Delivering the EU’s AI Continent Action Plan

The release of the European Commission’s AI Continent Action Plan marks an important moment in Europe’s ambition to become a global leader in AI. The strategy is bold, urgent, and refreshingly focused. Its five pillars—compute, data, talent, adoption, and regulatory simplification—form a coherent framework for building trustworthy AI systems aligned with European values and strategic interests.

This is a promising start. But as always, effective implementation is key.

The plan’s standout feature is the commitment to building up to five AI Gigafactories—massive compute hubs, each powered by around 100,000 advanced AI processors. Until now, Europe has lacked the computational muscle to train and deploy AI models at the scale of OpenAI, Google, or Meta. These Gigafactories could become a game-changer, directly addressing Europe’s most glaring gap in AI— access to sufficient compute for world-class AI development.

The immediate launch of a call for expressions of interest to host a Gigafactory signals welcome urgency. Yet, to get these facilities up and running within two years, a great deal of work lies ahead. To ensure progress is swift and strategic, the Commission can recommend Member States create special compute zones or streamlined permitting, as is already being tested in France and the UK. In parallel, hosting countries could reactivate nuclear reactors, extend the lifespan of existing ones, or look for alternative options – such as geothermal, hydropower, or gas or biomass facilities with high-quality carbon capture and storage systems – to quickly secure access to additional clean dispatchable energy, as the International Energy Agency highlighted in its recent report on Energy and AI.

The Action Plan also rightly prioritizes AI for security. Amidst the current geopolitical whirlwind,  Europe is facing rising threats—from cyberattacks to information warfare to other non-conventional threats. Countering such threats requires investing in sovereign AI systems that strengthen European resilience. 

Developing these systems needs to happen in a secure environment, as geopolitical adversaries may choose to sabotage or steal models through hybrid means. To safely develop sensitive AI applications, we recommend that one Gigafactory is designed to meet RAND Security Level 4 standards— a security level not yet reached by leading AI developers that would protect against well-resourced threat actors. Doing so would not only bolster European autonomy and resilience but would also position the EU as a leader in secure AI—a fast-growing, high-stakes domain.

Through the AI Factories, Gigafactories, and other upcoming investments under the Cloud and AI Development Act, Europe is laying the groundwork for its sovereign AI infrastructure. These are laudable initiatives. AI potentially offers a double dividend for citizens: better protection against modern threats like cyberattacks and disinformation as well as large productivity boosts. Research by Goldman Sachs shows that today’s AI models can already enable knowledge workers to be 25% more productive – future AI systems will only increase this number. But we can only cash these checks if we hold the pen—developing these critical technologies in Europe ensures we’ll never lose access to them when we need them most.

That said, the twenty-billion-euro question remains: how exactly will the envisioned infrastructure deliver the trustworthy European AI that President von der Leyen has called for?

The Action Plan references “training frontier models” and pursuing “moonshots” in science, healthcare, and biotech. But Europe currently lacks the institutional scaffolding to turn that ambition into action. Simply opening up the Gigafactories through call for actions likely won’t do; besides the French start-up Mistral, few European companies currently have the experience or capacity to fully leverage Gigafactory-scale compute. As Commission President von der Leyen put it, we need “the largest public-private partnership in the world for the development of trustworthy AI”— not simply to finance the datacenters, but also to optimally utilise them.

We therefore strongly support fast-tracking the creation of a dedicated talent and research institution to coordinate and guide compute usage. The proposed Resource for AI Science in Europe (RAISE) could serve this role—but only if it is established quickly and gets backed by a clear structure and a strong mandate. 

We suggest the Commission model RAISE on proven, high-impact entities like Focused Research Organizations and ARPA-style institutions, supported by Mario Draghi. As we outlined in a recent report, Building CERN for AI, such a hybrid institution could combine lean governance with bold ambition—coordinating decentralized research hubs while attracting highly talented in-house teams for frontier model training. In doing so, the EU can spur a self-sustaining ecosystem that builds AI systems which are both competitive and trustworthy.

Lastly, the Plan’s emphasis on sustainability is vital—especially as Europe aims to triple its data center capacity by 2030. But ambition must be matched by action. Without parallel investments in clean energy generation, storage, and grid resilience, the EU risks building compute infrastructure faster than it can power it, leading to potential delays or insufficient clean energy availability in other sectors. A sustainable AI future demands integrated planning across both the digital and climate sectors—from permitting to decarbonisation.

If implemented well, the AI Continent Action Plan could mark the beginning of a new chapter—one where Europe doesn’t just consume AI, but helps produce it, shape its future and, most importantly of all, reaps the benefits from it.