CERN for AI

President von der Leyen proposed CERN for AI as a pan-European institution focused on developing trustworthy AI. To deliver on this ambitious and timely mission, the institute would gather world-leading scientists and give them access to four 100.000 top-of-shelf GPU data centres, the Gigafactories. This will be the “largest public-private partnership in the world.”

CFG’s research strongly endorses and builds on these plans in the areas of international partnerships, research and governance structures, and legal frameworks.

Trustworthy AI refers to AI systems that are safe, reliable, and transparent in their operation. Current AI systems often act as “black boxes”—their decision-making processes are too complex for humans to understand. For instance, GPT-4 is rumoured to have 1.8 trillion parameters—mathematical dials that are tuned automatically and determine the model’s behaviour. We do not yet have techniques to disentangle how all these dials actually contribute to answers that make sense. This opacity creates risks when these systems are deployed in sensitive areas like healthcare or critical infrastructure. For instance, we cannot yet reliably flag when a model is lying or hallucinating. CERN for AI would pioneer new approaches to make AI systems more interpretable, reliable, and aligned with European values. This isn’t just about safety—trustworthy AI is also crucial for widespread adoption, especially in sectors where reliability is paramount.

CERN for AI would deliver three key benefits:

• Economic: It would help Europe compete in the rapidly growing AI market and boost productivity across sectors. McKinsey estimates that generative AI—only a subset of AI that typically excludes AI agents such as OpenAI’s recently launched Operator—will become a milti-trillion dollar market by 2030. Moreover, trustworthy AI could turbocharge lagging European AI adoption rates, bringing businesses from the current 13.5% closer to the 75% target by 2030.
• Security: It would provide Europe with domestic AI capabilities crucial for cybersecurity, counter-terrorism, and tackling disinformation. Depending on foreign powers for these critical capabilities is an unacceptable risk.
• Innovation: It would pioneer new approaches to making AI systems more trustworthy and reliable, setting global standards. Beyond these direct benefits, CERN for AI could catalyze a broader European tech ecosystem, similar to how the original CERN led to innovations like the World Wide Web. Homegrown AI could eventually support crucial innovation in areas like climate transition.

​​Yes, but time is critical. While American and Chinese companies currently lead in AI, the field is rapidly evolving, creating opportunities for new players. The recent Stargate Project has sparked concerns—will President von der Leyen’s proposed 200 billion-euro AI investment be enough and arrive in time? This pessimism is misplaced. Stargate’s funding is not yet secured, and AI’s new paradigm proves that innovation isn’t solely driven by vast data and computing power.

Europe is on the right path. Rather than replicating today’s AI models like ChatGPT, the EU is focusing on the AI of tomorrow: trustworthy, ethical systems—an ambition unmatched by other global powers. This opens a strategic niche in the AI value chain.

European values, especially the rule of law, are key to building lawful, ethical AI—essential in a field where values matter. Europe also boasts significant strengths: collective wealth, a large population, proprietary business data, and top-tier talent.

Under President von der Leyen’s leadership, Europe has begun its push for global AI relevance. However, this window of opportunity won’t remain open for long, making fast and well-executed implementation of CERN for AI crucial.

The majority of the funding (around €24.5 billion) would go towards computational infrastructure, including AI chips and the necessary interconnect systems. The remainder covers:

• Energy costs for running the compute infrastructure
• Personnel costs for approximately 3,000 staff members
• Physical infrastructure like buildings and cooling systems
• Operational costs including maintenance and governance.

This investment would give Europe a seat at the table.