The EU AI Continent Action Plan

Critical Questions by Laura Caroli

Published April 22, 2025

On April 9, the European Commission unveiled its "AI Continent Action Plan." The plan outlines the European Union's strategy to boost AI development and adoption across the continent in an effort to turbocharge the sector and enhance Europe's strategic autonomy. This vision is in line with the message that European Commission President Ursula von der Leyen conveyed at the February Paris AI Action Summit: world AI leadership is "up for grabs" and the European Commission will do anything in its power to place Europe as a serious contender for it. At the summit, von der Leyen also announced the InvestAI facility, which is intended to mobilize up to 200 billion euros of investments in AI, to signal the European Union's serious commitment to boost its AI sector.

The AI Continent Action Plan is structured in five pillars: computing infrastructure, data, skills, regulatory simplification, and adoption in key strategic sectors. While the plan has been received with a lot of interest and includes some notable actions, it begs the question of whether and to what extent it introduces new measures or merely repackages existing ones.

Q1: Is the AI Continent Action Plan the first of its kind?

A1: To understand its significance and value, the AI Continent Action Plan must be put in context with previous EU-wide initiatives around AI. Indeed, this is not the first major document outlining a European strategy on AI. The first comprehensive AI strategy dates to April 2018 (three years before the EU AI Act was even proposed). It tackled data, skills, adoption, and investment, alongside the possible need for regulation. It also launched the Coordinated Plan on AI, which was updated in 2021 and still constitutes the skeleton of the EU AI policy action, in coordination and synergy with member states. Already in the 2021 revision of the Coordinated Plan on AI, the ambition to establish the European Union as a global leader on AI was clearly spelled out, although achieving it in practice was far from reach at the time, and it still is now. The white paper on AI from February 2020 set the scene for what the EU vision for AI would look like: an ecosystem of excellence (supporting and leveraging Europe's strength in AI, such as startups, research, and talent), alongside an ecosystem of trust (the future EU AI Act). A subsequent key policy document informing the AI Continent Action Plan is the digital compass, presented by the European Commission in March 2021. The document laid down the European Union's digital vision for the decade 2021-2030 through concrete targets (e.g., on skills, adoption, semiconductor production, and so on).

The AI Innovation Package, launched in January 2024, seemed at the time a culmination of the European Union's policy actions on AI. Its main focus was to support AI startups and small and medium enterprises (SMEs) by facilitating access to funding, compute infrastructure, and data, with specific attention to leveraging generative AI. Importantly, the Innovation Package was the document that launched the "AI factories" (selected AI ecosystems), which are now at the heart of the AI Continent Action Plan.

This shows how the action plan is only the tip of the iceberg in a long-standing policy initiative to stimulate AI development and uptake in Europe.

Q2: Where is the money?

A2: Undoubtedly, the European Union has significantly stepped up its investments in AI. Until the Paris AI Action Summit, the European Commission had been investing around €1 billion per year on AI through public EU funds, to mobilize national and private investments up to €20 billion annually for the 2020-2030 decade. Certainly, the InvestAI facility announced by von der Leyen in Paris can be considered a marked shift compared to past years. The facility will see the direct involvement of the European Investment Bank, which is stepping up its action on AI investments. Of the €200 billion InvestAI will mobilize, €20 billion should be entirely dedicated to the announced AI gigafactories (see below). According to the commission's website on the initiative, "the gigafactories funded through InvestAI will be the largest public-private partnership in the world for the development of trustworthy AI." Despite the grandiose tone in the speeches and press releases, it is important to be aware of the breakdown of the announced "billions." Indeed, the €200 billion will largely come from the AI Champions Initiative, a private undertaking composed of over 60 European companies that pledged €150 billion to "frame a positive vision for Europe." The commission's contribution would top these up by another €50 billion, hence the total of €200 billion. Also, according to the commission, the €10 billion already dedicated to the AI factories is "already the largest public investment in AI in the world, and will unlock over ten times more private investment."

However, a more accurate breakdown of these figures paints a different picture, with only €7 billion already committed by the European Commission and the member states, a sum that can be redirected to the already existing EuroHPC Joint Undertaking. The Joint Undertaking, however, is only partly dedicated to AI Factories, while it is more broadly focusing on research and innovation based on supercomputers for a variety of use cases. This shows how announcements must always be taken with a grain of salt.

Q3: What is new about AI infrastructure?

A3: The heart of the plan is about the AI factories and gigafactories. As mentioned, AI factories were launched in January 2024 and are intended to be open AI ecosystems providing access to compute, data, and talent, in particular to AI startups and SMEs. The AI factories are selected and approved under the EuroHPC Joint Undertaking, the facility meant to pool resources and coordinate national actions on Europe's federated supercomputers. As of December 2024, seven multicountry consortia were selected to host the first group of factories, and six additional ones were announced in March 2025. According to the AI Continent Action Plan, the AI factories will be fully operational by the end of 2025. Also, the plan announces the procurement of nine additional AI-enabled supercomputers, and the upgrading of an existing one "with AI capabilities," to "more than triple the current EuroHPC AI computing capacity." Apart from launching these additional AI factories and supercomputers, the plan proudly boasts the launch of the gigafactories, which will be "large-scale facilities that develop and train complex AI models at an unprecedented scale" and with "massive computing power, exceeding 100,000 advanced AI processors." Ambition is certainly high. However, the initiative will take months to be launched (the official call is expected in Q4 2025) and even longer to be accomplished. The other key infrastructural element forming part of the plan is about enhancing EU cloud and datacenter capacity, with an emphasis on reducing dependence on non-EU infrastructure. To this end, the plan proposes a Cloud and AI Development Act, which will be aimed at "at least tripling the EU's data center capacity" within the next five to seven years. Interestingly, this will be coupled with action on reducing energy and water consumption of data centers, a sign that the European Union remains sensitive to the carbon footprint of massive AI infrastructure buildup.

Q4: What is new about semiconductors?

A4: The emphasis of the action plan on reducing dependencies on third countries for semiconductor production certainly stands out, especially in light of recent geopolitical developments. However, it must be noted that strategic autonomy and value chain independence were the very objective of the EU Chips Act, which was proposed in February 2022 and entered into force in September 2023 (and will soon be subject to a revision). The dedicated website itself introduces the act, saying "the European Chips Act will boost Europe's technological sovereignty."

When the Chips Act was proposed, the world was still in a Covid-19-riddled economy, where dependencies were becoming a real liability for Europe. At that time, the problem was the world shortage of chips and the substantial dependency on Taiwan. The Chips Act is therefore aimed at doubling the European Union's world market share in chips from 10 percent to 20 percent. The Chips Act is currently expected to be supported by €43 billion in investments until 2030, which the private sector is expected to roughly match.

Interestingly, the Chips Act website lists among its proposals "building semiconductor international partnerships with like-minded countries." In light of the current transatlantic tensions, new questions may arise as to what these countries are. In the AI Continent Action Plan's language, the goal "to reduce dependencies on critical technologies and strengthen sovereignty in cutting edge semiconductors" will inform the upcoming revision of the Chips Act.

This could indeed appear to shift the emphasis more towards reducing dependencies from an increasingly unreliable supplying partner, such as the United States, than from Taiwan or China. The publication of the AI Diffusion Rule by the Biden administration in January 2025 resulted in roughly half of Europe having limitations in its handling of U.S.-acquired chips. The European Commission understandably responded to this initiative with concern at the time. It is worth noting that the AI Continent Action Plan specifically refers to this European reaction in talking about "the latest geopolitical developments," which call for higher independence from other countries in semiconductor production. In any case, this quest for independence will likely take years to become a substantial possibility, even if Europe is more serious about it this time than with prior such calls.

Q5: What else is notable, and what is missing?

A5: As mentioned, the focus on skills was already present in numerous previous EU strategy documents around AI. However, what stands out in the AI Continent Action Plan is the mention of actions to encourage EU talent living abroad to return to the continent, alongside the relatively less new section on attracting top talent living abroad. In any case, reading both of these elements in light of current events, they seem more directed at attracting the many researchers who are reportedly starting to consider leaving the United States due to a lack of funding or interference in their research.

An element that seems to be missing is a greater focus on open-source AI, which is only briefly mentioned in the AI Continent Action Plan. This was especially unexpected, considering Ursula von der Leyen mentioned it in her speech at the Paris AI Action Summit as Europe's distinctive mark ("this European brand of open innovation"). Also, the focus on massive buildup of compute infrastructure through the AI gigafactories, and the willingness to advance frontier AI development, "including towards artificial general intelligence" seems at odds with what is emerging as a more suited way for European companies to stay competitive on AI: the emergence of "smaller" or "mini-models", trained at a fraction of the cost and with far less computing power than the big models the likes of GPT or Gemini require. When DeepSeek boomed in January 2025, European startups were not only listening, they were already implementing similar training techniques to achieve equivalent levels of efficiency in terms of costs and computing power.

Q6: What is the degree of proposed "regulatory simplification"?

A6: After months of rumors around the idea of a possible simplification of the EU AI Act (which is not even fully in application yet), many observers were expecting concrete proposals in this regard in the AI Continent Action Plan. Despite appearing in early leaked versions of the plan, explicit language on simplification seems to have been toned down in the officially released document. As reported by Politico, indeed, the "opportunity to minimize the potential compliance burden" in a leaked version became a mere "need to facilitate compliance with the AI Act" in the final text. The European Commissioner Henna Virkkunen has already announced that the EU AI Act will be part of a simplification effort to reduce reporting obligations, in particular for SMEs.

However, the AI Continent Action Plan only talks about a dedicated service desk, which will be set up within the EU AI Office to support companies and provide tailored advice on regulatory compliance and implementation. At the same time, the action plan also opens a public consultation on how to simplify the current rules in the EU AI Act, signaling the intention is there, but the commission is still assessing the extent of the simplification. In any case, the toning down of the language on regulatory simplification in the action plan could also be interpreted in light of the current transatlantic tensions. Indeed, the Trump administration has made it clear it views EU digital legislation as part of the "extortion" by foreign countries against U.S. companies and, therefore, increasingly part of the conversation and negotiation around tariffs. In this case, therefore, postponing clear proposals of regulatory simplification could be an attempt by the European Union to keep all the possible cards in its hands.

Q7: Is the plan really new, or is it a repackaging of previous initiatives?

A7: The plan formalizes ideas that were already included in the political guidelines presented by Ursula von der Leyen for her candidacy to her second term as European Commission president in July 2024. The AI factories; the Apply AI Strategy; the idea of a CERN, but for AI; and the Data Union Strategy all appear in the new strategy. Many of these ideas are also already part of the competitiveness compass announced in January 2025 and stem from the Draghi Report on enhancing EU competitiveness. Compared to von der Leyen's political guidelines, the AI Continent Action Plan certainly spells out, streamlines, and further boosts the announced actions.

However, the AI factories were already announced in January 2024, the supercomputers have been in the making for years, while the Chips Act has been in force since 2023. As explained above, the actual fresh money is also lower than it appears from public announcements. Indeed, when looking at EU funds on a certain policy, expectations must always be measured with the constraints under which the European Commission has to move. In particular, the announced amounts are always part of the Multiannual Financial Framework 2021-2027, forming the EU budget for the current seven-year financing period. EU funds can typically be coupled with national public investments and private capital.

However, as the European Commission noted in a recent document on the current challenges around the EU budget, "the capacity of the EU budget to mobilize private investment through risk-sharing instruments is hampered by limited appetite for risk", somewhat limiting the expected leverage effect of EU public funds for large-scale projects. In any case, in order to have a real picture of how serious the European Commission is about "turbocharging AI" in Europe, it will be key to watch the first draft proposals for the Multiannual Financial Framework 2028-2034, expected in the fall of 2025. The draft figures will show how much AI development is prioritized as opposed to other competing areas such as defense, migration, agriculture, social cohesion, and so on. In any case, the picture is not entirely bleak: the AI Index Report, annually published by Stanford, shows how the public investment gap between the European Union and the United States is progressively decreasing, at least looking at figures until 2023.

Laura Caroli is the senior fellow of the Wadhwani AI Center at the Center for Strategic and International Studies in Washington, D.C.

Programs & Projects