[Brussels, 9 October] – The European Commission unveiled its long-awaited AI in Science Strategy, described by Executive Vice-President Henna Virkkunen as a step towards the EU’s commitment to creating a ‘CERN for AI’. 

“There is no doubt the Commission is clear-eyed about Europe’s AI deficiencies and is showcasing serious ambition to address them, but this strategy is trying to do too many things at once,” said Max Reddel, Advanced AI Director at the Centre for Future Generations. “Europe has three clear options: help academics with compute access, build an ARPA-style innovation driver like Germany’s SPRIN-D, or create a true frontier AI lab with in-house talent. While spreading our bets  might feel like the safest approach, the past 20 years of EU innovation policy tells us that it is more likely to risk undercutting  all three.”

The European Commission has provided a sober, realistic assessment of Europe’s position in AI. Acknowledging that the EU is significantly behind the United States and China, faces chronic underinvestment, confronts a substantial compute gap, and operates within a fragmented ecosystem is a sound and realistic assessment.

The strategy also correctly links Europe’s AI capabilities to technological sovereignty and the shifting geopolitical landscape, recognizing that Europe must become an “AI continent“ to avoid losing unacceptable leverage to third countries. For instance, connecting RAISE to the upcoming AI gigafactories exemplifies the EU’s commitment to digital sovereignty by ensuring multiple levels of the tech stack are European.

Finally, CFG particularly appreciates the strategy’s dual approach, noting that advancing AI requires investment in both “AI in science” and “science for AI” – fundamental research in safety, robustness, and trustworthiness. This interplay is essential for both business adoption and risk mitigation.

Three design options for RAISE

Despite these positive features, RAISE will not be able to meaningfully advance all of the EU’s ambitious AI goals under its current design. As we see it, the EU must make a decisive choice between three options, and potentially adjust RAISE’s institutional design and funding accordingly.

• RAISE as an enabler of academic research. Academia suffers from a large and increasing compute gap with industry. As the Commission correctly identified, academic institutions need more funding and compute to pursue fundamental AI research at modern-day scale. With access to AI factory and AI gigafactory compute, RAISE could potentially achieve this goal. However, while investments in this area are not futile, the EU should not expect these to automatically yield large benefits to European sovereignty. After all, without an ecosystem of innovative European AI companies that can capitalise on academic research (Europe hosts only a 6% global share of AI players according to the strategy), benefits may remain largely theoretical or flow to third countries.

• RAISE as a driver of real-world innovation. RAISE introduces “Thematic Networks of Excellence” and promises to  support scientists in translating their most promising scientific and technological breakthroughs into real-world applications and new products and solutions. However, the current organizational structure’s multistakeholder, slow-moving, and likely consensus-based model is ill-suited to enable this goal. Rather than a passive, virtual network with a secretariat that coordinates Horizon Europe grants, an ARPA-style institution, like Germany’s SPRIN-D, would be a much more suitable vehicle. Such a model enables the active coordination, agile governance, commercialisation pathways and high-risk appetite necessary for breakthrough innovation.

• RAISE as a frontier trustworthy AI lab. While RAISE would support a European network of frontier AI labs, it is not set up to push the capability frontier of general-purpose AI itself. A full-fledged  ‘CERN for AI’ would both conduct fundamental research and apply those insights at scale, which cannot be done through a network alone. Under this model, RAISE would not only have to coordinate research across Europe, but also host  in-house talent focused on solving problems that demand rapid iteration and concentrated effort like managing large distributed training runs needed for frontier models. This option would require significantly more funding on much shorter timescales than the next MFF and may hence prove challenging within the EU’s current constraints.

“It is not too late for Europe to secure a seat at the table of global AI development,” said Daan Juijn, Senior Advanced AI Researcher at Centre for Future Generations. “ But this requires a clear decision as to what role the European Union wants to play, backed by policies that match the overarching  aim – and that has to be strongly backed by member states for any hope of rapid execution.”

A whole-of-Union AI approach will be needed to make RAISE’s operations and research uptake a success following any of the three design options. On the operational side, several member states and tens of businesses together are set to allocate 83%+ of AI Gigafactory compute, a key resource for RAISE’s operations. More ambitious funding scenarios for design options 2 and especially 3, particularly common borrowing mechanisms and resources from the next MFF referenced in the Strategy, hinge on widespread member state support. RAISE’s governance structure, though still being finalized, will likely operate under ERA rules, requiring buy-in from at least 14 member states for key decisions.

Effective research uptake and commercialization will depend equally on both governments and businesses. One strategic lever is ensuring sufficient inference compute through the upcoming Cloud and AI Development Act, which aims to triple EU data center capacity within five to seven years—contingent on strong member state support. Another critical factor is whether European businesses will experiment with the novel AI methods RAISE plans to develop. With current AI adoption among EU firms at only 13.5%, creating genuine demand for EU AI investments and research breakthroughs requires substantially broader business engagement.

CFG stands ready to engage constructively with the Commission, member states, and all stakeholders to further build on this strategy and ensure Europe secures its technological future.