The global race for artificial intelligence is now playing out on electricity grids, energy production capacities, data infrastructures and industrial supply chains. With its new roadmap on digitalization and artificial intelligence in the energy sector, published on June 3, the European Commission is making a major change in doctrine: AI is no longer considered as a simple digital subject, and is becoming a strategic industrial issue.
This development marks a break in the way Brussels approaches artificial intelligence. For a decade, the construction of European digital sovereignty has mainly revolved around regulation. From the GDPR to the AI Act, including the Data Act, the Data Governance Act and the Chips Act, the European Union has endeavored to define the rules of the game, protect data, regulate uses and secure certain technological capabilities deemed critical.
This approach allowed Europe to become a normative power, however it did not answer a question that has become central: how to create the material conditions allowing the emergence of a true European artificial intelligence industry?
The response provided by the Commission is that AI must be thought of as an industrial infrastructure based on four pillars: energy, computing, data and networks.
This new reading is a direct continuation of the Draghi report on European competitiveness. The former president of the European Central Bank stressed that the productivity gap between Europe and the United States could no longer be reduced through regulation alone. It called for an investment policy capable of supporting strategic technologies, energy and critical infrastructure. The roadmap published this week constitutes one of the first operational translations of this vision.
AI becomes a heavy industry
For several years, attention has focused on models, algorithms and data. Debates focused on the power of language models, reasoning capabilities or the risks associated with their deployment. If this phase is not completed, it is gradually supplanted by a more fundamental reality: the availability of infrastructure.
Data centers currently represent around 2.5% of European electricity consumption. According to estimates taken by the Commission, their installed capacity could increase from 12 GW to nearly 28 GW by 2030.
This growth comes at a time when the electrification of the economy is accelerating under the effect of climate and industrial policies. Electric vehicles, heat pumps, electrolysers and certain industrial activities are already competing for limited connection capacities in several European regions.
In this context, access to electricity is gradually becoming a factor of competitiveness as important as access to talent or capital.
This development brings artificial intelligence closer to sectors historically considered strategic. Building a next-generation AI campus now requires investments comparable to those of certain energy or industrial infrastructures. The economic logic is closer to that of a power plant or a port terminal than to that of a traditional software startup.
Europe joins the global AI infrastructure battle
While for several years, the United States and China have looked at artificial intelligence as a strategic sector requiring massive investments in physical infrastructure. Europe has favored the construction of a regulatory framework intended to govern uses and protect citizens. This divergence appears today as a real delay in a fast-paced competition and hyper growth.
In the United States, investments announced over the past eighteen months amount to hundreds of billions of dollars. OpenAI, Oracle and SoftBank launched Stargate. Microsoft, Amazon, Google and Meta are multiplying energy campus projects capable of powering tens of thousands of GPUs. Technology giants negotiate directly with electricity producers, invest in nuclear power and secure their energy supplies over several decades.
China is following a comparable trajectory. Beijing has integrated artificial intelligence into its national industrial planning and is simultaneously accelerating the deployment of electricity capacity, computing centers and digital infrastructure. The objective is less to produce the best models than to guarantee sustainable access to the computing power necessary for their development.
Faced with these two blocs, Europe risked appearing as a simple consumer market for technologies designed elsewhere. The roadmap published by the Commission reflects awareness: sovereignty in artificial intelligence depends as much on infrastructure as on algorithms. It remains to be seen whether this will be enough to accelerate the pace.
Data centers change status
This development explains why the Commission devotes a significant part of its strategy to data centers.
The change is particularly visible in the status that Brussels now attributes to them. Long considered as private infrastructures essentially the responsibility of technological players, data centers now appear as strategic assets whose development must be coordinated with national and European energy policies.
The Commission thus mentions the establishment of tripartite agreements bringing together energy operators, public authorities and data center operators. It wishes to encourage the use of long-term electricity purchase contracts, promote flexibility of consumption and accelerate the recovery of waste heat to supply urban networks.
Underneath, one idea emerges: data centers will no longer be able to simply connect to the network and consume the available electricity, and will have to participate in the balance of the energy system.
This approach is directly inspired by the difficulties observed in several European regions. Around Dublin, Amsterdam or Frankfurt, the concentration of data centers is already putting increasing pressure on electricity networks. Some connection requests now reach levels comparable to those of large industrial complexes. Brussels is clearly seeking to prevent these tensions from becoming a brake on the energy transition or industrial competitiveness.
The return of European industrial policy
This strategy also marks the assertive return of a European industrial policy. For several decades, the construction of the single market has been mainly based on competition, the opening of markets and regulatory harmonization. Large industrial programs had become relatively rare outside of a few strategic sectors such as aeronautics or space.
The succession of energy, geopolitical and technological crises has profoundly modified this approach.
The Chips Act for semiconductors, the AI Factories for computing capacities, the European projects in batteries or hydrogen already illustrate this evolution, and the energy-AI roadmap is part of this continuity, where certain infrastructures have become too strategic to be left solely to market dynamics.
Electricity networks, data centers, cloud infrastructures, artificial intelligence models or energy data spaces must now be considered as sovereignty assets.
The birth of European algorithmic sovereignty
For the first time, the Commission explicitly states that future artificial intelligence models used in the energy sector must be developed and mastered in Europe. The objective is to build specialized models capable of improving the forecasting of renewable production, congestion management, network maintenance and even investment planning.
This ambition reflects the emergence of a form of algorithmic sovereignty.
The challenge is not to directly compete with the general models developed by OpenAI, Anthropic or Google, but rather to guarantee that the critical systems on which European energy supplies will be based tomorrow do not depend exclusively on players outside the Union.
The logic is reminiscent of that which governed the creation of Airbus in aeronautics or Ariane in space. It is not necessarily a question of dominating the entire global market, but of maintaining mastery of technologies considered essential to the strategic autonomy of the continent.
Energy data, a new strategic asset
Brussels believes that the current fragmentation of energy data constitutes a major obstacle to the development of innovative services and efficient artificial intelligence models. Operators still have to deal with different interfaces depending on the Member States, heterogeneous access rules and varying levels of interoperability.
The Commission therefore wishes to create a framework allowing the simplified exchange of energy data on a European scale. Behind this technical objective lies a much broader ambition: to constitute a European strategic asset capable of fueling the development of future specialized AI models.
In a world where large models are trained on considerable volumes of data, Brussels clearly considers that European energy data can become a competitive advantage in the same way as computing infrastructures or electricity production capacities.
A reindustrialization strategy through AI
This strategy also constitutes an implicit response to European technological decline.
The Commission recalls that the United States produced forty reference AI models in 2024, compared to only three for the European Union. This statistic illustrates a broader problem: Europe remains competitive in research, but struggles to transform its scientific advances into industrial champions.
The document proposes to address this weakness not only through more funding for startups, but through the creation of an industrial environment favorable to the emergence of new sectors.
The objective is not only to create the next European AI champion, but to create the conditions enabling the entire European industrial fabric to use artificial intelligence to increase productivity, reduce its energy costs and strengthen its competitiveness.
The use cases highlighted concern electricity networks, nuclear power plants, renewable energies, buildings, mobility infrastructures and even administrative authorization procedures. AI is seen as a tool for transforming economic infrastructures themselves.
The real novelty of this roadmap is therefore not artificial intelligence. It is recognition that the global competition around AI will not be won only by the best models or the best researchers and will depend just as much on the ability to build power plants, develop networks, finance data centers, organize data sharing and secure critical infrastructure.
Europe no longer considers AI as just one digital sector among others and is starting to treat it as a strategic industry. In this new equation, electricity producers, network managers, data center operators and specialized model developers become components of the same industrial value chain. This is probably the most important change contained in this document.
