CEIAS
Who will pay for Europe’s AI sovereignty?
Oct 2, 2026 in CEIAS Insights

Who will pay for Europe’s AI sovereignty?

Europe wants greater technological sovereignty, but building sovereign AI infrastructure requires enormous amounts of capital. The EU’s own plans assume that much of this will come from private investment. As foreign investors help finance European AI infrastructure, the harder question is not simply where the money comes from, but what control comes with it.

Key takeaways:

  1. European AI sovereignty is about reducing strategic dependencies while expanding European capacity, rather than eliminating international technological interdependence.
  2. Europe’s AI ambitions will compete with other industries for scarce land and grid capacity. Eemshaven, a major energy and industrial hub in the northern Netherlands, shows why policymakers may need to consider the strategic value of AI infrastructure, not just how much electricity it consumes.
  3. Foreign capital can strengthen European AI sovereignty if it expands European computing capacity without transferring control over critical infrastructure, technology, data or access.

Semiconductors show how technological dependence can create geopolitical vulnerability, and why sovereignty is better understood as managing critical dependencies than eliminating them. But the same question increasingly applies to AI computing infrastructure, particularly the data centers and AI Gigafactories that provide the computing power AI systems depend on. As Europe turns to private and foreign capital to help finance that infrastructure, the key distinction is between ownership and control: who invests in or owns the infrastructure is not necessarily the same as who controls its technology, data, operations or access to computing capacity.

The Netherlands is home to ASML, whose lithography equipment is indispensable to manufacturing the world’s most advanced semiconductors. Yet European businesses and governments remain heavily dependent on cloud infrastructure provided by Amazon Web Services, Microsoft and Google. That puts Europe in an odd position. It possesses critical technology at one end of the AI supply chain while depending heavily on non-European companies further down it.

In an interview for this article, Eric-Mark Huitema, managing director of Dutch high-tech industry association High Tech NL, described this as a paradox: Europe supplies technology essential to producing advanced chips, but European companies then largely buy the resulting computing power back from non-European firms. He argues that Europe should pursue “strategic autonomy through mutual dependence”: remaining strong in critical areas such as semiconductor equipment, computing infrastructure and industrial technology while retaining alternative suppliers and sources of computing capacity.

The EU’s AI policy has also become more focused on computing infrastructure. In June 2026, the European Commission presented a new European Technological Sovereignty Package covering the technology chain from semiconductors to infrastructure, software, cloud and AI. This builds on the AI Continent Action Plan, which identifies computing infrastructure as one of five strategic areas for European AI development. The EU now has 19 AI Factories deployed around its EuroHPC supercomputers and in July launched a call for up to seven much larger AI Gigafactories.

The data-center dilemma

AI systems depend on physical computing infrastructure. Models require processors, data centers, cooling, electricity and high-capacity networks. That makes access to computing power a strategic issue in its own right. The Netherlands shows how difficult this becomes in practice: grid congestion has made new electricity connections scarce, forcing policymakers and grid operators to choose how to allocate limited grid capacity among competing needs. Large data centers therefore raise a broader allocation question: when electricity and land are scarce, how much strategic value does a computing facility need to justify using them?

The criticism becomes particularly acute when scarce Dutch electricity and land are used for infrastructure owned by some of the world’s largest foreign technology companies, such as Microsoft and Google, which operate major data centers in the Netherlands. But there is an obvious problem. If Europe limits the infrastructure required for large-scale computing while seeking greater independence from American cloud and AI companies, where exactly will European sovereign AI run?

These competing demands come together in Eemshaven, a major energy and industrial hub in the northern Netherlands. It hosts two hyperscale data centers, while the adjacent Oostpolder includes 200 hectares designated for high-tech manufacturing, battery production and data centers. Heleen van Wijk of Groningen Seaports, which manages and develops Eemshaven, argues that its energy infrastructure, available space and international connectivity position it well for further digital infrastructure.

Not every data center performs the same function. A colocation facility provides space, power, cooling and connectivity for customers’ computing equipment. A hyperscale facility may support global cloud services such as Microsoft, Amazon or Google. An AI Gigafactory, by contrast, combines large numbers of advanced processors with software, high-speed networking and specialized infrastructure for AI models. The EU’s planned Gigafactories are also intended to provide European companies, researchers and public authorities with access to their advanced processors and computing capacity.

Huitema argues that policymakers should recognize this distinction when allocating scarce resources. He proposes treating “Strategic AI Infrastructure” as a separate category: infrastructure that provides European companies and research institutions with access to advanced compute, contributes to technological autonomy and uses energy efficiently and flexibly.

Groningen Seaports is already considering some of these factors. Van Wijk said it does not apply a single selection criterion when allocating scarce land and grid capacity, but looks at a project’s overall value. She pointed to factors ranging from cooperation with universities and innovation to power purchase agreement structures, nature restoration and community funds. She added that “a demonstrable contribution to Dutch or European technological autonomy,” for example by making AI computing capacity strategically available, “could be a clear plus.”

In terms of technological sovereignty, this could mean assessing a project’s economic value, its contribution to European research and innovation, whether it expands European access to advanced computing capacity, and whether it can support the electricity system through flexible demand, renewable generation or battery storage. Van Wijk also pointed to the potential role of data centers as energy hubs, for example in combination with battery storage or hydrogen. That matters in the Netherlands, where new AI facilities would be competing with other industries for already scarce grid capacity. Policymakers need to decide which forms of compute justify consuming those resources. In Eemshaven, that policy debate comes down to something tangible: who gets the scarce megawatts and hectares.

Who pays for sovereign compute?

The EU’s own plans already depend heavily on private capital. InvestAI, the EU initiative to mobilize €200 billion for AI investment, includes a facility to mobilize €20 billion for AI Gigafactories. The latest Gigafactory initiative is private-led, with up to €10 billion in EU and national public funding intended to de-risk projects and unlock at least €20 billion in private investment. With Microsoft alone planning approximately $175 billion in capital expenditure in 2026, the scale of global AI investment makes substantial private capital difficult for Europe to avoid.

The need for private capital also brings foreign investment into the sovereignty debate. European investors will provide some of the financing, but so will large infrastructure investors, sovereign wealth funds and technology companies outside Europe. This raises the question of what conditions Europe should attach to foreign investment in AI infrastructure.

Groningen Seaports takes a similar view. Van Wijk said foreign capital is not in itself a concern: international companies are an important part of Eemshaven’s position in global value chains. With AI infrastructure, however, she said what comes with that capital matters more: “who ultimately controls the infrastructure, where does the knowledge end up, and who has access to the available computing power?” For Groningen Seaports, then, the investor’s nationality matters less than whether the investment gives a foreign actor control over the infrastructure, technology, data or access to computing capacity.

Foreign ownership versus foreign control

Foreign investment can strengthen European technological sovereignty if it gives Europe more computing capacity without creating new forms of foreign control. Its strategic value depends on the capabilities Europe gains and the dependencies the investment creates.

In 2025, France and the United Arab Emirates (UAE) agreed on a framework for up to €50 billion in joint AI investment, centered on developing a major AI campus in France. Under the 2025 plan, the campus was expected to reach 1.4 GW of capacity by 2030, with a first operational phase planned for 2028 and an investment of €8.5 billion.

Ownership alone says little about the control that comes with an investment. A foreign investor that supplies capital but has limited governance rights creates a different dependency from a technology company that simultaneously supplies the processors, operates the cloud platform and controls access to the facility. The France-UAE case shows why shareholder nationality tells only part of the story.

Keppel’s investment in the Netherlands shows a different form of foreign ownership. Singapore-based Keppel has invested in Dutch data-center infrastructure, including facilities in Amsterdam, Almere and Eindhoven. In 2021, Keppel DC REIT acquired the data-center campus in Eindhoven for €37.2 million. It was fully leased, with a global IT-services provider occupying its data-center space. Owning such an asset does not, on its own, give the investor control over the processors its customers use, their cloud platforms, the data they process or the workloads they run.

The France-UAE project, however, involves more than financial ownership. A large AI campus combines capital with scarce electricity, advanced processors, networking and access to computing capacity. Foreign investment in such a facility can expand European computing capacity, but control over these resources can also create new strategic dependencies.

What should Europe protect?

European policy should focus on the control rights and dependencies that come with foreign investment. The EU is already moving beyond a simple ownership test. The Commission’s 2026 Technological Sovereignty Package proposes a single EU-wide framework for assessing cloud and AI sovereignty. Its existing framework for procuring sovereign cloud services uses 48 criteria, covering issues ranging from jurisdiction and supply chains to operational control, technology, data and AI.

For strategic AI infrastructure, Europe could distinguish financial ownership from operational control, limiting investors’ authority over how infrastructure is operated and who can use it. Access matters as well. Infrastructure receiving public support or consuming scarce grid capacity could be required to provide meaningful access to European companies, researchers and public institutions.

Europe should also consider what would happen if a critical processor supplier, cloud provider or operator disappeared. If its loss would cause serious disruption and no realistic substitute exists, Europe remains dependent regardless of who owns the infrastructure. A foreign shareholder or technology provider should also not be able to restrict European access unilaterally during a political or economic dispute. This would not make the infrastructure entirely European, nor is that the objective. The goal should be to accept foreign investment without allowing a foreign actor to control something Europe cannot easily replace.

In Eemshaven, abstract questions about AI sovereignty become concrete choices about scarce land, electricity and computing capacity. The France-UAE partnership and Keppel’s investment in Dutch data centers show why foreign ownership is an inadequate measure of sovereignty. The strategic risk rises when financial ownership is combined with control over technology, operations, data or access to that capacity. Foreign investment and technological sovereignty are not necessarily in conflict. The real test is whether an investment leaves Europe with more control and more alternatives, or fewer.

Key Topics

Geoeconomics • Energy • Technology

office@ceias.eu

Dunajská 37
81108 Bratislava
Slovakia

Sign up for our newsletter to receive the latest news and updates from CEIAS.

All rights reserved.

© CEIAS 2013-2026