Topics
Europe is home to countless organisations, research institutes and industry associations working on the future of artificial intelligence. There is no shortage of conferences, declarations or strategy papers. But which initiatives can genuinely influence the direction of AI development? These five could help determine Europe’s position in the technological and economic competition of the next decade.
A citizen tells a digital service agent what they need to get done. The AI interprets the request, retrieves the relevant information and guides them through the next steps. Meanwhile, in a healthcare collaboration, institutions jointly develop AI models using their respective datasets without pooling sensitive patient records in a central database.
These capabilities are no longer years away. E-Group is already demonstrating working solutions: government stakeholders are being shown an operational service agent, while healthcare organisations are already using technology that allows data to remain where it is held. The company regularly showcases these capabilities in live demonstrations.
The question is therefore no longer simply whether this can be done. It is who owns the system, where the data is processed, who oversees its operation and, ultimately, who earns the revenue it generates.
The United States and China recognised years ago that AI is not just another IT product. It is a major source of economic and geopolitical influence. US technology giants dominate the most important global cloud services and AI platforms, while China is committing substantial public and private resources to building its own technology ecosystem.
For Europe, the answer is not necessarily to replace every American or Chinese solution. It is to avoid having to buy every strategically important technology from external providers.
Dependency is not merely a data protection issue. It is also an economic one. Without credible alternatives, customers have limited influence over pricing, service terms and the direction of technological development. Moreover, a substantial share of spending on AI accrues to technology providers—not necessarily to the places where the data was generated or where the work is performed.
Recognising this, Europe has launched several substantial initiatives. The following five differ in scale, structure and legal status. They include an EU office, major investment programmes, industrial collaborations and a specific technology project. What they share is the potential to deliver much more than another report on AI.
1. IPCEI-CIS and 8ra: A European Alternative Alongside Global Cloud Providers
One of Europe’s most important efforts to build technological autonomy operates under a less-than-memorable name. IPCEI-CIS—the Important Project of Common European Interest on Next Generation Cloud Infrastructure and Services—received EU approval in 2023. It launched with the direct participation of 19 companies from seven member states, including Hungary, backed by up to €1.2 billion in national public funding expected to mobilise a further €1.4 billion in private investment.
IPCEI programmes enable Member States to support strategically important cross-border projects that would be unlikely to proceed at the necessary scale or pace under normal market conditions. The objective is not simply to give one company a competitive advantage. It is to build shared technological capabilities whose benefits extend across the European economy.
The 8ra initiative, built around IPCEI-CIS, now connects more than 120 partners across 12 EU member states. Its central question is straightforward: can European cloud providers, data centres and computing systems closer to users operate as one large, interoperable network?
For a large enterprise, moving applications and data between cloud providers can be a significant undertaking. Systems are often incompatible, migration is expensive and switching providers can introduce business risk. Using open technology standards and a common architecture, 8ra aims to change that. It seeks to create European infrastructure in which data and applications can operate across different providers’ systems while the organisations responsible for them retain control.
Tangible results emerged in 2026. At Mobile World Congress in Barcelona, five major telecoms operators—Deutsche Telekom, Orange, Telefónica, TIM and Vodafone—jointly demonstrated a coordinated, cross-border European edge cloud. This does not amount to a full commercial rollout, but it is an important step from architectural concepts towards operational services.
Hungary is not simply watching from the sidelines. Antal Kuthy, E-Group’s founder and CEO, is a member of the IPCEI-CIS European industrial steering body and co-leads its third workstream, which focuses on advanced data processing. E-Group also plays a leading role in AI development within this workstream.
The company’s European role is backed by working technologies. COIOS is a sovereign agentic AI platform designed to give enterprises and institutions control over how they deploy and use AI. FedX, already used in several international projects, enables organisations to develop AI models collaboratively without transferring their raw data to a central system.
The Ambiti8n pilot, developed with E-Group’s participation and leadership, integrated technology components from eight organisations in four countries into a working application. This integration work has helped establish the company as a recognised technology contributor within IPCEI-CIS.
What to watch
The next test is whether successful demonstrations become widely available, commercially competitive services. If they do, technological autonomy for European businesses need not mean rejecting foreign providers. It could mean having a genuine choice.
2. EuroHPC and AI Factories: No AI Powerhouse Without Compute
Developing AI takes more than strong algorithms and talented researchers. It also requires substantial computing power, which remains concentrated in relatively few hands. One of the constraints on European AI development is its reliance on non-European technology companies for much of the infrastructure needed to train large models.
The EuroHPC Joint Undertaking—a supercomputing partnership involving the EU, European countries and private partners—is working to address that gap. Originally focused on supercomputing infrastructure for scientific and industrial applications, it has become a central pillar of Europe’s AI strategy.
Its most visible development is the network of AI Factories. The name can be misleading: these are not factories producing robots. They are supercomputing hubs that provide computing capacity, data management support and specialist expertise for AI development.
By 2026, the EU programme had selected or begun developing 19 AI Factories and 13 associated regional access points, including one in Hungary. The aim is to give European start-ups, research institutions and smaller businesses access to infrastructure that would otherwise be beyond their reach—not just to serve the largest and best-funded technology companies.
The next step could be AI Gigafactories. These proposed, much larger facilities would support the development and training of the most advanced, compute-intensive models. In 2026, the EU is working to prepare the establishment of up to seven such facilities, seeking to mobilise public and private investment on a scale of tens of billions of euros.
Their significance, however, is not simply the size of the data centres. For a European pharmaceutical researcher, automotive developer or industrial AI company, the breakthrough would be access to appropriate computing resources without depending on a global technology giant’s commercial priorities.
European ownership of infrastructure does not, by itself, deliver complete independence. Non-European manufacturers remain dominant in high-performance AI processors. The success of AI Factories will therefore also depend on Europe’s ability to build its own software, models, expertise and commercially viable services around the hardware.
What to watch
The key measures are how many businesses actually gain access to computing resources, what new technologies emerge and whether the facilities support enterprise applications as well as research. The number of processors installed is not, in itself, an economic outcome.
3. IPCEI-AI: Building a European AI Industry
Cloud infrastructure and computing capacity are necessary, but they are not sufficient. Owning data centres does not guarantee that European companies will develop the AI systems running on them—or that the commercial value generated by those systems will remain in Europe.
The proposed IPCEI-AI programme aims to address this gap. In September 2026, 18 European countries signed a joint manifesto setting out their intention to develop a new cross-border AI initiative. It is not yet an approved, funded IPCEI. It remains in preparation. Even so, its proposed objectives signal a shift in European thinking.
Strategically, IPCEI-AI would build on the foundations laid by IPCEI-CIS. Alongside European cloud and data processing infrastructure, it would prioritise the development of European AI technologies and industrial applications that run on that infrastructure. The two programmes are separate, but they target complementary capabilities.
E-Group is drawing on the experience and relationships it has developed through IPCEI-CIS. As part of the preparations for IPCEI-AI, it has entered into collaboration with 30 major European companies. The Hungarian company is therefore helping shape the new initiative not simply as a potential user, but as a developer with its own technology and an established European partner network.
The programme is not merely about producing European versions of popular chatbots. Its primary focus would be industrial and sector-specific AI: manufacturing automation, robotics, logistics, transport, energy, healthcare and other strategic fields. It would support models and intelligent systems built on European companies’ technologies, data and expertise.
This could be one of Europe’s genuine competitive strengths. Winning the race to build the most popular general-purpose language model is not the only route to success. European industry, healthcare, energy and public administration hold substantial expertise and data accumulated over decades. The question is whether Europe can turn those assets into intelligent systems that customers will buy.
One of IPCEI-AI’s key ambitions is to develop distributed, interoperable AI technologies. Rather than concentrating all computing capacity and data in a single central system, services from different European providers could work together while those providers retain their independence. This could closely complement the infrastructure being developed through 8ra.
What to watch
Watch national project selection processes, funding approvals and the first industrial applications. Some projects could begin in 2027, but actual investment and delivery will also depend on programme approval. The decisive test will be whether the collaboration can turn European research into competitive products and businesses.
4. HeliX and the European Health Data Space: Bringing AI to the Data
Discussions about AI often treat data as a raw material that can simply be collected and moved wherever it is needed. In practice, some of the most valuable data cannot be handled that way.
Hospital records, banks’ customer data and industrial production information are sensitive and subject to legal or commercial protections. Training shared AI models on data from several institutions may offer considerable scientific and economic benefits. Unrestricted data sharing, however, is not a viable approach.
HeliX—Health Ecosystem for Learning and Innovation Across Sovereign Data Space—is seeking a practical solution. Launched in 2025, the three-year European project has a budget of approximately €9.15 million and brings together 14 partners from Portugal, Italy, Poland and Hungary. Health Cluster Portugal coordinates the programme, with E-Group’s technologies contributing to its technical foundations.
HeliX is not starting from scratch. Several Hungarian medical universities already use E-Group’s related technology. The international project is therefore building on solutions already in use to create a healthcare collaboration environment that can be adopted more widely.
One of its foundations is federated learning: the AI model goes to the data, rather than the other way around. Training or analysis can take place locally within participating healthcare institutions, without transferring sensitive patient records to a central system. E-Group’s FedX technology supports this form of collaboration across institutions while allowing data to remain in place.
The approach does not automatically eliminate data processing risks, but it opens up a fundamentally different way to use sensitive data securely.
The potential benefits are substantial. Drawing on the experience of hospitals in different countries could support more accurate diagnostic tools, better risk assessment models and more effective drug research, without requiring large-scale transfers of patients’ personal data across borders.
The European Health Data Space provides the wider context. EU legislation that entered into force in 2025 will progressively establish new arrangements for the regulated, cross-border use of health data, primarily from 2029 onwards. HeliX is a separate project, but it aligns with those objectives by seeking to prepare a technology solution for broader commercial deployment.
Its budget is not comparable to the investments behind IPCEI or EuroHPC. Yet programmes focused on specific applications can reveal whether Europe’s broader strategies work in practice.
What to watch
The key questions are what commercially deployable results HeliX delivers by 2028 and how it connects to Europe’s emerging health data economy. Its potential relevance extends beyond healthcare: the approach could also be applied in other industries working with sensitive data.
5. The European AI Office: Setting the Rules of the Game
The first four initiatives primarily build technology, infrastructure or market opportunities. The fifth has a different role: shaping the conditions under which AI can be developed and used in Europe.
The European AI Office, part of the European Commission, is a central institution in the EU’s AI regulatory framework. Its responsibilities include supporting the implementation of the AI Act, adopted in 2024, supervising providers of general-purpose AI models and helping ensure consistent application of European rules.
Since August 2026, it has had substantive enforcement powers. It can request technical documentation from general-purpose AI model providers, assess compliance and initiate measures where necessary. Working with national authorities, it also helps enforce requirements for AI transparency and safety.
But the AI Office is more than an enforcement body. It also has a role in supporting AI development and adoption, contributing to guidance, codes of practice and cooperation mechanisms, and working alongside European AI development programmes.
Its task involves a difficult balancing act. Europe needs to protect people from AI-related risks while enabling its businesses to develop competitive technologies. Excessive administration and unclear requirements can inhibit innovation. Rules that are too permissive, however, can undermine trust, enable abuse and expose society to significant harm.
Human-centred AI is not simply an ethical aspiration. It is an economic issue too. Companies need to understand what data a system uses, who is accountable for errors, when a person can intervene and how automated decision-making can be reviewed. These questions become particularly important when AI moves beyond recommendations and begins executing tasks.
What to watch
Watch how the rules are applied in practice, how AI models are assessed, what compliance costs businesses face and how new safety requirements develop. The AI Office’s work will help determine whether Europe becomes an attractive place to develop trustworthy AI—or a cautionary example of excessive regulation.
Europe Does Not Need to Build Everything Itself
These five initiatives address different parts of the same challenge. EuroHPC builds computing capacity. IPCEI-CIS and 8ra lay the foundations for interoperable infrastructure. The proposed IPCEI-AI would support the development of European AI technologies. HeliX explores a practical route to using sensitive data. The AI Office shapes the rules for safe, accountable operation.
None will solve Europe’s competitiveness challenges on its own. Certainly not if new infrastructure remains underused, research fails to produce commercial products and businesses continue to find it easier to buy every solution off the shelf from global technology providers.
Success requires more than national and EU funding. It requires customers willing to buy European-developed technologies. The next phase will therefore be as much about commercial growth, enterprise partnerships and public procurement as it is about technology research.
The Hungarian example shows that this does not necessarily require waiting for tomorrow’s technology. COIOS already offers operational AI capabilities for enterprises and institutions in a controlled environment, while FedX provides an approach already in use for collaborative work with sensitive data. The next step is to extend these capabilities across as many processes as possible, with measurable results and clear accountability.
This is not an argument for Europe to build technological walls around itself. Giving up the benefits of global specialisation would make little sense. American and Asian technologies can continue to play important roles in the European economy. Genuine sovereignty does not require Europe to make everything itself. It requires credible choices—and the ability to avoid dependence on any single technology provider or country.
Those choices can have economic value even if European alternatives do not become market leaders. Credible competition alone can strengthen customers’ negotiating positions, put downward pressure on prices and encourage innovation.
There is, however, a larger issue at stake. AI does more than reduce costs and improve productivity. It is becoming part of economic activities that previously depended on human labour. If ownership of the enabling technologies—and most of the revenue they generate—moves elsewhere, Europe could find itself adopting AI while retaining only a limited share of the value it creates.
The real question in Europe’s AI race is therefore not whether we will have our own ChatGPT. It is whether Europe will primarily be a customer in the AI economy—or also an owner of the technology and the value it generates.
The answer will not come from the next conference in Brussels. It will come from the investments, operational systems and business results delivered over the coming years.