack a suitable metric and data) and a complementary approach such as ENERGY STAR (5). Option 3 is assessed in six components: PSU minimum efficiency (for example in scenario 3.2 Gold in 2019, Platinum in 2023, Titanium in 2026), operating temperature information or ASHRAE-based minimum requirements, idle power, SERT-based performance, information requirements and material efficiency. Scenarios 3.1, 3.2 and 3.3 differ in stringency.

The preferred option is 3.2. For servers, 2030 savings are about 2.4 TWh direct and 3.7 TWh indirect (infrastructure), 6.1 TWh in total (2.1 Mt CO2e); for storage 0.8 TWh direct and 2 TWh indirect, 2.8 TWh (0.9 Mt CO2e); in total 6.5 TWh in 2020 and 8.9 TWh in 2030. In the scenario comparison, 2030 server savings including infrastructure are 2.7 TWh for 3.1, 6.1 TWh for 3.2 and 15.3 TWh for 3.3. The most ambitious 3.3 assumes 100% adoption and is treated as overly optimistic, since actually raising operating temperature depends on operator behaviour.

Long-term projections are treated as uncertain because of rapid technological change. Scenario 3.2 assumes 30% of data centres adopt higher operating conditions; at 15% adoption savings including infrastructure fall by about a third or more (5.6 TWh for servers and storage), and at 45% they rise to 12.1 TWh (a 36% increase). The document proposes a review within four years and lists monitoring indicators such as PSU efficiency compliance, idle power, presence of temperature information and material efficiency items such as ease of disassembly.

Key points

1. EU-28 data centre electricity (servers, storage, network) was estimated at 78 TWh in 2015, about 2.5% of final electricity consumption.
2. Servers and storage alone used about 53 TWh in 2015, projected 55 TWh in 2020 and 78 TWh in 2030; including infrastructure 96 TWh in 2015 and 121 TWh in 2030.
3. The preferred option 3.2 saves 6.5 TWh per year in 2020 and 8.9 TWh in 2030.
4. In 2030 servers save about 2.4 TWh direct plus 3.7 TWh infrastructure (6.1 TWh, 2.1 Mt CO2e); storage 0.8 TWh direct plus 2 TWh indirect (2.8 TWh, 0.9 Mt CO2e).
5. Server and storage industry revenue was estimated at about EUR 55 billion in 2015.
6. IT energy share is assumed at 52% in 2010, 57% in 2015 and 65% after 2020; a PUE of 1.55 is used for the 2030 business-as-usual case.
7. The EURECA project found average annual server utilisation of around 20% across more than 350 public data centres.
8. Energy labelling was discarded for lack of a suitable metric and data for B2B products, and self-regulation because industry proposed none.
9. Scenario 3.2 assumes 30% of data centres adopt higher operating temperatures through the information requirement; at 15% savings fall to 5.6 TWh, at 45% they rise to 12.1 TWh.
10. A review within four years is proposed; monitoring covers PSU efficiency, idle power, operating temperature information and disassemblability.

Using it in design

• In the what-if module show direct and infrastructure savings separately for IT refresh scenarios: the document finds infrastructure savings exceed direct IT savings, which supports showing that every watt saved in IT also saves cooling and power infrastructure energy.
• Use the declared operating condition class (A1-A4) when choosing inlet temperature targets for a new hall; the document's purpose for this information requirement is to let operators raise operating temperature safely.
• For low-utilisation server room retrofits, treat roughly 20% average server utilisation as a starting assumption and show the savings potential of consolidation and virtualisation.
• Idle power matters, but industry experts argue it is not a sufficient efficiency proxy by itself; compare servers on idle power together with SERT-based active state efficiency.
• Whether operators actually raise operating temperature depends on behaviour; model the adoption rate (for example 15%, 30%, 45%) as a sensitivity parameter in temperature-increase savings.

Who should read it

Engineers, policy analysts and researchers who want the reasoning, energy figures and economic impacts behind the server and storage ecodesign regulation.

Document contents

1. Introduction
2. Problem definition
3. What is the scale of the problem?
4. What are the underlying drivers of the problem?
5. Who is affected, in what ways and to what extent?
6. How are existing policies and legislation affecting the issue?
7. Baseline scenario: How will the issue evolve in absence of intervention?
8. Should the EU act?
9. Policy objectives
10. Policy options
11. Minimum efficiency requirements for power supply units
12. Operating temperature information requirements
13. Requirement on the idle mode energy consumption
14. Information or MEPS requirement on servers' performance (e.g. SERT-based)
15. Material efficiency requirements
16. Option 5: Complementary regulatory approaches - compulsory Energy Star
17. Analysis of impacts
18. Comparison of options
19. Evaluation and monitoring
20. Annexes

Related standards

• Regulation (EU) 2019/424

Read the document

Publication details

Authors

European Commission

Publisher

European Commission

Year

2019

Document number

SWD(2019) 106 final (CELEX 52019SC0106)

Document language

English

Source type

Government / public

Access

Open licence — downloadable

Licence

EU reuse (2011/833/EU) — reuse permitted with acknowledgement of the source

Licence evidence

eur-lex.europa.eu

Citation