EU strategy paves way for Raise, a distributed CERN for AI across Europe

EU AI in Science Strategy sets up Raise: a distributed "CERN for AI"

The European Commission will adopt its AI in science strategy on October 7, announced at Research and Innovation Days in Brussels on September 17. The plan lays the groundwork for Raise - the Resource for AI Science in Europe - a distributed infrastructure to accelerate AI in science and science for AI.

Raise will pool data, infrastructure, talent, and funding. "Through the different interactions we had with the scientific community, we thought the best way forward is to have networking of what exists in Europe and strengthening of what exists, instead of creating new infrastructures," said Maria Cristina Russo, director for prosperity at the Commission's research and innovation directorate general.

Why a distributed "CERN for AI"

Ursula von der Leyen promised a "CERN for AI" in July 2024. The Commission's advisors backed a distributed model to ensure broad access. "All parts of Europe would be included and have transparent and easy access," said Nicole Grobert, chair of the expert group.

Raise will connect existing sites, including the EU's AI factories and upcoming AI gigafactories, rather than build a single, central hub.

AI factories now, gigafactories next

Thirteen AI factories have already been selected across Europe. These sites are building ecosystems around AI-optimized supercomputers, providing compute, data, and skills.

"Now, our task is to make sure that the customers and researchers that are coming end up in the right AI factory [...] but also to have a good knowledge of the other large AI infrastructures in the EU," said Petra Dalunde, co-director of the Swedish AI factory.

The Commission will launch a call for AI gigafactories - specializing in training very large AI models - toward the end of the year.

What's next

• October 7: expected adoption of the AI in science strategy.
• November 3-4 (Copenhagen): inaugural Raise event with more details on governance and access.
• Late 2024: official call for AI gigafactories.
• Within two weeks of the Brussels event: adoption of the "apply AI strategy" to support new industrial and public sector uses, confirmed by research Commissioner Ekaterina Zaharieva.

What this means for research leaders

• Map workloads to AI factory capabilities: HPC-enabled model training, data access, and support services vary by site.
• Prepare for gigafactory calls: define use cases for very large model training, identify partners, and outline data provenance and governance.
• Budget for compute and storage: include persistent storage, model checkpoints, and data egress in proposals.
• Strengthen data pipelines: prioritize standardized formats, documentation, and reproducibility to move fast across sites.
• Build consortiums early: cross-border access is central to Raise; align on IP, ethics, and publication policies now.
• Upskill teams: ensure your lab can use distributed HPC clusters, GPUs, and MLOps workflows.

Access and collaboration

Raise is intended to make access transparent across Europe and reduce fragmentation. Expect clearer entry points via AI factories, with routing to the best-fit site and cross-site collaboration.

For event context, see the Commission's Research and Innovation Days information page: Research & Innovation Days.

If you're planning training for research teams, explore role-based programs: AI courses by job.