Two recent reviews of data centre energy in the EU landed within weeks of each other. The first, from the IEA's Electronic Devices & Networks Annex, examined the policies regulating data centre energy efficiency. The second, from the EU Joint Research Council, examined the underlying data on how much energy data centres actually use. Read together, they paint a picture of an industry being asked to disclose, comply and reduce on the basis of regulations that miss the biggest available lever, and data that does not robustly measure even what those regulations cover.
What the data review found
The JRC report estimates EU data centre electricity consumption at 45 to 65 TWh in 2022, between 1.8 and 2.6% of total EU energy. Telecoms adds another 25 to 30 TWh. Together, the sector accounts for somewhere between 2.8 and 3.8% of total EU energy.
The point of the report is less the headline number and more the uncertainty around it. Three recent EU studies on the same question produced answers ranging from 39.5 to 104 TWh, a 260% spread driven by methodology differences and unverified assumptions. Of the 27 EU member states, 17 are rated as having "Low" or "None" quality energy data on their data centre sector. Only 4 are rated "High" (those 4 cover 77% of EU data centre capacity, so geographic coverage is reasonable; per-country granularity is not).
The concentration is striking on its own. Germany, France, the Netherlands and Ireland account for roughly two-thirds of EU data centre energy use while representing less than 40% of the EU population. Ireland alone draws 18% of national electricity from its data centre sector. The top 12 EU markets account for 95% of total use.
Telecoms data is materially better quality (over 90% from operator data), which serves as a useful counterpoint: when an industry reports its own metered consumption, the analysis improves. When the question relies on estimates from installed capacity, the answer can vary by a factor of three depending on which assumptions the analyst chooses.
What the policy review found
The IEA review covered the energy efficiency regulations currently in force. Minimum Energy Performance Standards exist in four jurisdictions: China, Germany, France and Japan. All four are focused on PUE reductions. None of them set requirements for IT efficiency.
The same review modelled five energy-efficiency scenarios to test their potential impact:
- Migrating workloads to cloud
- Reducing PUE
- Increasing utilisation through virtualisation
- Increasing server efficiency
- Shutting equipment down at low utilisation
Several of those scenarios rest on assumptions that do not hold in practice. Cloud is not always more efficient (cloud utilisation is frequently below 40%, and migration carries security and cost implications the analysis omits). Virtualisation is not rare in enterprise data centres; it is universal. Equipment shutdown at low utilisation is structurally unavailable to cloud providers because elastic scalability depends on warm capacity.
The one scenario that does have meaningful headroom, on the IEA's own modelling, is server efficiency. The model projects up to a 45% reduction in data centre energy demand by 2027 if that lever is pulled. From the work we have done across hundreds of customer estates, that figure is on the optimistic end, but it is not unrealistic. The opportunity exists.
The legislation does not exist to support it. There is no mandatory IT efficiency standard in any of the four MEPS jurisdictions. The EU Energy Efficiency Directive references the ITEE metrics from the EU Code of Conduct, but the requirement is reporting, not improvement. Some public-sector procurement frameworks nod at IT efficiency without requiring validation or impact assessment before purchase.
Where the two reviews land together
The combined picture is uncomfortable. The data quality gap means the EU does not robustly know how much energy its data centre sector consumes, with country-level estimates varying by up to 80%. The policy gap means the regulations that do exist focus on the building (PUE) rather than the IT load running inside it, which is where the largest efficiency lever sits.
The argument structures itself almost automatically. If the biggest available efficiency lever is server-level IT efficiency, and the policies regulating data centres do not require its measurement or improvement, and the data underneath those policies cannot reliably tell us where we stand, then the sector is being asked to comply and reduce against a frame that misses the work that would actually move the number.
What needs to change
Three things, in order.
- Country-level data of usable quality. Telecoms manages it. Data centres can. It requires regulators to insist on metered consumption reporting rather than estimates derived from plate ratings.
- IT-efficiency metrics integrated into the regulatory frame, not just the reporting frame. Naming ITEEsv and ITEUsv in a directive is not the same as requiring improvement against them. The EU EED is the closest current example of moving in this direction.
- Procurement guidance with teeth. Public sector procurement specifies many things; efficiency of the IT being purchased is rarely one of them. Validation and impact assessment before purchase is the smallest change that would have the largest immediate effect.
Until those three move, the discussion about data centre energy in the EU will continue to be a discussion about buildings while the load inside them does most of the consuming.