This is new topic related to an earlier question by @mrchrisadams.
The original question was about recent plans by the European Commission to make Europe a global center for AI processing (Rankin 2025). That issue has a number of strands — so let’s try the following breakdown:
- what kind of growth in data centers (DC) is being sought?
- is that growth feasible and likely from perspectives of economics and social license?
- have the associated electricity demands been factored into current energy scenarios and downstream analyses?
- is that analysis accessible, transparent, reproducible, and conformant to open science?
- what conclusions can be drawn?
Please excuse any inaccuracies, misconceptions, and omissions in the following treatment, but rather, consider this a first stab to get the ball rolling.
Data centres
There is quite a lot of material published online about the projected uptake of data centres in Europe — with all the usual caveats. I won’t cover this aspect, except to say that the European Commission recently outlined its own ambitions in this regard (Rankin 2025).
Associated demand
The IEA (Spencer et al 2025) has just published are report that examines, among other things, data centre growth in the United States and globally. Harvey (2025) covered this report and quotes former Green MEP Claude Turmes as saying that the IEA view was too rosy and failed to offer advice on protective regulations. Regardless, if you look at the graphs showing growth prospects in the IEA report, you will see considerable variance.
One significant issue that I have not seen addressed is the high price of electricity in Europe relative to other locations. Normally one would expect AI‑based data centres to be located near cheap electricity, in the same way that cryptocurrency mining operations usually are.
The social license for new high‑voltage transmission is another key consideration. The Armagh to Tyrone 400kV high-voltage link crossing Northern Island is a case in point, now likely subject to judicial review despite construction having commenced (O’Carroll 2025). Community campaigners complain that the finished line is primarily there to service energy‑hungry datacentres in the Republic of Ireland (an observation that may not be remotely accurate).
The impact of additional electricity demand from data centres on stated net‑zero targets for the European Union is also highly material. And can only be revealed through systems modeling.
Energy system public policy scenarios
Turning to the question of whether or not data centres do indeed feature in future energy system scenarios for Europe (essentially the original question from @mrchrisadams).
To my knowledge, the European Commission 2025 energy system reference scenario is not yet public (I emailed the Commission on 10 April 2025 for a status update). So I cannot comment on the inclusion of data centres in that baseline scenario. That scenario is also tailored to the closed‑source PRIMES
framework. The Commission intends to make about 90% of the related inputs to PRIMES
public — a decision that was criticized by academic and civil society representatives present at that 22 October meeting last year (covered here) because the lack of complete information makes reproducibility impossible. This is a serious matter.
A recent study (Stöcker et al 2025) focusing on Germany but based on a European‑wide closed‑source model but did not explicitly factor in data centre growth (for disclosure, my wife Berit worked on that report). The report nonetheless provides a good example of how analysis related to data center demands might proceed. The ENERTILE
model from Fraunhofer ISI and the associated LTS T45-Strom*
scenario are covered in appendix 7.1. Incidentally, the study was reported by Der Spiegel newspaper (mha/dpa 2024) as follows:
However, model calculations also showed that baseload power plants could be integrated into an energy system dominated by solar and wind energy. Their electricity could be used for electrolysis during periods of low demand, thus reducing hydrogen imports. [translated]
PRIMES
As mentioned, the workhorse energy system model for the European Commission is PRIMES
. A recent think tank report by McWilliams et al (2025) called out PRIMES
for its lack of transparency (page 4):
Since 1990, the PRIMES model has been increasingly used by the European Commission as a reference tool for assessing major energy and climate policy decisions (E3 Modelling, 2018). PRIMES is owned by a United Kingdom consultancy and the data, code and assumptions underpinning the model are not made publicly available. This makes it impossible for external stakeholders to properly evaluate European Commission modelling for important issues such as the 2030, 2040 and 2050 climate targets. Part of the issue is that certain datasets needed to run the model are only available from commercial data providers and public distribution is not permitted. This is the case, for example, for much data on commodity or fuel prices from energy exchanges, which is often only available on a commercial basis.
Discussion
It seems that the recent aim of the European Commission to promote a European AI‑based data processing hub has not been factored into current energy system scenarios – at least not that I am aware of. Moreover, the current workflow and analysis used by the Commission is not reproducible, let alone replicable, by independent researchers — this represents a major problem, both technically and in terms of democratic norms.
Closure
@mrchrisadams: please reply if you have more questions or comments.
If anyone else wishes to chip in, please do so! It would be great to hear from the PyPSA‑Eur folk, for instance, on this particular question.
References
Harvey, Fiona (10 April 2025). “Energy demands from AI datacentres to quadruple by 2030, says report”. The Guardian. London, United Kingdom. ISSN 0261-3077.
McWilliams, Ben, Simone Tagliapietra, and Georg Zachmann (19 February 2025). Europe’s energy information problem — Policy brief issue 07/25. Brussels, Belgium: Bruegel AISBL.
mha/dpa (3 December 2024). “Stromversorgung laut Studie auch ohne Grundlastkraftwerke sicher” [Study shows that the power supply is secure even without baseload power plants] (in German). Der Spiegel. Hamburg, Germany. ISSN 2195-1349.
O’Carroll, Lisa (8 April 2025). “Northern Ireland faces court case over £300m north-south power pylon plan”. The Guardian. London, United Kingdom. ISSN 0261-3077.
Rankin, Jennifer (9 April 2025). “EU to build AI gigafactories in €20bn push to catch up with US and China”. The Guardian. London, United Kingdom. ISSN 0261-3077.
Spencer, Thomas, Siddharth Singh, Davide D’Ambrosio, Hugh Hopewell, Vincent Jacamon, Alex Martinos, Nicholas Salmon, and Brent Wanner (April 2025). Energy and AI — World Energy Outlook Special Report. Paris, France: IEA Publications. CC‑BY‑4.0 license.
Stöcker, Philipp, Berit Erlach, Sven Wurbs, Frank-Detlef Drake, Manfred Fischedick, Jutta Hanson, Hans-Martin Henning, Wilhelm Kiewitt, Jochen Kreusel, Albert Moser, Wolfram Münch, Karen Pittel, Albrecht Reuter, Dirk Uwe Sauer, Wolf-Peter Schill, Hartmut Spliethoff, Cyril Stephanos, Christoph Weber, and Anke Weidlich (February 2025). Nuclear fission, natural gas, geothermal energy, nuclear fusion. The future role of baseload power plants — Discussion paper. Germany: acatech, Leopoldina, Akademienunion. German language version.
▢