The claim that loosening Germany’s Energy Efficiency Act for data centres would cost the economy billions comes from a study co-authored by AlgorithmWatch, the group campaigning against the change, with Umweltinstitut München and Fraunhofer ISI. None of the published sources offers an independent or government cost estimate.
How this is computed
- Key claims supported: 12 of 24
- Type of source: –
- Independent confirmation: None
Rates the source, not the topic and not our analysis.
Why it matters
Our reading: The cost question is central to the Bundestag debate on the draft. So far only one side has put numbers on it. AlgorithmWatch’s figures illustrate how small changes to efficiency limits might add up as data centre construction grows, but they remain modelling by an interested party until the assumptions are reviewed. The only non-AlgorithmWatch source, an MIT Technology Review essay, addresses public attitudes to AI, not the economics of the rules.
What the sources say
- Conditionally supported AlgorithmWatch states that the federal government adopted the draft law on 10.7.2026 (Drucksache 21/8027). According to AlgorithmWatch, it amends the Energy Efficiency Act and the Energy Services Act as well as several ordinances.
- Conditionally supported According to AlgorithmWatch, a study it produced with Umweltinstitut München and Fraunhofer ISI concludes that weaker efficiency requirements would add billions of euros in costs for the German economy and prolong dependence on fossil energy. The results rest on modelling to 2045.
- Conditionally supported The same study calculates that weakening the current efficiency limit alone would require extra supply equivalent to four gas power plants. It puts the figure at seven if a planned, ideal-condition value replaced the measured one. The press release does not give plant size or utilisation assumptions.
- Conditionally supported According to the study, even at an electricity price of 5 ct/kWh the avoidable consumption up to 2045 would cost 3.75 billion euros. The figure belongs to the seven-gas-plant scenario.
- Conditionally supported AlgorithmWatch says the draft adds broad trade-secret exceptions to the duty to publish energy indicators for individual data centres. It argues this contradicts the law’s intent; that assessment is its own interpretation.
- Conditionally supported MIT Technology Review cites a Gallup poll from May: 71% of US adults said they would oppose a new AI data centre in their area, against 53% for a new nuclear plant. This is a survey of US attitudes, not a measure of costs, and it does not address German regulation.
The key caveat
Every cost and energy figure traces back to AlgorithmWatch, which co-authored the study and campaigns against the draft. The sources do not lay out the study’s assumptions, and no independent review of them is cited.
Open questions
- Has the economics ministry published its own estimate of the energy and cost effects of the amendment?
- Which plant sizes, utilisation rates and expansion assumptions underlie the four- and seven-gas-plant scenarios?
- How many of the data centres AlgorithmWatch counts as planned or under construction will actually be built?
What’s next
Watch how the Bundestag economics committee proceeds with Drucksache 21/8027 after its public hearing.
Sources: AlgorithmWatch, 24 September 2026: 160.000+ Unterschriften an Bundestag übergeben – für Klimaschutz statt KI-Wahn · AlgorithmWatch, 1 October 2026: Stellungnahme zur öffentlichen Anhörung anlässlich der geplanten Novelle des Energieeffizienzgesetzes · AlgorithmWatch, 2 October 2026: AlgorithmWatch kritisiert im Wirtschaftsausschuss: Weniger Effizienz bei Rechenzentren kostet die Wirtschaft Milliarden · MIT Technology Review – KI, 5 October 2026: People really hate AI, so why can’t they get enough?
Produced by Zeitgeist’s AI agents (research, editing, standards check) with Claude. No human review.