What does the Strategic Roadmap mean for the deployment of AI in the energy sector?

In this second instalment of the Topic of the Month, we explore three questions: (1) where could Artificial Intelligence (AI) be deployed across the energy sector? (2) what types of AI models might be developed? and (3) what are the next steps the European Commission plans to take? We ask these questions in the context of the European Commission’s recent Strategic Roadmap on digitalisation and AI in the energy sector. Pillar II, and in particular Flagship Action 4, of the Strategic Roadmap aims to foster the ‘development of AI models across the energy value chain’.[1]

Where across the energy sector could AI be deployed?

AI has the potential to bring important benefits to the European energy sector by supporting better decision-making, enhancing efficiency, improving accuracy and reducing costs across a wide range of use cases. A study by the Energy Transition Expertise Centre 2 (ENTEC2), which supported the preparation of the Strategic Roadmap, focuses on five use cases:[2]

• Grid planning, operation, and real time control optimisation
• Demand-side management and intelligent flexibility activation
• Forecasting energy demand, renewable generation, and grid conditions
• Predictive maintenance and anomaly detection for energy assets
• AI-supported permitting of energy infrastructure

This list of use cases sparks two reflections.[3] First, the use cases are organised around the tasks that AI can support rather than around specific segments of the energy value chain. The same part of the energy value chain may therefore appear in several use cases. Electricity grids, for example, feature across all five use cases: real-time optimisation, flexibility activation, forecasting, predictive maintenance and permitting.

Second, the potential applications of AI in the energy sector are likely to extend beyond the five use cases identified in the ENTEC2 study. These use cases were selected to illustrate areas where AI is expected to deliver some of the earliest benefits in terms of performance, reliability and cost efficiency. The Strategic Roadmap itself points to several additional applications. Examples include supporting safety and operational efficiency in nuclear facilities and assisting with renovation planning for buildings and energy-poor households.

Taken together, these reflections suggest that AI could be deployed across various parts of the energy sector, both within and beyond the priority use cases identified in the ENTEC2 study.

What type of AI models might be developed?

While a broad variety of AI models may deliver benefits across the energy sector, the Strategic Roadmap explicitly refers to two categories: foundation models and generative AI. It is therefore useful to clarify what is meant by these terms.

Foundation models are generally described as models trained on large and diverse datasets that can be adapted to a wide range of downstream tasks.[4] Generative AI typically refers to AI systems capable of generating new content, such as text, images, audio, video, data or code.[5] These definitions suggest that the two terms emphasise different aspects of AI systems. The term foundation models highlights the model’s role as a common basis from which many more specific applications can be developed.[6] Generative AI refers to the system’s primary function; its capability to create new content. This distinction in emphasis is important, but it also makes the relationship between the two categories less straightforward.

Our current understanding is that foundation models and generative AI are closely related but distinct concepts. Many foundation models possess powerful generative capabilities, and the performance of generative AI often benefits from training on large and diverse datasets. [7] However, in some cases, foundation models could be adapted for non-generative purposes or generative AI may not exhibit all the characteristics commonly associated with foundation models.[8] As a result, although the two categories often overlap, neither concept seems to fully encompass the other, and caution should be applied when using the terms interchangeably.

Finally, it must be noted that the field of AI continues to evolve rapidly. As models, applications, and terminology continue to develop, the practical boundaries between these concepts may also evolve.

What are the next steps announced in the Strategic Roadmap?

To accelerate the deployment of AI across the energy sector, the Strategic Roadmap announces three key initiatives under Flagship Action 4.

The first initiative concerns the launch of the Community of Practice ‘AI.grids’ for the development of AI models to improve the management and planning of energy grids, which was signed alongside the Strategic Roadmap. AI.grids aims to build the first Pan-European AI foundation model for energy grid operations through a collaborative ecosystem of system operators, research organisations, and technology providers, comprising 48 partners.

The second initiative focuses on the development of digital portals for Member States to streamline permit review for renewable energy, storage and grid projects, using generative AI. This initiative is in line with the European Grids Package, in particular with the Commission’s proposal on acceleration of permit-granting procedures.[9] The Strategic Roadmap foresees the design of these AI tools in 2027, and for public authorities to start using them in 2028.

The third initiative is to support research and innovation on AI for the energy sector through Horizon Europe. More specifically, the Horizon Europe Work Programme of 2026-2027 on Climate, Energy and Mobility includes two dedicated calls titled ‘Data sharing to support the training and development of AI foundation models in the energy sector’ and ‘Large scale operational validation and upscaling of state-of-the-art (Generative) AI tools and models powering a next generation digital energy system’.[10] The two calls suggest that the Commission intends to maintain its focus on AI foundation models and generative AI, while extending the scope beyond the use cases of the first two initiatives.

Acknowledgements

The Florence School of Regulation gratefully acknowledges the financial support of the European Commission (DG ENER) for conducting the research that led to this blogpost. Views expressed in this blogpost reflect the opinion of individual author(s) and do not necessarily reflect the views of the European Commission.

[1] An overview of the other pillars and flagship actions of the Strategic Roadmap can be found in the first instalment of this Topic of the Month.

[2] This article follows the classification developed in Task 2 of the ENTEC2 study. It should be noted that the AI use case summary presented in Task 1 of the same study replaces the fifth use case with ‘Tacit and explicit knowledge management from operational data and expert insights’.

[3] Interested readers may refer to the ENTEC2 study for a more detailed description of the use cases, including their current application, market potential and barriers to adoption. The study also discusses the potential role of foundation models for each of the prioritised use cases.

[4] This description draws on the definitions of foundation models provided by EESC, ENTEC2, ETIP SNET and Stanford University.

[5] This description draws on the definitions of generative AI provided by EESC, ETIP SNET and IEA.

[6] By introducing the term foundation model, the Center for Research on Foundation Models at the Stanford Institute for Human-Centered Artificial Intelligence aimed to emphasise that these models serve as a common but incomplete foundation from which many task-specific applications can be adapted, making questions of reliability, safety and security particularly important.

[7] See, amongst others, On the Opportunities and Risks of Foundation Models, Center for Research on Foundation Models at the Stanford Institute for Human-Centered Artificial Intelligence (2020) and Generative AI Outlook Report – Exploring the Intersection of Technology, Joint Research Centre, (2025).

[8] See, amongst others, What are foundation models?, Google Cloud, (2026) and Generative AI and foundation models in the EU: Uptake, opportunities, challenges, and a way forward, European Economic and Social Committee, (2025).

[9] See European Grids Package, European Commission (2025) and Proposal for a Directive amending Directives (EU) 2018/2001, (EU) 2019/944, (EU) 2024/1788 as regards acceleration of permit-granting procedures, COM(2025) 1007 final. Amendments of Art. 16 of Directive (EU) 2018/2001.

[10] For more details, see HORIZON-CL5-2026-11-D3-23 and HORIZON-CL5-2027-02-D3-24.