AI has become an instrument of power. Europe will not win by imitating others, but by capitalizing on its assets: energy, trusted data and quantum.
The recent restrictions imposed by the United States on access to certain strategic technologies do not constitute a simple episode of trade tensions.
They mark a much more profound change.
For several decades, globalization was based on a conviction: innovations would circulate freely, interdependencies would promote prosperity and technology would remain a factor of cooperation.
This era is coming to an end.
Semiconductors, cloud infrastructures, computing capabilities, business modelsartificial intelligencedata or tomorrow quantum technologies will become instruments of power. They now determine the capacity of States to produce wealth, to protect their interests and, more fundamentally, to maintain their freedom of decision.
Power changes its nature.
As the next French elections approach, one development nevertheless deserves to be welcomed.
For the first time in a long time, technological sovereignty is the subject of a relatively broad consensus. Most politicians now recognize that Europe cannot sustainably depend on infrastructures, platforms and technologies designed elsewhere.
This is progress; but a consensus does not constitute a strategy.
The real question is no longer whether more sovereignty is needed but to define what model of power Europe wishes to build in the 21st century to guarantee its exercise.
Sovereignty no longer begins at the borders. It starts in infrastructure.
For two centuries, the power of nations was built around the control of natural resources, energy, industry and military capabilities.
The 21st century adds a new dimension: knowledge itself becomes a strategic infrastructure.
Artificial intelligence models are not simple software; they already assist doctors, engineers, researchers, magistrates, administrations and armies.
They gradually contribute to producing knowledge, guiding decisions, detecting risks and arbitrating choices.
In other words, computing centers become cognitive infrastructures.
Tomorrow, depending on foreign artificial intelligence could produce effects comparable to energy or industrial dependence.
Political autonomy will depend as much on the control of these infrastructures as on the control of physical infrastructures.
Making energy a strategic lever
We often talk about GPUs.
We talk much less about their real raw material.
Energy.
Each generation of artificial intelligence models requires ever-increasing computing capacities, themselves directly linked to the availability of abundant, stable and competitive energy.
From this angle, France has an exceptional advantage.
Its nuclear fleet probably constitutes one of the main European strategic assets in the artificial intelligence economy.
This advantage should no longer be seen as a simple economic opportunity.
It must become an industrial policy lever.
Access to this energy can no longer be thought of as a simple commercial service.
When a foreign player chooses to set up a mega data center on European territory to benefit from this energy competitiveness, it is legitimate to expect compensation: technology transfers, establishment of research centers, opening of computing capacities to European companies, participation in the financing of sovereign infrastructures, training of talents, creation of intellectual property in Europe.
This is not about protectionism.
It is a question of applying to our strategic assets the same lucidity that is now practiced by the other great powers.
Moving from power to confidence
Since the emergence of generative AI, global competition has largely been based on quantitative logic: always more GPUs, more data and more parameters.
This unsustainable trajectory in the medium term is not, however, the only possible one.
In health, industry, energy, research, justice or administration, the quality of an artificial intelligence system depends less on the volume of knowledge it absorbs than on the quality of the information on which it relies.
Europe has a unique heritage: scientific, industrial, administrative, medical and cultural data.
Some of the richest documentary bodies in the world.
The real challenge is therefore perhaps not to build the largest model.
It is to build the best knowledge.
This requires European data spaces that are collectively governed, documented, anonymized when necessary, legally usable, versioned, audited and regularly enriched.
In other words, transforming data protection into a competitive advantage.
Europe will probably not win the battle for the most powerful model.
On the other hand, it can become the continent that develops artificial intelligence in which the world will place the most trust.
Quantum can change the rules of the game
Most analyzes assume that advances in artificial intelligence will continue to depend primarily on the accumulation of computing power.
This hypothesis deserves to be questioned.
Quantum computing opens up another perspective.
It could accelerate certain optimization calculations, improve scientific simulation, transform the discovery of new materials or new drugs, strengthen learning capabilities and profoundly modify the algorithmic architectures of tomorrow.
The real breakthrough will perhaps not come from ever more gigantic models.
It will perhaps come from more intelligent, more specialized, more efficient models.
Europe already has leading research in the quantum field.
By articulating this scientific advance with high-quality data and competitive energy, it could offer a radically different approach: moving from brute force logic to precision logic.
Building a strategic European knowledge complex
None of these ambitions will succeed without an overall vision.
Europe has no shortage of talent.
It already has champions in semiconductor production equipment, electronic components, research, the cloud, artificial intelligence and cybersecurity.
What it lacks is a common architecture.
Just as states built military-industrial complexes in the last century, Europe must now build a strategic knowledge complex.
This would bring together the energy infrastructure, semiconductors, computing capacities, sovereign clouds, data spaces, open software, cybersecurity, artificial intelligence, quantum computing, research, training and financing necessary for the hypergrowth of its companies.
It is no longer a question of thinking about each sector separately; it is about considering the whole as a system because sovereignty never results from an isolated technology.
It arises from the control of an entire value chain.
Invent another idea of power
Basically, Europe would be making a mistake by simply seeking to reproduce American or Chinese models.
Her story shows that her greatest successes came when she offered a different vision.
Airbus has never been a copy of Boeing.
CERN has never sought to replicate American laboratories.
Europe has always succeeded when it transformed its singularities into competitive advantages.
Artificial intelligence today offers a comparable opportunity: not to become the continent of the most gigantic models but that of the most reliable models.
Not that of unlimited consumption of resources, but that of efficiency.
Not that of opacity, but that of trust.
Not that of dependence, but that of autonomy.
Technological sovereignty is not an end in itself.
It becomes the condition for the exercise of all other sovereignties: economic, energy, scientific, democratic and political.
It is probably on this terrain that the real competition between the great powers will be played out over the coming decades.
And this is where Europe can still choose to no longer play by others’ rules, but to invent its own.
