Arthur Mensch, co-founder of Mistral AI, recalled an electrical reality: not a shortage of production, but a physical incapacity of networks to transport and deliver electricity where the AI needs it.
artificial intelligence is above all a matter of… electricity.
Behind every AI model, every query, and every data center, there is immense energy consumption. As he summed it up: data centers transform electricity into intelligence.
This sentence perhaps says more about the economic future of France than many industrial plans. Because in this new global AI economy, competition will not only be based on models, software or talent. It will also depend on the capacity of States to produce abundant, competitive and ideally carbon-free electricity. And in this area, France has a rare advantage: its engineers… and its low-carbon electricity.
For a long time, digital sovereignty has been thought of from the perspective of semiconductors, the cloud or technological platforms. But the meteoric rise of artificial intelligence is profoundly changing the situation. The energy needs of digital infrastructures are exploding. Future AI data centers will consume massive amounts of electricity, to the point of becoming energy as well as digital infrastructures tomorrow.
This transformation is no longer theoretical. In the United States, it is already causing a real supply crisis. In Northern Virginia, which has become the world’s leading data center hub, server consumption already represents around 26% of the state’s total electricity. Faced with this pressure, local operators have been forced to slow down the awarding of new contracts in order to avoid the risk of cuts for individuals. Other states like Oregon, Iowa and Nebraska are facing similar tensions on their networks. To circumvent these limits, technological giants are now developing their own energy production capacities directly on site, outside the public electricity network, what is called “behind the meter”, by mobilizing energy mixes ranging from gas to renewables, sometimes even nuclear.
Therefore, a major challenge appears: how to fuel this technological revolution without further saturating our national electricity network?
The answer may already be before our eyes, or rather on our roofs and parking lots. By multiplying solar, green and local energy production points as close as possible to needs, France could not only relieve its network, but also build the energy foundations of its digital sovereignty. Photovoltaic shade structures, industrial roofs and even battery storage could thus become strategic infrastructures of the AI economy. What yesterday was still a matter of commercial development or local energy transition could tomorrow become an issue of national technological competitiveness.
The rise of AI could accelerate the transformation of parking lots, warehouses and commercial areas into real energy production micro-sites capable of powering critical digital uses. Because local energy infrastructure could well become one of the invisible but essential links in French competitiveness in artificial intelligence.
One thing is certain: the battle for AI will not be won in algorithms alone. It will also play a role in our ability to produce enough electrons to fuel this new industrial revolution. France perhaps has a historic opportunity: to converge energy transition, industrial sovereignty and the development of artificial intelligence. On condition of rapid investment in electrical infrastructure capable of powering this new digital economy.
