The LUMI-AI supercomputer being built in Finland will run on Advanced Micro Devices (AMD) Instinct MI430X chips under a $420 million EuroHPC contract awarded Monday to Bull, the French systems vendor owned by Eviden.

Key Takeaways

  • EuroHPC awarded Bull a 387.8 million euro contract to build the LUMI-AI supercomputer in Finland

  • The system uses AMD Instinct MI430X accelerators paired with sixth-generation EPYC processors

  • The original LUMI supercomputer cost roughly 144 million euros and went online in 2021 in Kajaani

  • LUMI-AI is expected to come online within the next two years, per EuroHPC’s typical buildout timeline

The LUMI-AI Supercomputer’s Costly Leap From CPUs To AI Chips

EuroHPC Joint Undertaking, the public-private partnership that pools European Union and national funding to build shared supercomputing infrastructure, selected Bull for the project. The 387.8 million euro contract was detailed in a GlobeNewswire release posted Monday.

The system pairs AMD Instinct MI430X accelerators, chips built to train and run large AI models, with sixth-generation EPYC processors that manage data traffic.

The company said in the announcement that the buildout would strengthen Europe’s sovereign AI capacity and accelerate scientific discovery once operational.

How Finland’s Original LUMI Machine Set The Price Bar

The original LUMI supercomputer went online in 2021 at CSC IT Center for Science’s data hall in Kajaani, built mainly for physics and climate research. That machine cost roughly 144 million euros, funded jointly by EuroHPC and a ten country Nordic and Central European consortium.

The LUMI-AI supercomputer, the AI-dedicated expansion announced Monday, costs almost three times as much.

That jump reflects how much pricier AI accelerator chips have become than the general-purpose processors that powered the original build five years ago.

Sovereign Compute Becomes Europe’s Priciest Bet

The LUMI-AI supercomputer is expected to come online within the next two years, EuroHPC’s typical buildout timeline for major systems. It arrives as European governments push to reduce reliance on American cloud providers such as Amazon, Microsoft and Google for AI training capacity.

Whether the investment pays off depends on demand from European researchers who might otherwise rent capacity from US hyperscalers.

The chipmaker’s stake in the LUMI-AI supercomputer deepens its footprint in state-backed AI infrastructure, following similar deals in Saudi Arabia this year.