AMD Instinct MI430X GPUs, EPYC CPUs Power €387.8M LUMI-AI Supercomputer

The EuroHPC Joint Undertaking has signed a €387.8 million procurement contract with French vendor Bull to deliver LUMI-AI, a next-generation supercomputer announced August 31, 2026 as powered by AMD Instinct MI430X GPUs and 6th Gen AMD EPYC 256-core processors. The system will be installed at CSC – IT Center for Science’s new data center in Kajaani, Finland, with deployment scheduled for the second half of 2027.

The contract covers acquisition, delivery, installation and maintenance of the system. Funding is shared equally between the EuroHPC JU — whose half comes from the European Union’s Digital Europe Programme — and the six-country LUMI AI Factory consortium of Finland, Czechia, Denmark, Estonia, Norway and Poland, according to the EuroHPC JU. The system will be owned by the EuroHPC JU and hosted by the consortium.

Expected Performance and System Architecture

The organizations behind the project state that LUMI-AI is expected to provide ten times the AI capacity and nearly twice the high-performance computing capability of the current LUMI supercomputer. The official LUMI release attributes the gap between those two gains to newer GPU architectures, which it says are significantly more performant in AI-driven tasks, while traditional computing performance describes highly precise floating-point calculations used in scientific simulations.

The liquid-cooled system will be based on Bull’s BullSequana XH3500 architecture, which Bull describes as purpose-built for advanced AI workloads. Additional technology partners include IBM, providing a storage solution based on IBM Storage Scale, and Nokia, contributing data center networking expertise alongside Bull’s BXI interconnect, which Bull describes as Europe’s only high-bandwidth, low-latency interconnect designed for HPC and AI environments.

According to AMD’s product page, the Instinct MI430X is designed for scientific HPC and sovereign AI, combining FP64 accuracy for simulation and modeling with AI capability for inference, training and fine-tuning. AMD lists the GPU at up to 288 TFLOPs of hardware-based peak theoretical FP64 performance, up to 9.2 PFLOPs of peak theoretical FP4 performance, and up to 432 GB of integrated HBM4 memory with up to 23.3 TB/s of peak theoretical memory bandwidth. AMD says the MI430X supports the open ROCm software platform and is expected to be available in 2027, with the US Department of Energy’s Discovery system at Oak Ridge National Laboratory and Europe’s Alice Recoque and Herder systems also expected to deploy it.

“Europe’s next wave of AI innovation will depend on infrastructure that is open, efficient and trusted,” said Thomas Zacharia, senior vice president of Global Public Sector and Strategic Partnerships at AMD.

Role in the LUMI AI Factory

LUMI-AI will serve as the computing backbone of the LUMI AI Factory, supporting industrial AI innovation, academic research, AI training and inference, and large-scale scientific simulation. The EuroHPC JU states the system will support next-generation AI workflows characterized by large-scale, dynamic and often confidential datasets.

The LUMI AI Factory has provided AI computing resources and services to European small and medium-sized enterprises and startups since April 2025, using the current LUMI supercomputer across manufacturing, health and life sciences, and communication technologies and networks, while also serving climate research, materials science and language technology communities. The EuroHPC JU says LUMI-AI is primarily designed to serve startups and SMEs for research and innovation, will also be accessible to the wider research community, and that supercomputing resources will be allocated jointly by the EuroHPC JU and the consortium in line with their respective investments. It is the sixth AI Factory procurement contract the EuroHPC JU has signed for a next-generation AI-optimized supercomputer.

The computing environment will be complemented by LUMI-IQ, a quantum computer to be delivered by IQM Quantum Computers. The LUMI consortium says LUMI-IQ will be located in the same data center and tightly integrated with LUMI-AI, forming a hybrid platform for workloads combining HPC, AI and quantum computing.

Site, Construction and Energy Design

LUMI-AI will be housed in CSC’s new data center in the Renforsin Ranta business park in Kajaani, on the site of a former paper mill. CSC states that construction of the expansion began in January 2026, with completion expected in summer 2027, and that the current LUMI supercomputer was inaugurated in Kajaani in 2022.

The organizations state that LUMI-AI will be powered entirely by renewable energy and that its liquid-cooling design will capture excess heat for reuse in Kajaani’s district heating network. Bull says its patented Direct Liquid Cooling technology uses warm water to cool the system, and the LUMI consortium says the approach will significantly increase the share of district heating generated using excess heat from CSC’s data centers.

Bull was selected following a call for tender the EuroHPC JU launched in May 2025, and Finland was selected as the hosting site in December 2024. The EuroHPC JU states the system is expected to be installed and made available to users in 2027.