By Dr. Tawhid CHTIOUI, Founding President of aivancity, the leading school for AI and data
The Illusion of Data Centers in AI Strategy
Over the past two days, during the artificial intelligence summit, the French government has made a series of announcements regarding the development of data centers, which are supposed to be the cornerstone of the country’s digital sovereignty. Rather than fostering a strategic and innovative vision for AI, these announcements reveal a profound misunderstanding of the real issues at stake. By focusing efforts on physical infrastructure, they overlook the fact that AI also relies on algorithm optimization, data quality, and data governance. This approach is not only reductive but also outdated in the face of new technological dynamics.
While AI requires computing power, the issue is not limited to the construction of server farms (Thirty-five sites in France are “ready for use” to host data centers for artificial intelligence, announced Clara Chappaz, the minister in charge of AI, on Thursday following an interministerial committee meeting chaired by François Bayrou; The Élysée Palace also announced on Sunday evening a private investment plan totaling 109 billion euros by 2031, primarily for data centers).
Reducing AI to data centers is tantamount to confusing the technical infrastructure with the fundamental challenge of AI.
AI is, above all, a transformation of practices, business models, and organizations. The real challenges lie in its impact on work, society, the economy, and governance.
A data center is simply a logistical tool, just like roads are for transportation. No one talks exclusively about power plants when discussing the digital revolution, so why confine AI to a debate about infrastructure?
Data centers won't make France a leader in AI
There is no denying that data centers play a strategic role in digital sovereignty, particularly by ensuring that data is hosted within the country’s borders and by providing greater control over foreign cloud giants. In a context where control over digital infrastructure is becoming a geopolitical issue, it is legitimate to seek to strengthen the country’s capacity to store and process its own data, thereby avoiding excessive dependence on American and Chinese hyperscalers such as AWS, Microsoft Azure, or Alibaba Cloud. However, this approach must not come at the expense of a broader reflection on the country’s economic and technological structure.
In this context, how do these investments in data centers fit into the European strategy to develop a sovereign cloud infrastructure and a unified European database by 2030? The European Union aims to build an integrated digital framework that allows member states to access a shared, secure infrastructure independent of major American and Chinese players. This vision relies on coordination at the continental level, where interoperability, shared data management, and the optimization of existing infrastructure must take precedence over the duplication of national data centers, which are costly and sometimes redundant.
Given this, is France’s strategy aligned with this vision, or does it risk creating an isolated, ineffective approach that is incompatible with the European architecture currently being developed? If the goal is truly digital sovereignty, wouldn’t it be better to invest in infrastructure solutions compatible with this European vision, rather than multiplying national initiatives that could, in the long run, be called into question by changes in the European framework?
However, if the goal is to strengthen digital sovereignty, the priority should not be the large-scale construction of data centers, but rather the development of a sovereign digital ecosystem that incorporates appropriate infrastructure, disruptive technologies, and a strategy for optimizing existing resources.
Relying solely on physical infrastructure without making significant investments in research, innovation, and technological expertise risks reinforcing a model in which France would merely store data without extracting its true economic and strategic value.
Furthermore, AI does not rely solely on owning data centers, but on access to computing resources and data. Cloud computing has revolutionized business models by enabling companies, both large and small, to access computing power on demand, thereby avoiding heavy investments in physical infrastructure. Today, the bulk of the value in AI lies not in servers, but in software architecture, algorithm optimization, and data governance. In other words, a country that focuses on optimizing and intelligently leveraging its data will gain a decisive advantage over one that merely stacks servers.
The real solution, therefore, does not lie in a fruitless dichotomy between AI and data centers, but in building a hybrid model in which appropriate infrastructure is combined with resource optimization and the development of skills among economic actors.
It is imperative that France’s strategy not be limited to a mere digital real estate project, but that it incorporate an ambitious macroeconomic vision capable of positioning the country as a creator of value rather than merely an infrastructure operator.
A model that consumes excessive energy and is incompatible with environmental goals
At a time when environmental issues have taken center stage, the proliferation of data centers makes no ecological sense. These facilities consume vast amounts of energy and natural resources, particularly water for cooling. The exponential growth of data centers is causing major environmental problems, particularly in terms of energy consumption, CO₂ emissions, and the increasing vulnerability of national power grids.
The government claims that these new data centers will be powered by clean energy, thereby reducing their carbon footprint. Certainly, the use of renewable sources such as solar, wind, or nuclear power could limit the environmental impact of these facilities. However, this approach solves only part of the problem. On the one hand, clean energy production remains insufficient to meet the exponential demand from data centers without affecting other strategic sectors. On the other hand, even when powered by low-carbon energy, these data centers remain extremely water-intensive for cooling and require massive infrastructure to maintain their performance.
Rather than piling up servers and consuming ever-increasing amounts of energy, there is an urgent need to rethink the energy efficiency of AI models. This involves optimizing algorithms, making better use of smart data, and utilizing specialized processors (TPUs, low-power GPUs) designed to minimize energy consumption. Decentralization, combined with embedded AI, offers a far more efficient alternative: it allows calculations to be performed as close as possible to users and connected devices, thereby avoiding unnecessary data exchanges with remote data centers and reducing their environmental impact.
Finally, one question remains: Is France truly committed to a sustainable and innovative AI model, or is it simply trying to justify a massive investment in infrastructure that, despite the clean energy argument, remains energy-intensive and could become obsolete in the medium term?
A contradiction with France's AI strategy
France’s artificial intelligence strategy, embodied in particular by the France 2030 Plan, aims to make the country a leader in edge AI by focusing on decentralized, lightweight technologies optimized to run directly on local devices and infrastructure. Edge AI prioritizes local computations on smart devices (smartphones, connected objects, vehicles, industrial machines, etc.), without relying on massive data centers. This approach aligns with industrial priorities across sectors such as automotive, aerospace, healthcare, and defense. For example, an autonomous vehicle cannot rely on a data center hundreds of miles away to make a decision in real time. It must be capable of processing data immediately, using optimized algorithms integrated directly on board.
Yet today, the government is heavily promoting the development of centralized data centers—an approach that runs counter to the very principles of edge AI. These heavy, energy-intensive, and costly infrastructures perpetuate a centralized computing model that stands in opposition to the efficiency and flexibility inherent in edge AI.
We don’t know what to think anymore: France has a clear ambition to become a leader in edge AI, yet at the same time, it is investing heavily in centralized data centers, which run counter to that very strategy. Is this a sudden reversal, a lack of coordination among decision-makers, or simply a complete lack of coherence? Have the implications of these choices been properly considered, or are we witnessing a policy being pursued without a clear vision? This contradiction raises serious questions about the coherence and thought process behind these decisions, which seem to be slowing down the transition to decentralized AI and increasing our dependence on cloud giants rather than reducing it.
The Incompatibility of Data Centers with Quantum Computing
Massive investment in data centers is not only a strategic mistake in light of the rise of embedded AI, but it is also fundamentally incompatible with the rise of quantum computing.
Quantum computing is not merely an evolution of classical computing, but a profound paradigm shift that completely transforms the way we process information. Unlike the centralized architectures of data centers, it enables complex calculations to be performed in a fraction of the time and with a fraction of the energy required. A single quantum processor can complete tasks in a matter of minutes that would take several weeks using classical infrastructure.
Investing heavily in data centers today amounts to betting on technology that could become obsolete in the medium term, particularly with the rise of quantum computing. Recent advances in this field, such as Google’s “Willow” chip, unveiled in December 2024, which can solve problems in a matter of minutes that would take current supercomputers millions of years to process, illustrate the rapid pace of progress. Furthermore, French companies such as Quandela have presented ambitious roadmaps, planning to network quantum computers by 2028.
These developments raise a fundamental question: why is France investing heavily in infrastructure whose relevance may soon be called into question? Is this a short-term gamble, while waiting for quantum technology to reach industrial maturity? Is this a decision driven by political and economic considerations, aimed at revitalizing a struggling sector? Or is it simply a belated reaction, an attempt to make up for lost ground, but with an approach that fails to truly account for the technological trends of tomorrow? One might also wonder if this is not the result of repeated changes in government and a lack of coordination, with divergent visions between those advocating for embedded AI and those still banking on centralized infrastructure.
In any case, it is essential to clarify this strategy, as continuing down this path could lead to massive investments that quickly become obsolete, while other countries gain the upper hand with more flexible models better suited to emerging technologies.
Who will benefit from these billions invested in data centers?
The announcement of these massive investments raises a key question: who will actually benefit from these investments in data centers? Contrary to what one might think, these facilities will not primarily benefit the French tech ecosystem, but rather a handful of well-established players—mostly foreign—who control the bulk of the market for semiconductors, cloud infrastructure, and data center management software.
Unsurprisingly, the primary beneficiaries will be American cloud computing giants such as AWS (Amazon Web Services), Microsoft Azure, Google Cloud, and Oracle, which already provide the technological backbone for the majority of data centers around the world. They will be the ones selling software licenses, management solutions, artificial intelligence platforms, and hosting services, thereby capturing a significant share of the revenue generated by these infrastructures.
But beyond the cloud, we must also consider semiconductor and hardware manufacturers, which are essential to the construction and operation of these data centers. Today, the market is overwhelmingly dominated by Nvidia, Intel, AMD, and Broadcom, which design the processors and GPUs essential to the computing power of modern data centers. Nvidia, for example, has become a key player in the field of specialized artificial intelligence chips, and every new data center built represents a financial windfall for this American company, whose solutions are unmatched in the market. Added to this are suppliers like Cisco and Dell—also American companies—which equip data centers with servers and communication networks.
By focusing public and private resources on developing this infrastructure, France is simply increasing its dependence on these foreign giants. Rather than fostering the emergence of local players capable of offering competitive alternatives, this strategy locks in a market that is already dominated, where added value, intellectual property, and the most lucrative margins largely elude French and European companies…
Investing in data centers is like building coal-fired power plants in the age of renewable energy.
Betting on massive data centers in 2025 is like investing in a fleet of airships in 1940, just as jet aviation was about to revolutionize everything. The future belongs to those who optimize and innovate, not to those who stack servers. The AI of the future will be hybrid, frugal, and decentralized. France must focus on algorithmic intelligence, energy efficiency, and the autonomy of AI systems, while ensuring sovereign data protection through appropriate and controlled infrastructure.
What should France do?
If France wants to remain competitive and maintain its sovereignty, it must absolutely avoid this misguided path and focus its efforts on optimizing existing infrastructure, developing embedded AI, and accelerating research in quantum computing. It must support companies specializing in optimized processors, invest in quantum research centers, and build a sovereign, hybrid, and intelligent cloud. Priority must be given to hybrid architectures that combine advances in embedded AI, edge computing, and quantum computing.
It is also essential to reorient investments by emphasizing the added value that AI can bring in terms of productivity, innovation, and societal impact, rather than focusing on the scale of infrastructure.
Support for Businesses and AI Deployment
The French business sector, composed mainly of small and medium-sized enterprises, lags significantly behind in the adoption of AI (see the findings of theaivancity Observatory on Responsible AI Practices). To bridge this gap, it is imperative to establish targeted support programs designed to assist these companies in integrating AI solutions tailored to their specific needs. This could include grants, tax credits, or tax incentives to encourage the implementation of AI technologies. Furthermore, the creation of collaborative platforms would facilitate the sharing of resources and knowledge among businesses, research centers, and academic institutions, thereby fostering open innovation.
Training and Skills Development
Education is a cornerstone of AI development in France. Although announcements have been made—such as the goal of training 100,000 experts and incorporating AI education into high schools—these initiatives are struggling to translate into concrete action due to a lack of sufficient investment.
But the issue of training cannot be limited to AI experts alone. Artificial intelligence is now a cross-disciplinary skill that is essential in all academic and professional fields. It is no longer conceivable for doctors, lawyers, architects, designers, managers, financiers, or engineers to practice their professions without at least a basic understanding of AI tools and their impact on their respective fields.
AI is redefining professional practices, transforming decision-making processes, and changing the skills required in the job market. It is therefore essential to incorporate broad-based AI training into all academic disciplines, so that every student, regardless of their major, can acquire the necessary foundations to understand, use, and leverage these technologies. This involves not only creating specific courses on the uses and challenges of AI, but also adopting an interdisciplinary approach, where AI is integrated into existing programs in law, medicine, social sciences, finance, the arts, engineering, and more.
Without this fundamental transformation of higher education, France risks ending up with a small technological elite that masters AI, while the rest of the workforce will be dependent on it without understanding either its mechanisms or its implications. Training experts is essential, but educating an entire society capable of understanding and leveraging AI is a far more strategic challenge.
It is therefore crucial to finally allocate significant resources to developing ambitious training programs, both at the secondary and higher education levels, as well as reskilling initiatives for workers whose jobs are likely to be transformed by AI. Collaboration with private-sector partners can also help address the skills gap.
Promoting Research and Innovation
France has a high-quality AI research ecosystem, but it is essential to strengthen the links between academic research and the industrial sector. This involves supporting collaborative projects, funding innovative startups, and promoting technology clusters specializing in AI. Furthermore, the implementation of supportive regulations and an environment conducive to experimentation would encourage innovation while ensuring the ethical and responsible use of AI.
By focusing its efforts on these strategic priorities, France will not only be able to catch up in the adoption of AI, but also position itself as a leader in the development of innovative, sustainable solutions that deliver significant value to society.
A pivotal moment: investing in talent, not just in concrete
But the announcements made at this summit must not amount to yet another misguided technological promise, such as the massive investments announced for data centers. If we continue to prioritize costly, energy-intensive infrastructure without a genuine strategy for building skills and adopting AI across the entire economy, then this summit will be nothing more than a missed opportunity.
If these announcements are limited to the same old promises—ambitious but vague figures on training, development goals without a clear implementation plan, and a glaring lack of engagement from all stakeholders in the AI ecosystem—then this initiative will only repeat past mistakes. After having focused on infrastructure and servers, it is imperative that France finally focus on talent, applications, and innovation. Without this, the summit will be nothing but a massive disappointment—a missed opportunity to make AI a true economic and societal transformation, rather than merely an infrastructure issue.
It is no longer enough to simply state ambitions or announce impressive figures regarding AI training and development. What is missing today is a genuine, structured implementation backed by resources commensurate with the stakes—one that both fosters an environment conducive to innovation and widely disseminates AI expertise across all sectors of society. AI won’t wait. Other countries are moving quickly and aren’t content with mere rhetoric. France must make this moment a true strategic turning point, or risk remaining a spectator to the revolution that is already shaping the future…