The European Union is advancing its competitiveness and innovation capacity through the creation of AI Factories and AI Gigafactories, designed to support the development and training of Europe’s complex AI models. The four Southern European countries in the PromethEUs group (Spain, Portugal, Italy, and Greece) are hosting or participating in some of the first factories launched under the EU’s Digital Europe Programme and EuroHPC initiatives.
Building large-scale infrastructure is the first step, but past European tech cycles show that hardware alone does not guarantee competitiveness. The real impact comes when this power is embedded across companies, labs, and public institutions. High Performance Computing (HPC) needs to be integrated within the wider economic fabric.
What Are AI Factories and Gigafactories?
AI Factories are computing and data hubs embedded in Europe’s most powerful supercomputers. Their function is to provide compute power, datasets, and technical assistance for startups, companies, researchers, and public administrations. The design is to allow different sectors to experiment with AI without outsourcing compute or depending on foreign platforms.
AI Gigafactories are still in planning. The European Commission has expressed 67 expressions of interest from 16 EU countries. A gigafactory’s purpose is to consolidate the compute capacity required for large foundation models, installing over 100,000 GPUs per facility and enabling training workloads that until now are only feasible in the US or China. They are significantly more expensive: the European Commission aims to fund them through blended financing mechanisms under InvestEU and InvestAI, with around €20 billion earmarked to support up to five AI Gigafactories, as part of a broader €200 billion investment goal for European AI.
Southern Europe’s Compute Capacity and its Deployment
Spain now hosts one of the EU’s top-tier supercomputers, MareNostrum5, and co-leads one of the first AI Factories in Europe.
Portugal has launched Deucalion (a national supercomputer in Guimarães) and is formally linked to the Barcelona AI Factory through a shared interface.
Italy is hosting the AI Factory at the Bologna Technopole, built around the Leonardo supercomputer, one of the five fastest in the world.
Greece, through the GRNET center in Athens, is setting up the Pharos AI Factory.
These are important infrastructure projects, but compute on its own is not enough. One of Europe’s main assets is its entrepreneurial and industrial fabric. If AI is to be developed and deployed in line with European priorities, it must be integrated into this fabric.
A more relevant indicator than infrastructure is the actual use of AI by enterprises. The new investments in compute infrastructure need to be deeply integrated within Europe’s industrial fabric, so that AI innovation comes from companies and labs that are working closely with the AI factories and gigafactories.
Two Challenges: Capital and Services
We are in the early phase of a technological revolution: focused on building infrastructure and testing the capabilities of new systems. The creation of AI Factories and Gigafactories fits squarely within this phase. However, the competitive advantage is gained when that capacity is broadly accessible and systematically used across the economy.
The challenge ahead is making sure that a wider range of European actors can participate in the development of advanced AI systems, experiment with foundation models, and shape applications at the technological frontier. This requires a more deliberate effort to connect infrastructure with talent, regulatory clarity, and long-term funding instruments.
- Stable private and public investment
Advancing into the next phase requires stable investment flows throughout the innovation cycle, which would include funding in all stages: exploratory R&D, deployment and scale. In 2023, AI startups in the EU raised around €8 billion, compared to nearly €60 billion in the United States. Within Europe, Southern countries remain significantly underfunded, with limited access to growth capital for deep tech ventures.
In order to make the new or existing funding mechanisms effective (InvestAI, Horizon Europe, EIB), national governments will need to act as co-investors, helping to de-risk frontier research, support testing environments, and accelerate private sector involvement.
- Infrastructure Needs a Services Layer to Deliver Value
A Gigafactory with 100,000 high-end chips does not guarantee that companies or labs will build advanced AI. What connects compute to the economy is the services layer:
- User interfaces that simplify access so teams with different technical levels can work without friction.
- Scheduling systems that distribute workloads efficiently, preventing long queues and maximizing use.
- Data tools to manage and move large or sensitive datasets across projects safely.
- Technical support teams that help smaller firms integrate their data and workflows.
The risk is building the hardware first and figuring out these systems later. Ideally, services and infrastructure should be designed together so everything is aligned from the start.
References:
Dealroom. (2025). Startups backed by the EU’s Framework Programmes. https://dealroom.co/reports/startups-backed-by-the-eus-framework-programmes
European Commission. (2024). AI factories under the Digital Europe Programme. https://digital-strategy.ec.europa.eu/en/policies/ai-factories
Eurostat. (2025). Usage of AI technologies increasing in EU enterprises. https://ec.europa.eu/eurostat/web/products-eurostat-news/w/ddn-20250123-3
Perez, C. (2002). Technological revolutions and financial capital: The dynamics of bubbles and golden ages. Edward Elgar.