Digital Sovereignty: Why Europe Is Not Winning the Cutting-Edge Models Race
European digital sovereignty is necessary to ensure security in Europe, whether in the public or private cyberspace. Indeed, the Old Continent suffers from structural disadvantages in its competition with the United States and China in the field of AI, particularly when it comes to pursuing a strategy based on the race for scale and raw power1. First, European venture capital lacks the investment capacity of U.S. funds or the Chinese government. Second, Europe has more limited energy resources and imposes stricter emissions targets on itself, which complicates the deployment of energy-intensive data centers. Finally, as we highlight in the SKEMA Publika report Artificial intelligence and European sovereignty, the European market remains fragmented, which complicates the implementation of a coherent strategy and undermines Europe’s ability to compete on a global scale.
What is “Alternative AI”?
The term “alternative AI” refers to a political and technological shift already taking shape in Europe that aims to bridge the digital sovereignty gap. It is that of Europe, which should not compete exclusively in the race for massive frontier AI models, but focus on alternative AI systems tailored to local needs. European companies do not need powerful AI on a raw scale, but alternative solutions that are cheaper and better prepared to meet the demand of the European ecosystem, whether it is agriculture, services, finance, industry or the public sector.
“Alternative AI” is AI focused on deployment and governance rather than on model size, combining several building blocks. The first is small language models (SLMs), which are better suited for specific use cases and local hosting. The second building block relates to the multimodal dimension; it is AI that processes multiple types of data (text, images, video, etc.) and not just text. Open-weight models represent the third building block; these are models whose parameters are accessible. The fourth building block is Edge AI, where processing takes place on the device, at the industrial site, in the hospital, or near the data source, rather than systematically in the cloud. The fifth building block is RAG (Retrieval-Augmented Generation), which enables the AI model to connect to reliable sources such as internal documents, a regulatory database, and so on. The final building block involves implementing a portfolio of models rather than using a single AI model for all of an organization’s needs.
Beyond the need to innovate, the real challenge lies in the adoption gap
Europe is lagging behind in terms of AI adoption by businesses, which makes it all the more crucial to take steps to develop its digital sovereignty. While AI adoption is certainly progressing rapidly, the actual benefits observed remain limited and uneven. In early 2026, 43% of U.S. employees reported using generative AI as part of their work, compared to 26% in Italy and 36% in the United Kingdom. Its use was also less widespread in Germany, France, and Italy2. In the eurozone, 38% of companies estimated that they had reached an advanced level of adoption, while 33% were still at the stage of occasional or experimental use3. Thus, beyond the need to develop an alternative AI, these figures highlight a priority: accelerating the integration of AI into concrete processes, particularly in sectors where European companies already possess solid industrial expertise.
Where Europe Has a Real Advantage: Specialized Deployment of AI
Cutting-edge AI is characterized by significant economies of scale, concentrated computing power, and highly polarized ecosystems, all areas in which Europe starts at a disadvantage. Thus, Europe’s most credible strategy does not lie in directly imitating the American or Chinese giants of cutting-edge AI, but rather in the specialized deployment of AI systems tailored to sectors where it already possesses institutional, industrial, and regulatory strengths.
Industry, healthcare, finance, and the public sector all share high demands for reliability, traceability, confidentiality, interoperability, and domain expertise. In these areas, the value of AI depends less on the raw size of the model than on its ability to integrate into real-world processes, leverage reliable sector-specific data, and operate within a robust governance framework. Thus, Europe could gain an advantage by coordinating its efforts as a collective of stakeholders strategically adopting AI, provided it maintains sufficient capabilities in cutting-edge AI to evaluate the best systems and adapt them to its own needs.
Sovereign Data: The “New Working Capital” of Enterprise AI
Proprietary data is not the “new oil,” but rather the “new working capital” of enterprise AI, that fuels processes and supports business activities. As a result, RAG (Reasoning, Argumentation, and Governance), multimodal retrieval, data spaces, and assistants based on domain-specific knowledge make it possible to transform institutional knowledge into directly actionable resources.
The European Data Governance Regulation and initiatives related to the Data Union are therefore of particular importance: they can reduce barriers to the rapid mobilization of data, under conditions that are both secure and legally compliant.
AI Act and Governance: Making Trust a Production Technology
The European AI Regulation should not be viewed merely as a compliance requirement, but as a production infrastructure. The Data Union Strategy combines greater European control with openness toward trusted partners, under fair and secure conditions. A viable approach would involve promoting reliable cross-border data flows, documented traceability, open and reusable components, local deployment where necessary, and sufficient computing capacity to limit critical dependencies.
The European Commission’s 2025–2026 agenda outlines a response: AI factories and future AI gigafactories, an action plan for the AI continent, a strategy for the Data Union, sector-specific deployment programs, and the governance framework of the AI Regulation. These initiatives could lay the groundwork for a European strategy centered on adoption and deployment, even as the public debate remains largely focused on cutting-edge models.
Nevertheless, significant counterarguments remain: Europe risks not investing enough in cutting-edge capabilities; open models do not eliminate its dependence on American or Asian hardware; and a growing gap in quality or ecosystem could reduce the benefits derived from its deployment expertise.
The paper’s 4 recommendations for European decision-makers
1. Make AI factories truly dedicated to deployment
2. Use public procurement as a lever for building trust
3. Help companies transition from purchasing a single model to managing a portfolio of models
4. Share validation infrastructure in sectors with high trust requirements
Digital sovereignty refers to the capacity of a state – or, in this case, the European Union – to act in cyberspace to enhance its autonomous capacity for decision-making and organization, action, and assessment, while ensuring the security of government data.
Sovereign AI is an artificial intelligence system that is entirely controlled and operated by the country, region, or organization that built it. This means it has control over the data, infrastructure, and models.
European artificial intelligence is not as advanced as that of the United States and China due to limited investment and dependence on foreign infrastructure, as well as strict regulations that may hinder deployment in the short term.
The AI Act is the European regulation for the development and market advancement of the use of artificial intelligence systems in the EU, reducing at the same time risks for citizens such as manipulation, exploitation of vulnerabilities or discrimination based on biometric or social criteria.
A small language model (SLM) is a model designed to require significantly less computing power, offers lower latency, or requires less hosting capacity than state-of-the-art, very large-scale systems, particularly for specialized or on-premises deployment.
- Bick et al., Differences in AI adoption in Europe and the US (CEPR, 2026). ↩︎
- AI Office publishes frontier AI expert findings on EU competitiveness, sovereignty and security. (2026). Shaping Europe’s Digital Future. https://digital-strategy.ec.europa.eu/en/library/ai-office-publishes-frontier-ai-expert-findings-eu-competitiveness-sovereignty-and-security. ↩︎
- Ferrando et al., Adopting and investing in AI: evidence from euro area firms in the SAFE (ECB, 2026). ↩︎
