Europe is racing to build the infrastructure its AI ambitions depend on — without becoming permanently dependent on infrastructure it doesn't control. That tension sits at the heart of one of the continent's most consequential technology initiatives: the European Commission's plan to build up to five AI Gigafactories, each capable of supporting at least 100,000 GPUs and backed by hundreds of billions of euros in public and private investment.

An AI supercycle collides with a sovereignty problem

Global cloud revenues are climbing, AI investment is measured in trillions, and agentic AI is widely seen as the next major inflection point — the kind of shift some compare to the arrival of the smartphone. At the same time, a countertrend is gaining momentum: cloud repatriation. Organizations worried about data privacy, cost, and control are increasingly choosing hybrid setups that combine public cloud economics with the ability to keep sensitive workloads under their own jurisdiction.

Layered on top of both trends is a hard physical constraint: power. The United States is already commissioning single data centers with roughly 1 gigawatt of capacity. Europe is several years behind on that curve, which is pushing the continent toward a different architecture altogether — a federated network of smaller, distributed data centers rather than a handful of mega-sites, designed to work around local power limits instead of waiting for them to disappear.

What an AI Gigafactory actually is

Announced under the EU's AI Continent Action Plan, the AI Gigafactories initiative isn't simply about building bigger data centers. It rests on four strategic goals:

1. Protection — securing sovereign European AI infrastructure so the continent doesn't repeat what happened during the rise of public cloud, when ownership concentrated almost entirely outside Europe.
2. Federation — integrating gigafactories into the existing European AI and cloud-edge continuum rather than building isolated, standalone facilities.
3. European models — supporting the development of local, industry-specific, or language-specific AI models, not just chasing frontier-model parity with the US and China.
4. Open access — making compute capacity available to citizens, SMEs, startups, and public institutions, not only large incumbents.

Sovereignty gets a formal framework

In October 2025, the European Commission published a sovereignty framework defining eight dimensions for evaluating cloud and AI providers. Four stand out as particularly consequential for buyers:

• Strategic sovereignty — who ultimately owns and controls the provider, and under which jurisdiction.
• Supply chain sovereignty — how much of the underlying compute, storage, and networking hardware actually originates in Europe.
• Operational sovereignty — independence from foreign control over day-to-day operations and support.
• Technological sovereignty — openness across the technology stack, avoiding vendor lock-in.

These dimensions give European enterprises and public bodies a concrete way to evaluate claims of "sovereign AI" rather than taking them at face value — a distinction that matters as more providers adopt the label.

Where the money and priority sectors are headed

The EU has identified three sectors as early off-takers of gigafactory capacity: healthcare, mobility and automotive, and defense, security and space. The plan includes pre-purchasing guaranteed GPU hours for these sectors under a "buy European" principle, designed to give European AI infrastructure early, committed demand rather than leaving it to compete purely on price against hyperscalers.

Momentum is already visible at the national level. Spain and Portugal have signed a memorandum of understanding to jointly pursue a gigafactory site in the Iberian Peninsula, and similar coalitions are forming elsewhere across the continent. The initiative's first phase — an expression-of-interest window — closed in June 2025; the next milestone, a formal call for proposals, is expected to open soon.

Why the networking layer matters more than it seems

Much of the sovereignty conversation focuses on chips and compute, but the network connecting GPUs, storage, and data centers to each other is just as central to the sovereignty equation — and it's an area where very few European vendors compete at scale. Open, programmable data center fabrics, support for open operating systems, and quantum-safe networking are increasingly treated as sovereignty requirements in their own right, not just performance features.

The bigger picture

Europe's AI Gigafactories initiative is as much an industrial policy bet as a technology project. Its success will depend on execution as much as ambition: whether the call for proposals attracts credible consortia, whether power and permitting keep pace with GPU demand, and whether "sovereign" infrastructure can stay commercially competitive with hyperscale alternatives.

Nokia and Pedab recently hosted a webinar exploring this topic in more depth, featuring Nokia's David Gong on data center networking and AI infrastructure. Watch the full recording below for a closer look at the sovereignty framework, the gigafactory roadmap, and audience Q&A.

Want to learn more about Nokia's AI and data center networking solutions? Reach out to local Pedab representative.