EU opens bidding for AI gigafactories in push to expand compute

Europe makes a large compute bet, but on a different scale than the US

The European Commission has opened bidding for what it calls up to seven AI gigafactories, launching one of the bloc’s most explicit attempts yet to expand domestic computing capacity for advanced artificial intelligence. The plan combines public and private money, with up to 10 billion euros expected from EU and national sources and at least 20 billion euros in private investment.

The headline number is large by normal public-technology standards. It is also modest by the standards now set by the global AI infrastructure race. According to the candidate report, Europe’s roughly 30 billion euro package stands against more than $600 billion that major US technology companies alone plan to spend on data centers this year. That contrast explains both the ambition and the vulnerability of the initiative: Europe is trying to build strategic compute capacity, but it is doing so while its biggest rivals are investing at a far greater pace and scale.

What the AI gigafactories are meant to do

The proposed facilities are designed to give startups, companies, research institutions, and government agencies access to the infrastructure needed to train and run large AI models. In practical terms, the project is about compute availability. For years, Europe has faced the same basic bottleneck confronting much of the rest of the world: cutting-edge AI development increasingly depends on access to large clusters of advanced chips, power, networking, and facilities that can support them.

The Commission’s approach is to build shared capacity rather than leave access entirely to foreign hyperscalers or a fragmented patchwork of national programs. Eighteen member states, including Germany and France, are participating. Applications are due by November 12, 2026, and construction of the first facilities is set to begin in 2027.

The project sits within the EU’s broader “AI Continent” strategy, which frames AI infrastructure as an economic and geopolitical necessity rather than a discretionary industrial subsidy. That framing reflects a hardening consensus across advanced economies: compute is no longer just a technical input. It is becoming a strategic asset tied to industrial competitiveness, research capability, and digital sovereignty.

Why the hardware partnerships matter

The Commission has also signed letters of intent with AMD, Nvidia, and Qualcomm to secure access to hardware. That detail is important because data-center plans are only credible if hardware supply follows. In the current market, advanced AI chips remain one of the most constrained inputs in the stack, and capacity announcements mean little without realistic pathways to procurement.

The inclusion of multiple chip suppliers also signals a practical posture. Rather than anchoring the effort to a single hardware vendor, the Commission appears to be keeping the program broad enough to support different technical architectures and procurement channels. That does not eliminate supply risk, but it reduces the chance that the initiative becomes dependent on one company’s roadmap or allocation choices.

For European startups and research institutions, that could be one of the most meaningful aspects of the plan. A shortage of available compute does not just slow frontier-model development. It can distort the entire innovation pipeline by forcing smaller players to wait longer, pay more, or relocate critical workloads to providers outside the region.

The scale problem remains

Even so, the numbers in the report make the competitive gap hard to ignore. If major US technology companies are planning to spend more than $600 billion on data centers in a single year, Europe’s total package for up to seven AI gigafactories is small by comparison. The candidate source describes the EU effort as roughly 20 times smaller.

That disparity does not automatically make the European plan ineffective. Public infrastructure programs do not need to match private-sector hyperscaler spending line for line to create value. If the facilities are well targeted, reliably funded, and efficiently allocated, they could still improve access for exactly the groups Europe says it wants to support: startups, research organizations, established companies, and state institutions.

But the scale gap does shape expectations. The EU is not currently attempting to outspend the largest US firms. It is trying to ensure that Europe does not become structurally dependent on compute capacity built elsewhere. That is a narrower goal, and probably a more realistic one. The challenge is that AI capability is increasingly cumulative. Regions with larger, faster, and more abundant compute tend to attract more talent, more training runs, more experimentation, and more downstream companies.

That means Europe’s initiative may be judged less on the raw number of gigafactories than on whether those facilities meaningfully reduce waiting times, widen access, and create a credible platform for homegrown AI development. If they do not, the program risks being seen as symbolically important but operationally insufficient.

A policy shift from discussion to procurement

One of the more significant aspects of the announcement is that it moves the conversation from abstract strategy to procurement and construction. Europe has spent years debating AI regulation, competitiveness, and sovereignty. Opening the bidding process turns those debates into a concrete infrastructure program with deadlines, participants, and named industrial partners.

That transition matters because AI policy is increasingly measured by physical follow-through. Access to compute depends on land, power, cooling, chips, and financing, not just white papers and legislative ambitions. By setting a November 2026 application deadline and a 2027 construction timeline for the first sites, the Commission is putting a schedule behind a strategic objective that until recently could sound mostly rhetorical.

There is still a long distance between announcing a framework and delivering functioning facilities. Private capital must materialize at the projected level. Hardware availability has to hold. National and EU coordination will have to remain intact across a multiyear buildout. But as a policy signal, the launch is clear: Europe no longer wants to treat frontier AI compute as someone else’s infrastructure problem.

What comes next

The first test will be whether the bidding process attracts serious, executable proposals that can turn pooled funding into actual capacity. The second will be whether those future facilities are accessible enough to support the research and industrial ecosystem the Commission says it wants to strengthen.

For now, Europe has defined the problem correctly: advanced AI requires large-scale computing infrastructure, and relying entirely on outside providers carries economic and strategic risk. The unresolved question is whether a 30 billion euro package can create enough capacity, fast enough, to matter in a market moving at hyperscaler speed.

• The European Commission has opened applications for up to seven AI gigafactories.
• Up to 10 billion euros in EU and national funding is expected to attract at least 20 billion euros in private investment.
• Applications are due November 12, 2026, with first construction planned for 2027.

This article is based on reporting by The Decoder. Read the original article.