Europe Signs €387.8 Million Deal for a Tenfold AI Computing Upgrade
EuroHPC signed a €387.8 million contract for LUMI-AI, a Finnish supercomputer expected to offer ten times the current LUMI system's AI capacity in 2027. The purchase names its major hardware and intended public users, but final scale, energy demand and measured performance remain undisclosed.
Europe has turned another part of its public AI-infrastructure plan into a signed purchase. EuroHPC awarded Bull a €387.8 million contract to build LUMI-AI in Finland, promising ten times the current system's AI capacity when the machine becomes available in 2027.
The 30-second summary
- What happened? EuroHPC signed a contract covering the acquisition, delivery, installation and maintenance of a new AI-focused supercomputer.
- Why does it matter? European startups, smaller companies and researchers are meant to gain shared access to computing for model training, inference and scientific simulation.
- What is the catch? The machine is not operating, its full configuration and absolute performance have not been published, and the tenfold claim compares it with the existing LUMI system.
Key Number: €387.8 million is the total contract value, split equally between EuroHPC and a six-country consortium led by Finland.
A contract rather than another computing promise
The European High-Performance Computing Joint Undertaking said on August 31 that it had signed the procurement contract after a tender launched in May 2025. The agreement covers the machine, installation, delivery and maintenance, with deployment planned for the second half of 2027 at CSC's data centre in Kajaani.
The public body will pay half of the cost through the Digital Europe Programme. Finland, Czechia, Denmark, Estonia, Norway and Poland will fund the other half through the LUMI AI Factory consortium. EuroHPC says it has now contracted six AI-optimised systems as part of a wider network of 19 AI factories.
The step is distinct from an earlier NewTqnia report about Brazil funding two proposed AI supercomputers. Brazil's announcement still left major procurement details unresolved. LUMI-AI now has a selected builder, named components, a signed budget and an access model, although delivery remains more than a year away.
What Bull plans to install
The system will use Bull's liquid-cooled BullSequana XH3500 architecture, AMD Instinct MI430X accelerators and sixth-generation AMD EPYC processors with 256 cores. IBM will supply storage technology, while Nokia and Bull will provide networking and interconnect components.
According to the LUMI AI Factory description, the machine should offer ten times the AI capacity and nearly twice the traditional high-performance-computing capability of the current LUMI supercomputer. That difference reflects the stronger advantage of modern accelerators in lower-precision AI workloads than in the highly precise calculations commonly used for scientific simulation.
Who is supposed to use it?
LUMI-AI is intended to train and run large AI models, process confidential datasets and support simulations in areas such as health, climate, manufacturing, materials and language technology. Its multi-tenant design and API-based access are meant to let several user communities share the infrastructure without operating their own supercomputers.
EuroHPC says startups and small or medium-sized companies will be primary users alongside universities and other researchers. Computing time will be divided by EuroHPC and the six-country consortium according to their investments. That makes the allocation rules, queues and practical cost of access as important as the peak hardware specification.
Cooling the machine and heating the city
Bull plans to cool LUMI-AI with warm water and send captured waste heat into Kajaani's district-heating network. The facility is also expected to use renewable electricity. Reusing heat can reduce wasted energy, but it does not remove the system's electricity demand or establish its efficiency before real workloads are measured.
NewTqnia has previously examined why AI's climate impact depends on what the computing actually enables. Cleaner power and recovered heat improve the infrastructure side of the equation, while the models and industrial applications determine much of the wider effect.
Before we overstate the result
The contract is real, but the performance is still a target. EuroHPC and Bull have not published the final accelerator count, total power demand, independently measured benchmarks or production reliability. The tenfold AI figure is relative to today's LUMI and cannot be compared directly with another country's system without a common benchmark. Installation in 2027 also leaves normal procurement, construction and integration risks.
What happens next
The decisive milestones will be the final configuration, installation tests and public access rules. LUMI-AI is also expected to connect with the planned LUMI-IQ quantum computer, but useful hybrid workloads will need software and experiments beyond simply placing two machines at one site. Until commissioning data appear, the signed contract proves Europe's spending commitment, not the promised computing performance.
Verified topics and entities
Sources and citations4 sources
External references used to support the reporting in this article.
Published by
NewTqnia Artificial Intelligence Desk
An institutional editorial team within NewTqnia
