Europe is moving ahead with another major investment in sovereign artificial intelligence infrastructure. The EuroHPC Joint Undertaking has signed a procurement contract for LUMI-AI, a new supercomputer that will be built in Kajaani, Finland, by Bull, the French supercomputer manufacturer now fully owned by the French state. The contract is valued at €387.8 million, or roughly $449 million, and represents the sixth AI Factory procurement deal under EuroHPC's expanding programme.
Key facts at a glance
- Contract value: €387.8 million (~$449 million).
- Manufacturer: Bull, owned by the French state since 31 March.
- Location: Kajaani, Finland, at the existing LUMI data centre.
- Funding: split between EuroHPC and the LUMI AI Factory consortium, with contributions from Finland, Czechia, Denmark, Estonia, Norway and Poland.
- Processors: AMD Instinct MI430X accelerators and sixth-generation AMD EPYC CPUs with 256 cores.
- Architecture: BullSequana XH3500 with Bull's BXI interconnect and patented warm-water direct liquid cooling.
- Storage and networking: IBM Storage Scale and Nokia data centre networking.
- Expected deployment: second half of 2027.
- Capacity: roughly ten times the AI capacity of the existing LUMI system, and nearly double its conventional high-performance computing capability.
A shared European investment
The new machine is not funded by Finland alone. The cost will be split evenly between EuroHPC and the LUMI AI Factory consortium, which includes Finland, Czechia, Denmark, Estonia, Norway and Poland. This arrangement means the supercomputer will physically sit on Finnish soil but will be accessible to researchers and companies from all participating countries. The consortium model has become a recurring feature of European supercomputing, allowing smaller and mid-sized governments to pool resources and gain access to infrastructure that no single country could easily fund or operate on its own.
LUMI-AI is expected to deliver around ten times the AI capacity of the current LUMI supercomputer, which has been operating in Kajaani since 2021. The new system will also nearly double the conventional high-performance computing capability of the existing installation. Deployment is scheduled for the second half of 2027, a timeline that might appear slow for an AI market moving at breakneck speed, but it is consistent with the complexity of building large-scale supercomputers.
Hardware details and European components
The hardware choices reveal a mix of American processing power and European engineering. LUMI-AI will be equipped with AMD Instinct MI430X accelerators and sixth-generation AMD EPYC processors with 256 cores, all built around Bull's BullSequana XH3500 architecture. The accelerators, which are essential for AI workloads, come from an American vendor, illustrating a persistent dependency that Europe is still trying to address.
However, the system includes several components developed in France. Bull is supplying its BXI interconnect, the high-speed network that links the nodes together, as well as its patented warm-water direct liquid cooling technology. Both have been developed in France and are seen as strategic technologies in their own right. The cooling system is particularly relevant in Finland, where low ambient temperatures make energy-efficient cooling easier, but where the density of modern AI chips demands sophisticated liquid cooling regardless of climate conditions.
The wider supplier list also points to a deliberate effort to keep more of the infrastructure within European borders. IBM Storage Scale will provide the storage layer, and Nokia is supplying the data centre networking. Nokia's involvement brings a Finnish company into a system that will be installed in Finland, adding a local dimension to the project.
Bull's new state-owned status
One of the most significant aspects of this procurement is the ownership of the manufacturer. Bull was bought outright by the French state on 31 March, in a transaction that followed Atos's sale of its Advanced Computing business. The enterprise value of the deal was up to €404 million, including €104 million in earn-outs. The French government is now the company's sole shareholder, a situation that would have been highly unusual in the supercomputing industry only a few years ago.
Bull's strategic importance goes beyond commercial supercomputing. The company operates Europe's only supercomputer manufacturing plant, located in Angers, and builds the systems used to model France's nuclear deterrent. This dual-use character helps explain why the French government was unwilling to see the business fall into foreign hands. The company generated roughly €700 million in revenue in 2025, and its order book includes both military and civilian clients.
Emmanuel Le Roux, Bull's chief executive, described LUMI-AI as an important milestone for European AI and for Bull's leadership in energy-efficient, complex AI infrastructures. Anders Jensen, who runs EuroHPC, framed the purchase as the sixth step in a broader sequence of AI infrastructure investments, rather than an isolated project. Both statements reflect a growing consensus in European policy circles that AI capacity is a strategic resource.
Kajaani as a supercomputing hub
Finland's Kajaani has become an unlikely hub for high-performance computing. The city, located in the north of the country, offers two critical advantages: abundant cheap, low-carbon electricity and a cold climate that makes it significantly easier to cool large computing installations. Finland has spent years building the infrastructure needed to turn these natural conditions into a competitive advantage for data centres and supercomputing.
The existing LUMI supercomputer has already demonstrated the value of this approach. Kimmo Koski, who runs CSC, the Finnish research organisation that will host LUMI-AI, emphasised that the new system is a continuation of what the LUMI consortium has already built, not a project starting from scratch. That track record is part of the reason six countries are prepared to pool funding around a facility located in only one of them.
With Germany's Jupiter now joining the European supercomputing landscape, LUMI remains one of the continent's most established large-scale systems. The addition of LUMI-AI will further cement Kajaani's position as a key site for European scientific computing and AI research.
AI Factories and the bigger European strategy
The LUMI-AI contract is part of a broader EuroHPC strategy to create a network of AI Factories across Europe. The first seven AI Factory sites were selected in December 2024, followed by six more in October 2025. Procurement contracts have been signed progressively since then, and LUMI-AI is the sixth of those agreements.
The AI Factory model is designed to provide researchers and companies with access to specialised AI computing resources, complementing the general-purpose supercomputers that EuroHPC has already deployed. The hope is that these facilities will support the development of European AI models, applications and services, reducing the continent's reliance on cloud providers and computing platforms from outside the European Union.
Beyond the AI Factories, Europe had planned a more ambitious programme of gigafactories. The gigafactories were intended to take the effort to another scale, with a €30 billion programme covering seven much larger sites. Progress on that part of the strategy has been slower than hoped. The gigafactory programme has struggled with delays that have frustrated some of the partners it needs, and the projects themselves are considerably more complicated than building an individual supercomputer at an established site.
The gap between the AI Factories and the gigafactories is significant. Even the AI Factories that are now being procured remain relatively small compared with what the biggest private technology companies are building. A €387.8 million supercomputer represents only a fraction of the investment behind a single hyperscaler campus. Europe's response to that imbalance has so far relied heavily on pooling resources between governments rather than trying to match the private sector project for project.
Sovereignty and dependence
The decision to buy from a state-owned supplier is perhaps the clearest indication of how much the European approach has changed. Three years ago, government ownership of a major supercomputer manufacturer would have been an unusual detail in a procurement story. Today, it is a central part of the strategic discussion around technological sovereignty.
Yet the LUMI-AI system also highlights the limits of that sovereignty. The accelerators at its core are still made by an American company, AMD. While Europe has developed significant capabilities in interconnect technology, cooling, system integration and some areas of software, the most advanced AI chips remain firmly outside European control. That combination captures both the progress Europe has made towards technological sovereignty and the parts of the stack where it remains dependent on foreign suppliers.
European policymakers have long recognised this dependency as a vulnerability, particularly as AI becomes more central to economic competitiveness and national security. Several initiatives are underway to develop European chip manufacturing and AI accelerator designs, but these efforts are still in their early stages. The LUMI-AI contract shows that Europe can assemble a world-class AI supercomputer using its own integration skills and infrastructure, but it also shows that the processor technology that powers such systems is still imported.
Timeline and delivery
The contract for LUMI-AI is now signed, the site is established, and the hardware has been chosen. The next phase will involve detailed system design, manufacturing, installation and testing. Bull will need to coordinate with AMD, IBM, Nokia and a range of other suppliers to deliver the system on schedule.
For EuroHPC, the sixth AI Factory contract is a milestone, but it also raises questions about the pace of delivery. Some observers have argued that Europe's public funding cycles and procurement rules are too slow for the fast-moving AI sector. By the time LUMI-AI is deployed in the second half of 2027, the AI landscape could look very different from today. Yet the need for sovereign, energy-efficient computing infrastructure is unlikely to diminish.
The real test will come when LUMI-AI is operational, when Europe will find out how much of its AI infrastructure strategy can translate from coordinated procurement into computing capacity that researchers and companies can actually use. For now, the project stands as a concrete example of what European governments can achieve when they combine financial resources, political will and industrial capability.