Energy Consumption in Data Centres and Broadband Communication Networks in the EU (JRC135926, EUR 31841 EN)

Overview

Where demand for digital services is rising rapidly and official energy statistics do not exist, the report uses existing literature and public data sources to estimate the electricity consumption of data centres and telecommunication networks in the EU-27 in 2022. Data centres used an estimated 45-65 TWh (1.8-2.6% of total EU electricity use) and telecommunication networks 25-30 TWh (1-1.2%); together 70-95 TWh, or 2.8-3.8%. The report was published in 2024 by Kamiya and Bertoldi and informed the design of the EED Article 12 reporting scheme.

The method is a simplified country-based model. Eight countries with credible country-level estimates (Belgium, Denmark, Finland, France, Germany, Ireland, the Netherlands, Sweden) rely directly on them and cover nearly 80% of EU-27 data centre capacity. Five countries (Austria, Italy, Luxembourg, Poland, Spain) are modelled from installed IT capacity or IT floorspace with assumptions on utilisation, PUE and power density. The rest are given a first-order estimate from economic, energy and digital indicators. Because of the uncertainty, optimistic and pessimistic assumptions are added to the central estimate to give a range for each country.

The results show concentration: Germany, France, the Netherlands and Ireland account for nearly two-thirds of the region's data centre energy use with less than 40% of the population. Data centres represent 18% of national electricity use in Ireland, 5.2% in the Netherlands, 4.8% in Luxembourg, 4.5% in Denmark, 3% in Germany, 2.3% in Sweden and 2.2% in France. Large (colocation and hyperscale) centres account for about 65% of the total and enterprise centres about 35%. Networks are more uniform at 0.5-1.5% of national electricity use.

The limits are explicit. Most country-level estimates are modelled; only the Irish and Dutch estimates are based primarily on measured (electricity meter) data, and the number and installed capacity of smaller data centres are highly uncertain. The authors conclude by asking for standardised data centre definitions and classifications, energy consumption data by size and type, and better data on AI-specific energy use.

Key points

1. EU data centres used an estimated 45-65 TWh of electricity in 2022, 1.8-2.6% of total EU electricity use.
2. EU telecom networks used an estimated 25-30 TWh in 2022 (1-1.2% of EU electricity); together 70-95 TWh, 2.8-3.8%.
3. Germany, France, the Netherlands and Ireland account for nearly two-thirds of data centre energy use with less than 40% of the population; Germany, France, Italy and Spain account for about 60% of digital infrastructure energy use.
4. Data centre share of national electricity use ranges from as low as 0.4% in some countries to 18% in Ireland; for networks it stays at 0.5-1.5%.
5. Large (colocation and hyperscale) data centres use about 65% of the total and enterprise data centres about 35%.
6. Globally, 2015-2022 data centre and network energy grew only 2-8% CAGR while workloads grew about 33% CAGR, due to efficiency gains.
7. Published EU data centre estimates differ widely (e.g. 39.4 TWh for 2018 bottom-up vs 104 TWh for 2020) and each source is rated for quality; one national estimate assumes an average PUE of 1.7.
8. Eight countries were estimated from country-level figures (nearly 80% of capacity), five from IT capacity and the rest from other indicators; only the Irish and Dutch estimates are based primarily on measured data.
9. AI likely accounts for less than one-quarter of data centre energy use today; GPUs and ASICs are expected to raise rack power density and cooling needs, but AI-related energy use is very poorly understood.

Using it in design

• In grid connection and permitting discussions account for regional concentration (data centres are 18% of national electricity in Ireland); in dense markets grid capacity and permitting can be a critical risk.
• Use the consumption ranges in this report for market scale and policy context, not for a single facility energy budget; most country estimates are modelled.
• Build the assumption of rising rack power density and cooling needs for AI workloads into the capacity plan; the report states GPU and ASIC use will increase them.
• Keep your own energy data classified by size and type (enterprise, colocation, hyperscale); the report names the lack of standard classification as a main obstacle to comparison.
• Document the PUE and sources used in your own studies; the report shows that one national estimate assumes an average PUE of 1.7 and that each source is rated for quality.

Who should read it

Planning, investment and policy teams who need the scale, regional distribution and policy context of data centre energy demand.

Document contents

1. Abstract
2. Executive Summary
3. 1. Introduction: 1.1 Background, 1.2 Objectives and approach
4. 2. Literature review
5. 2.2 Modelling approaches
6. 2.3 Data centres
7. 2.4 Telecommunication networks
8. 2.5 Digital technologies and services
9. 3. Data collection and modelling
10. 3.2 Data centres
11. 3.3 Telecommunication networks
12. 3.4 Summary and limitations
13. 4. Results and discussion: 4.1 Data centres, 4.2 Telecommunication networks, 4.3 Summary
14. 5. Conclusions and recommendations
15. Annex

Related standards

• Directive (EU) 2023/1791
• EN 50600-4-2

Read the document

Publication details

Authors

George Kamiya; Paolo Bertoldi

Publisher

European Commission JRC

Year

2024

Document number

EUR 31841 EN / JRC135926

Document language

English

Source type

Government / public

Access

Open licence — downloadable

DOI

10.2760/706491

Licence

CC BY 4.0 — share with attribution

Licence evidence

op.europa.eu

Citation