Powering AI

The publication was produced with the support of the Friedrich-Ebert-Stiftung Just Climate Competence Center in Brussels. You can find the full paper on FES' website.

Europe’s planned AI compute build-out is likely to reshape its electricity systems, but current EU policy proposals don’t recognise that yet. 

Our analysis finds that AI-optimised data centres should not be treated like conventional digital infrastructure. They concentrate high-density accelerator hardware, increasingly run time-sensitive AI services, and often seek to locate close to major demand hubs. This creates a new class of industrial-scale electricity demand: one that needs large volumes of reliable power, with limited short-term flexibility unless training and batch jobs are actively managed around system conditions. 

Three bottlenecks 

We identify three emerging bottlenecks that policymakers need to address:  

• The risk of idle capacity and stranded grid investment. AI infrastructure plans can outpace actual demand or utilisation. If expected use remains uncertain, costly generation and grid investments may be built for capacity that doesn’t materialise.  

• AI data centres as an underused source of grid flexibility. Some AI workloads, especially training, can be scheduleds more flexibly. Without clear incentives and minimum requirements, however, this potential is unlikely to be used, even where it could help relieve congestion and absorb surplus renewables.  

• Siting and permitting rules that overlook local constraints. Current approaches, including proposed data centre acceleration zones under the Cloud and AI Development Act (CADA), still treat AI clusters as conventional data centres. They do not adequately account for grids constraints, resource limits, or competition for scarce low-carbon electricity. 

Recommendations 

We argue that AI clusters must be recognised and regulated as distinct industrial actors on the grid. To that end, we propose three directions for EU policy:  

1. Define AI clusters as a separate category in relevant policy and regulatory frameworks. 

2. Make public support and fast‑track treatment to transparent utilisation and verifiable flexibility commitments. 

3. Align siting decisions for AI‑specific capacity with grid development plans and low‑carbon generation. 

These choices will determine whether Europe’s AI sovereignty agenda strengthens its path towards climate neutrality – or makes that transition harder.