EU Launches €30 Billion AI Gigafactories Tender in Major Infrastructure Push

The European Union has officially opened bidding for up to seven "AI Gigafactories" in a €30 billion public-private drive to build sovereign computing infrastructure and reduce dependence on foreign cloud providers. Managed by the EuroHPC Joint Undertaking, the tender offers up to €10 billion in public support and runs until 12 November 2026, marking Brussels’ transition from AI regulation to aggressive industrial strategy.
Yet the significance of this initiative lies less in the sheer size of the investment than in what Brussels is attempting to change. For years, Europe has tried to shape artificial intelligence primarily through regulation. The Gigafactories represent something different: an industrial strategy built around physical computing infrastructure, designed to give European organisations the ability to develop advanced AI without depending on foreign cloud providers.
Whether that also produces genuine technological sovereignty remains a far more complicated question.
Beyond Regulation
Europe has never lacked AI research. Its universities, research institutes and emerging companies have repeatedly demonstrated an ability to produce competitive models and innovative architectures.
The difficulty has come later.
Training, refining and deploying frontier AI systems requires enormous computing capacity, and that capacity has largely been rented from American cloud providers. European developers have frequently found themselves building their most advanced systems on infrastructure operated by Amazon Web Services, Microsoft Azure or Google Cloud.
That arrangement carries consequences extending well beyond commercial costs.
Reliance on foreign cloud platforms means capital leaves Europe, sensitive workloads operate within ecosystems governed by external legal jurisdictions, and critical digital infrastructure becomes concentrated in a small number of providers. Those dependencies create vulnerabilities for governments, industry and research alike.
The Gigafactories are intended to establish an alternative foundation.
Rather than replacing Europe's existing network of 19 AI Factories, they expand it dramatically. Each new facility must deliver between three and four times the AI processing capacity of today's installations while integrating advanced software environments, cloud platforms, high-speed networking and energy-efficient data-centre designs operating according to circular-economy principles.
Access is expected to extend well beyond large corporations. Start-ups, scale-ups, SMEs, industrial companies, universities, research organisations, defence bodies and public authorities are all intended users, provided operations comply with European requirements on data protection, cybersecurity, safety and ethics.
That breadth is deliberate. Brussels is not merely financing larger supercomputers. It is attempting to create a European computing layer that serves both commercial innovation and strategic resilience.
Infrastructure Becomes Industrial Policy
The programme also reflects a broader change in European thinking about technological competitiveness.
For much of the past decade, debates around digital sovereignty focused on legislation, competition policy and regulatory frameworks. AI infrastructure now occupies a similar strategic position to transport corridors or energy networks.
Computing capacity has become an industrial asset.
Where these facilities are eventually built will depend on more than available land. Regions capable of offering abundant low-carbon electricity, reliable grid connections, efficient cooling systems and suitable data-centre infrastructure will hold a significant competitive advantage during the selection process.
That inevitably creates new divisions within the Union.
Member states with robust energy infrastructure and favourable operating conditions stand to attract high-value investment, highly skilled employment and long-term technological ecosystems. Countries facing expensive electricity, constrained grids or restrictive environmental limitations may find themselves excluded from hosting the most strategically valuable digital infrastructure of the coming decade.
The competition is therefore no longer solely about software innovation. It increasingly revolves around access to power, cooling capacity and physical computing density.
Breaking the Hyper-Scaler Trap
If the programme succeeds, one immediate effect could be a reduction in Europe's operational dependence on American cloud hyperscalers.
Instead of continuously renting computing resources from external providers, European organisations would gain access to domestic infrastructure specifically designed for frontier AI development.
That matters especially for smaller companies and research institutions.
High-performance AI computing has often remained prohibitively expensive for start-ups and academic laboratories. State-supported facilities offering shared access could lower those barriers while keeping sensitive workloads inside European infrastructure governed by European rules.
The programme's financing structure reflects the scale of the ambition. Public authorities are expected to contribute up to €10 billion, while private investors are anticipated to provide at least twice that amount. Such leverage indicates that Brussels is not attempting to replace private capital but to direct it towards infrastructure that commercial markets alone have been reluctant to build.
The resulting ecosystem would reduce one layer of external dependence.
Not all of them.
The Sovereignty Gap
Building European data centres does not automatically produce European technological independence.
The most obvious constraint remains hardware.
The world's most advanced AI processors continue to be designed and manufactured outside Europe. Even if Gigafactories operate entirely within the Union, much of their computing capability will still depend on imported semiconductor technologies and global manufacturing chains.
There is an equally significant dependency hidden deeper inside the software stack.
Nvidia's CUDA ecosystem has become the dominant platform connecting AI software with underlying hardware. Machine-learning frameworks are optimised first for CUDA, while developers accumulate years of specialised code, deployment pipelines, engineering expertise and optimisation routines around it.
That creates powerful network effects.
Replacing Nvidia hardware is therefore rarely a matter of purchasing different processors. Migrating to alternative platforms frequently requires rewriting software components, validating model behaviour again and accepting possible performance compromises during the transition.
For many organisations, those engineering costs outweigh any savings achieved by buying different hardware.
The result is that software lock-in becomes as strategically important as hardware availability.
Competition Beyond Nvidia
European policymakers appear aware of that problem.
The tender requires multi-vendor hardware architectures rather than systems tied permanently to a single supplier. That requirement creates opportunities for alternative semiconductor providers, including AMD, Intel, Qualcomm and European developers pursuing RISC-V-based technologies.
Yet hardware diversity alone will not guarantee meaningful competition.
Without broader support for open software ecosystems capable of reducing dependence on CUDA, many organisations may continue to favour Nvidia simply because their existing development environments already revolve around it.
The challenge therefore extends below the visible infrastructure.
Physical facilities can be financed relatively quickly. Software ecosystems evolve far more slowly because they depend on accumulated expertise, developer communities, optimisation libraries and production experience built over many years.
This is precisely where Europe's sovereignty ambitions become most difficult.
A New Measure of Strategic Power
The Gigafactory initiative signals that Brussels increasingly sees advanced computing infrastructure as a strategic capability rather than a purely commercial service.
That shift reaches beyond artificial intelligence itself.
High-density computing increasingly underpins industrial competitiveness, scientific research, public administration and defence applications. Decisions about where computing capacity resides are becoming intertwined with questions of economic security and national resilience.
Europe's investment acknowledges that reality.
Whether it ultimately reshapes the continent's technological position will depend on more than constructing larger facilities. Reducing reliance on foreign cloud providers is one objective, and the programme directly addresses it. Escaping dependence on imported chips and proprietary software ecosystems is another challenge entirely - one that cannot be solved simply by placing more servers on European soil.
Sources: European Commission
