News
DealsCorvex agrees USD 103 million take-or-pay deal to perform AI inference with Mid-Atlantic GPUsSovereignNeysa scales AI infrastructure in India with AMD Instinct MI350X GPUs on TytanInferenceMicrosoft Foundry adds Baseten, DeepSeek AI inference (MSFT:NASDAQ)ClustersDataOne and Rowan University sign engineering co-op program agreement for DataOne's Vineland AI factoryClustersLook inside Cal Poly's $3 million AI Factory (Video)DealsOpenAI Taps AMD Chips for 457 MW of AI Capacity in Six-Year TensorWave DealClustersCal Poly launches an AI data center on campus. Take a look insideInferenceArista takes aim at the next AI network battleground: distributed inferenceDealsOpenAI Signs 457 MW AI Data Center Deal as AMD Expands Its Challenge to NvidiaClustersBehind Dell AI Factory’s NVIDIA & AMD Enterprise ExpansionDealsCorvex agrees USD 103 million take-or-pay deal to perform AI inference with Mid-Atlantic GPUsSovereignNeysa scales AI infrastructure in India with AMD Instinct MI350X GPUs on TytanInferenceMicrosoft Foundry adds Baseten, DeepSeek AI inference (MSFT:NASDAQ)ClustersDataOne and Rowan University sign engineering co-op program agreement for DataOne's Vineland AI factoryClustersLook inside Cal Poly's $3 million AI Factory (Video)DealsOpenAI Taps AMD Chips for 457 MW of AI Capacity in Six-Year TensorWave DealClustersCal Poly launches an AI data center on campus. Take a look insideInferenceArista takes aim at the next AI network battleground: distributed inferenceDealsOpenAI Signs 457 MW AI Data Center Deal as AMD Expands Its Challenge to NvidiaClustersBehind Dell AI Factory’s NVIDIA & AMD Enterprise Expansion
EveryNeocloud
Group photo with Henna Virkkunen, 5th from the left, during her visit of LUMI the super computer and the new CSC Data Center

LUMI

EuroHPC petascale machine in Kajaani, Finland, built by HPE on AMD hardware

Photo: Europäische Kommission - Audiovisueller Dienst, Alessandro Rampazzo, C, CC BY 4.0, via Wikimedia Commons

LUMI is a petascale supercomputer housed in a former paper mill in Kajaani, Finland, and switched on in June 2022. Hewlett Packard Enterprise built it for the EuroHPC Joint Undertaking and a consortium of eleven European countries, and it became the fastest machine in Europe in January 2023.

Share on WhatsApp

  • 10,240 AMD Radeon Instinct MI250X accelerators
  • 379 petaFLOPS measured, 531 petaFLOPS peak
  • 117 petabytes of storage at 2 terabytes per second
  • Runs on hydroelectric power and heats nearby buildings
  • €144.5 million budget shared across eleven countries

What it is

LUMI, short for Large Unified Modern Infrastructure, is a petascale supercomputer sitting in the CSC data centre at Kajaani, Finland. Hewlett Packard Enterprise supplied the hardware, an HPE Cray EX system mixing AMD EPYC processors with Radeon Instinct accelerators. Most of its arithmetic throughput comes from the graphics units rather than the CPUs, a choice made on cost per unit of performance. The machine draws 7.1 MW, holds 1.75 petabytes of memory and offers 117 petabytes of storage. In January 2023 it took the title of Europe's quickest computer, and the June 2025 TOP500 listing placed it ninth worldwide.

How it came about

Pilot work on the CPU section began in 2021, with the first batch of projects chosen by June of that year and a September start planned. Full service, graphics partition included, followed in 2022. The system went live on 13 June 2022 and reached the top of the European rankings in January 2023. On 31 August 2026 the EuroHPC Joint Undertaking and the LUMI AI Factory consortium gave Bull, a French state-owned firm, a €387.8 million deal to design, deliver, install and look after a fresh AI-focused supercomputer.

The compute

The build pairs 64-core AMD EPYC CPUs with 10,240 AMD Radeon Instinct MI250X accelerators, giving 362,496 processor cores and more than 144 million graphics cores. Memory totals 1.75 petabytes. Storage splits into a 7-petabyte flash tier and an 80-petabyte conventional tier, both running the Lustre parallel file system, plus a 30-petabyte Ceph-based data handling service. Combined capacity reaches 117 petabytes with 2 terabytes per second of aggregate input/output bandwidth.

Where the GPUs are

Everything sits in the CSC facility at Kajaani, northern Finland, occupying 150 square metres inside a disused UPM paper mill. That site drew on hydroelectric power throughout, and waste heat is captured to warm nearby buildings. Grid reliability there is notable: across 38 years of mill operation the location recorded just one two-minute interruption.

LUMI
A drawing made by EveryNeocloud from what this page says. It is a diagram, not a picture of the real provider.

Customers and deals

Half of the machine's time belongs to the EuroHPC Joint Undertaking, and a fifth of that share is set aside for companies and small and medium enterprises. The remaining half is divided among the eleven consortium nations in proportion to what each paid in. Pilot projects were picked for the CPU partition before the graphics section opened.

Money and ownership

EuroHPC Joint Undertaking and the LUMI Consortium share the bill, with a total budget of €144.5 million. Consortium members are Finland, Belgium, the Czech Republic, Denmark, Estonia, Iceland, Norway, the Netherlands, Poland, Sweden and Switzerland. A separate €387.8 million contract went to Bull in August 2026 for a new AI-optimised system.

Its place in AI

The name is the Finnish word for snow. LUMI reached first place among European supercomputers in January 2023 and sat fifth on the November 2022 TOP500 with a measured 309.1 PFLOPS. By June 2025 it ranked ninth. It belongs to the EuroHPC family of machines, alongside Leonardo in Bologna, Italy.

Questions and criticism

Its power demand is 7.1 MW, though the site runs entirely on hydroelectric generation and recycles heat into local buildings. Access is shared between a European body and eleven national funders, so allocation follows financial contribution.

Questions people ask about LUMI

What does LUMI stand for?

The acronym expands to Large Unified Modern Infrastructure. The word itself is Finnish for snow, fitting for a machine installed in Kajaani, a town in central Finland.

Who paid for it?

Costs were split between the EuroHPC Joint Undertaking and the LUMI Consortium, a group of eleven European nations. The overall budget came to €144.5 million, and each country's share of compute time reflects its financial contribution.

How fast is it?

Measured performance is 379 petaFLOPS, with a theoretical ceiling of 531 petaFLOPS. The November 2022 TOP500 list recorded 309.1 PFLOPS and fifth place; by June 2025 the machine sat ninth on that ranking.

Who can use it?

Half the capacity goes to the EuroHPC Joint Undertaking, which reserves a fifth of its portion for industry and SMEs. The other half is distributed among the consortium countries according to how much each contributed to the budget.

What is happening next?

On 31 August 2026, EuroHPC and the LUMI AI Factory consortium contracted Bull, a French state-owned company, for €387.8 million to design, deliver, install and maintain a new supercomputer tuned for AI work.

Good reading and sources

Where this page comes from

The facts on this page come from the Wikipedia article "LUMI", written by its contributors and used under the CC BY-SA 4.0 licence, and from its Wikidata record (CC0). EveryNeocloud rewrote the text in its own words, with AI assistance, and compares every page with the article to make sure no passage is copied. It may be reused under the same CC BY-SA 4.0 licence with a link back.

It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.

Photo credits on this page

The small pictures above come from Wikimedia Commons under free licences. Some are cropped to fit.

The weekly digest

Every Monday: the provider of the day, new pages on EveryNeocloud, and the conferences and anniversaries of the week ahead.