Europe Energy Briefs – Europe's growing AI ecosystem

A complete ecosystem is being built around AI in Europe, with significant implications but also potential for the energy sector.

At the heart is the AI Act, which was adopted by the European parliament in June 2024 with the aim to establish a harmonised framework for AI systems in the EU and with the first rules coming into effect in February 2025.

The AI Act follows a risk based approach with AI systems categorised according to their level of risk. As electricity grids are a critical infrastructure, AI systems for managing these are categorised as ‘high risk’, whereas for example GenAI that may be implemented for customer service is ‘limited risk’.

This high risk categorisation brings a new set of stringent obligations, particularly for DSOs and TSOs, which also need to take account of any national requirements.

Also of interest
Redefining AI from villain to hero in the energy transition

Thus DSOs and TSOs are required to conduct risk assessments, implement mitigation and classify their AI systems – and they are not identical as in addition to the management of their grids, DSOs also interact directly with customers.

For example, for DSOs impact areas identified are consumer interaction, demand management, local grid stability, DER integration and cybersecurity. For TSOs the impact areas identified are large-area grid stability, cybersecurity and cross-border energy trading.

The recent ETIP SNET paper on ‘Unlocking the potential of AI and generative AI in smart grids’, which devotes a chapter to Europe’s regulatory landscape and should be referred to for more information, highlights that the distinct roles of DSOs and TSOs necessitate tailored compliance strategies and requires careful planning to meet staggered deadlines.

To aid compliance, an AI Act service desk is due to be launched by the Commission in July 2025 to serve as a central hub for information, guidance and support tools.

EU continent action plan

With the AI Act starting to be implemented the European Commission in April launched its AI continent action plan with a series of initiatives for the EU “to become a global leader in AI”.

Five ‘pillars’ are identified around computing infrastructure, data access, adoption, skills and regulatory simplification, of which for the latter already noted is the AI service desk.

In terms of the infrastructure a network of 13 AI factories are currently being deployed around Europe's supercomputers to train and develop large AI models, with the first of these due in 2025-2026.

Then there is the plan to develop up to five AI gigafactories, with at least four times more computing capacity than the AI factories, to train and develop complex AI models at scale, i.e. with hundreds of trillions of parameters.

Given the scale of these, each requiring an investment estimated up to €5 billion ($5.6 billion), they will take longer to deploy and so far an expression of interest call has been released with the formal call for establishment due in Q4 2025.

Once in operation these facilities will be open to AI startups, industry and researchers under specific access conditions.

Cloud capacity is another component of the infrastructure and the Commission is proposing an at least tripling of the EU’s data centre capacity to meet its needs within the next five to seven years.

With the obvious impact on the grid, the forthcoming roadmap for digitalisation and AI in the energy sector is due to propose measures for the integration of data centres into the energy system as well as address other related issues such as electricity grid optimisation, energy efficiency in buildings and industry and demand side flexibility.

Similarly an upcoming water resilience strategy is planned to look at reducing the water footprint of these installations and increasing their circularity through water reuse, efficiency and dry cooling.

Alongside this computing infrastructure is the need for data for feed it and the action plan proposes the creation of data labs as integral components of the AI factories, bringing together and federating large, high quality data volumes from different sources, including the common data spaces.

Clearly, clear rules will need to be in place in line with those applicable to the data spaces, such as the energy data space.

Innovation and skills

The third ‘pillar’ is adoption and the energy sector is listed as one with among the largest untapped potential and will be addressed in a forthcoming ‘Apply AI’ strategy. Among the components is a developing network of innovation hubs to support AI uptake.

Then finally there is the all important skills development for AI, with a forthcoming roadmap, an ‘AI in education’ initiative to develop AI literacy at primary and secondary levels and an AI skills academy to support education and research, particularly at tertiary level.

So – a lot of policy proposals and activities in the pipeline that the energy sector can tap into in the years ahead, but what about the here and now?

As part of its remit the ETIP SNET paper has formulated a roadmap with short, medium and long term actions. Underlying these is the need for pooling of expertise and cooperation among stakeholders to share experiences, build expertise and generally advance the development of AI in the region’s energy sector.