EU Orders €388 Million AI Supercomputer

EU Orders €388 Million AI Supercomputer

Ondřej Barták
Ondřej Barták
Entrepreneur and Programmer
2. 9. 2026
4 minutes reading · 8 views
Listen to the article
Audio version of the article
EU Orders €388 Million AI Supercomputer

The European Union has awarded a contract for a new supercomputer designed for artificial intelligence. The tender was won by the French company Bull, with the contract valued at €387.8 million, approximately $449 million. It is the largest order in the company’s history. The machine, called LUMI-AI, will be built in the Finnish city of Kajaani and launched in the second half of 2027. Compared with the current LUMI supercomputer, it will handle ten times as much artificial intelligence work and nearly twice as many conventional scientific computations.

The order was placed by the EuroHPC Joint Undertaking, an EU organization that coordinates Europe’s publicly funded supercomputers. LUMI-AI is the sixth machine acquired under the so-called AI Factories program and will be built right next to the existing LUMI supercomputer in Finland. EuroHPC will remain the owner of the system, the LUMI AI Factory consortium will host it, and it will physically operate in the data center of the Finnish computing center CSC.

The funding is split equally. EuroHPC is providing half of the funds, while the other half has been contributed by the six countries in the consortium, namely Finland, Czechia, Denmark, Estonia, Norway, and Poland. Czechia is therefore one of the project’s co-owners, even though the machine itself will be located on Finnish soil, while researchers and companies from all partner countries will be able to use its computing power.

Europe needs another machine

Demand for computing power for artificial intelligence significantly exceeds the capacity currently available to EuroHPC. “At present, we have to reject some requests because we are unable to meet all the demand,” Evangelos Floros, who is responsible for infrastructure at EuroHPC, told the media. According to him, the pressure should ease as additional systems are gradually brought online.

The network that LUMI-AI will join has been under development for years. The EuroHPC Joint Undertaking has a long-term budget of €8.2 billion and is gradually building a network of specialized AI Factories centers connected to its supercomputers. Two examples show that this infrastructure is actually being used in practice. OVHcloud confirmed to Reuters that it had completed the first phase of training its foundation model on Germany’s Jupiter supercomputer. Meanwhile, Italian startup Domyn, which leads the European Commission-backed EUROPA consortium, is training its own model using European computing capacity.

What LUMI-AI will be able to do

The new machine will be capable of training advanced models, subsequently running them, and performing conventional scientific simulations. Bull will build it on its BullSequana XH3500 architecture, which it designed specifically for demanding artificial intelligence computations.

Computing power will be provided by AMD chips, specifically next-generation AMD Instinct MI430X accelerators and sixth-generation AMD EPYC processors with 256 cores. IBM will provide the storage, while Finnish company Nokia will supply the data center’s network infrastructure. The individual components will be connected using Bull’s BXI technology, the only European high-bandwidth, low-latency solution developed specifically for supercomputers and AI.

Kimmo Koski, head of the Finnish computing center CSC, mentioned another benefit. According to him, LUMI-AI, combined with the LUMI-IQ quantum computer, will help European scientists and companies conduct research in areas such as healthcare, climate, and the development of new technologies.

When the machine will become operational

The system is scheduled for deployment in the second half of 2027. Although that may sound like the distant future for artificial intelligence technology, it is a typical timeline for supercomputer contracts of this scale.

The Finnish city of Kajaani offers inexpensive electricity with a low carbon footprint and a climate that makes cooling large computing facilities significantly easier. Finland has also long based its strategy on turning these natural conditions into a competitive advantage for data centers.

Cooling is a discipline of its own for these facilities. LUMI-AI will use Bull’s patented solution, which delivers warm water to the chips instead of air. Waste heat produced by the machine will be captured, with the excess directed into Kajaani’s district heating network, increasing the share of environmentally sourced energy. The entire facility will run on renewable electricity in a building constructed according to strict environmental standards, making the system one of the world’s most energy-efficient large AI centers.

A small facility compared with American ones

In global terms, however, Europe’s AI Factories centers remain relatively modest. A supercomputer costing nearly €388 million represents only a fraction of the investment flowing into the vast campuses of private data center operators. So far, Europe has responded to this disparity primarily by pooling funds from individual governments rather than attempting to compete with the United States project by project.

Another program was intended to bring greater ambition. The €30 billion Gigafactories project was supposed to cover seven much larger sites. However, preparations have dragged on, and the delay is already making some of the partners Europe needs for the plan nervous. Smaller machines at established sites are easier to build. EuroHPC selected the first seven AI Factories sites in December 2024, followed by another six in October 2025, and the relevant contracts have been signed gradually ever since.

Sources: thenextweb.com and euronext.com

Advertisement

Content created with help from UpTier.

SEO and GEO on autopilot. UpTier’s multi-agent systems write and optimize content for search engines and AI answers.

Discover UpTier ↗

Category:AI
Did you enjoy this article?
Discover more interesting posts on our blog
Back to blog

Related posts

Amazon Is Buying Billions in Nvidia AI Chips but Has Nowhere to Plug Them InAmazon Is Buying Billions in Nvidia AI Chips but Has Nowhere to Plug Them In
Amazon is buying billions of dollars’ worth of Nvidia AI servers, but some are sitting in warehouses due to a power shortage. The company is paying for chips that are not yet generating revenue.
4 min read
22. 10. 2026
Artificial intelligence is learning the language of plants. Bombadil reads electrical signals from stems and treesArtificial intelligence is learning the language of plants. Bombadil reads electrical signals from stems and trees
Low-cost sensors capture electrical signals from trees and crops, and AI translates them into practical advice. Growers can detect stress earlier and save water.
6 min read
22. 10. 2026
Altman Outlines Three Endgames for the AI Race. In One, a Company “Melts” Everyone Else’s GPUsAltman Outlines Three Endgames for the AI Race. In One, a Company “Melts” Everyone Else’s GPUs
Sam Altman described three possible endings to the race for superintelligence: from international oversight and joint research to brutally cutting rivals off from chips.
5 min read
22. 10. 2026
Přihlaste se k odběru našeho newsletteru
Zůstaňte informováni o nejnovějších příspěvcích, exkluzivních nabídkách, a aktualizacích.
CodedTrip

Operated by CodedTrip LLC, USA.

YouTube
TikTok