Invited to Imperial College as part of the Franco-British Summit on Artificial Intelligence, President Emmanuel Macron has developed his vision on the future of European AI and his ability to secure critical resources. In a reasoning revealing a mastery of the subject, he recalled that among them, energy occupies a central place, in the same way as fleas or calculation capacities. A position that anchors the AI strategy of France in a material and industrial framework, far beyond the debates on regulation or ethics usually heard.
Infrastructure reasoning
“” “If you do not produce in the long term pilotable, low carbon and competitive energy, you have no chance of staying in the race“Says Emmanuel Macron in the face of an assembly of researchers, entrepreneurs and political leaders. The French president no longer dissociates the development of the artificial intelligence of the infrastructures that make it possible.” Having calculation capacities in a data center is extremely demanding in terms of energy. “
In this perspective, electricity production becomes a key factor in digital sovereignty, and on this ground, Emmanuel Macron believes that France has a strategic asset: “In 2024, we exported 90 terawatt hours, largely thanks to our historic program in nuclear.” This figure, rarely put forward in the context of innovation policies, should now establish itself as an indicator of industrial resilience.
The choice of nuclear is part of a broader trend, faced with the explosion of needs related to AI, several countries accelerate the securing of their low carbon energy resources. In the United States, long-term agreements between cloud giants and nuclear energy producers are increasing. The United Arab Emirates, on their side, deploy an integrated strategy with G42, which combines intensive calculation, sovereign cloud and localized energy supply. In this landscape, France intends to make its electric mix a lever of technological attractiveness, it remains to correct the current weaknesses of the French nuclear park but also invest in a nuclear program and develop the calendar to be in line with the new needs generated by AI. According to the RTE report published in June 2025, 15 reactors out of 56 were still stopped for extended maintenance. The cost of recovering the nuclear program (EPR2, SMR) far exceeds initial forecasts, with critical delays on certain sites. France therefore remains energetically excess short -term, but this “comparative advantage” is more than instable.
AI, a heavy industry
The link between AI and Energy is all the more structuring since the rise of foundation models is accompanied by an explosion of calculation needs. “When you represent 20 % of global demand in model training, but only 3 to 5 % of calculation capacities, you are in a situation of critical dependence”, alerts Emmanuel Macron. Here it refers to Europe’s structural delay on training infrastructure, and the need for catching up, both hardware (fleas, servers, packaging) and software.
This growing energy requirement pushes states to review their industrial strategy. Macron thus evokes the construction of “AI Factories” in France, supplied as a priority by a controlled energy, not carbonated, and which is weakly subject to geopolitical tensions. Nuclear, who has become central to French energy policies, thus finds a new justification in the digital context, remains to be seen what nuclear is so much the energy need will quickly increase in power in the event of a deployment of AI Factory and that the calendar of the French program is stalled on distant deadlines.
By 2030, some scenarios estimate that only European calculation centers could consume the energy equivalent of 10 to 15 % of the French nuclear park.
Land, dead angle of the IA strategy
An essential point, but absent from the presidential intervention, lies in the absence of land proposals that meet the industrial ambitions displayed. The infrastructure necessary for the deployment of artificial intelligence (data centers, calculation clusters, storage platforms) require large terrain, connectable to the very high voltage electrical network, often located on the outskirts of large cities or nearly pilotable energy production units. However, in France as in several European countries, the scarcity of logistics land and associated regulatory constraints make these locations particularly complex. Zan, ICPE standards, environmental rankings, permit to develop, procedures lie down, arbitrations freeze, and certain strategic projects may be postponed or abandoned under the pressure of appeal or excess environmental precaution. Several actors evoke an “blocking ecologism”, where the fear of artificialization or the short-term carbon footprint prevents the establishment of technologies with long industrial and climatic potential in the long term. In reverse, other areas of the globe, from emirates to certain regions of the United States, set up “Fast Track” land corridors, with derogatory plans and unified governance to accelerate the establishment of IA infrastructure. Without a revision of its land approach, Europe is likely to design a digital sovereignty that it will have neither space nor the conditions to bring out.
Strategic autonomy: the compute triptych, energy, capital
By connecting artificial intelligence and electricity production, the head of state sketches a systemic vision of European technological sovereignty. Energy security is no longer just a climate or industrial issue, but a prerequisite for bringing out technological champions. In his eyes, the current dependence on American or Taiwanese suppliers (NVIDIA, TSMC, FOXCONN) must be gradually attenuated by investments in semiconductors, supercomputers and energy sources.
“” “We build AI factories, we are attiring investors, but even today, more than 80 % of flea packaging is made by NVIDIA ”he specifies. The desire to attract these actors on European soil, as shown by the partnership between Mistral AI and Nvidia, aims to reassemble the value chain without renouncing strategic alliances.
This approach is in line with the France 2030 plan and European initiatives around the alliance on semiconductors, European cloud and high-performance computing joint Undertaking (EuroHPC). It underlines a conviction shared in several capitals, without control of physical layers (fleas, energy, cooling, land) software innovation will remain significantly dependent.
A vision posed, a site still to consolidate
By replacing artificial intelligence in the material realities of energy, and calculation, Emmanuel Macron sketches technological sovereignty based on tangible assets. This perspective, which goes beyond the algorithmic considerations, responds to a transformation of AI into a heavy, structuring, and intense capital industry. But between the ambitions expressed and their concrete achievement, several conditions remain to be met, align the energy calendars, do not obscure the fundamental question of land, and consolidate the value chain in a frantic cooperative. At a time when other powers advance without delay or obstacle on these projects, only the ability of France and Europe will have to transform this vision into concrete achievements, a true indicator of the strategic autonomy that President Macron aims.
