The European continent is accelerating the build-out of its own artificial intelligence computing infrastructure. The Brussels-funded EuroHPC joint venture has announced an order for the LUMI-AI supercomputer worth approximately 450 million dollars. The contract is the largest in the history of the French company Bull, which a few months ago was purchased by Paris from the loss-making Atos consortium for up to 404 million euros. In effect, the French state has redirected assets it had just acquired into a strategic project aimed at closing the technological gap with the United States and China.
Funding architecture and project geography
Funding for LUMI-AI is split evenly: half of the money is provided by the EuroHPC joint venture, and the other half by the LUMI AI Factory consortium, which includes six European countries. The new system will be installed at the Finnish supercomputing center, directly next to the already operational LUMI supercomputer, enabling the creation of a single high-performance cluster. Commissioning of the equipment for industrial use is scheduled for the second half of 2027. By design, LUMI-AI is intended for large-scale training and inference of AI models, as well as for running complex multidisciplinary scientific simulations — from climate modeling to the development of new materials.
Context: how Bull became an instrument of state AI strategy
The LUMI-AI contract takes on particular significance against the backdrop of recent corporate reshuffling in France. Just a few months before the agreement was signed, Paris completed the buyout of the Bull business unit from the Atos consortium, which had been operating at a loss. The deal was worth up to 404 million euros. In practice, the state has turned Bull into a "venture" asset for advancing a sovereign AI agenda: the company receives the largest contract in its history, while France gains a technological base independent of American and Chinese suppliers. Experts note that this "state buyout — state order" model is becoming a standard scenario for European technology companies amid geopolitical competition.
The EuroHPC network: 19 centers, 12 supercomputers, and expansion plans
LUMI-AI is not an isolated project but the latest element of a large-scale network that EuroHPC has been building since 2021. The total budget of the initiative for the 2021–2027 period is 8.2 billion euros. To date, the organization has already deployed a network of 19 so-called "AI factory" centers, built on 12 supercomputers, including Germany's Jupiter. EuroHPC is currently preparing a selection of up to seven new locations, signaling an intention to double the continent's computing power by the end of the current decade. It was on Jupiter that Europe's largest cloud provider, OVHcloud, completed the initial training of its foundational AI model, while the Italian startup Domyn, which leads the EUROPA consortium, gained access to EuroHPC systems to train its own 400-billion-parameter model.
European AI players: OVHcloud and Domyn in the race for the lead
Both companies — OVHcloud and Domyn — do not disclose detailed technical specifications of their models, but their strategic goal is clear: to become European leaders in advanced artificial intelligence. For OVHcloud, already working with the Jupiter infrastructure, access to LUMI-AI will open up opportunities to scale inference and fine-tuning of models for corporate clients. For Domyn, with its 400-billion-parameter model, the EuroHPC contract means access to computing resources comparable to those of the largest American laboratories, which is critical for competitiveness in the foundation models segment.
Contradictory data
Cross-checking of sources revealed a discrepancy in the wording of the contract value. The Ukrainian outlet RBC.UA (source 1) and the main news text cite the cost of LUMI-AI as "around 450 million dollars." Meanwhile, the business publication Delo.UA (source 2) names the figure as "390 million euros." Most likely, the two refer to the same contract expressed in different currencies: at an euro-to-dollar exchange rate in the range of 1.15–1.16, exactly 390 million euros is equivalent to approximately 450 million dollars. Nevertheless, the absence of a single canonical source (an official EuroHPC or Bull press release) with the exact figure in one currency creates uncertainty, which the editorial team records as a caveat. Until official confirmation is available, both figures are presented as alternative formulations of the same funding volume.