All articles

EuroHPC Explained: High-Performance Computing in the EU

EuroHPC explained: how the EU's Joint Undertaking pools resources to build world-class supercomputers like LUMI, Leonardo and JUPITER, plus the HPC concepts — parallelism, clusters, exascale — behind them.

Prep4EU Insight No single member state can easily fund a world-class supercomputer. EuroHPC is the EU's answer: pool the money, build shared exascale machines across Europe, and give researchers and industry access to computing power that rivals the US and China — a question of digital sovereignty as much as of science.

What it is

High-Performance Computing (HPC) means using supercomputers — vast clusters of tightly connected machines working in parallel — to solve problems far beyond the reach of an ordinary computer: climate modelling, drug and materials discovery, engineering simulation and training large AI models. The EuroHPC Joint Undertaking (EuroHPC JU) is the EU body that makes this happen at European scale. It is a legal and funding partnership between the EU, participating countries and private partners, created in 2018 and reinforced in 2021, that co-invests in building and operating supercomputers and in the research, applications and skills around them.

The point is pooling. Rather than each country trying — and mostly failing — to fund a top-tier machine alone, EuroHPC combines EU budget with national contributions to procure shared systems hosted in different member states but available to users across Europe. The result is a fleet of some of the world's most powerful computers operating under EU governance.

How it works in practice

Three things make HPC, and EuroHPC, work.

1. Massive parallelism. A supercomputer is built from thousands of nodes (each a server with many CPU cores, often paired with GPU accelerators). A problem is split into pieces that run simultaneously across them. Performance is measured in FLOPS (floating-point operations per second): petascale systems do 10¹⁵ FLOPS, and exascale systems reach 10¹⁸ — a billion billion calculations per second.

2. Fast interconnect. Parallelism only pays off if the nodes can exchange data quickly, so HPC clusters use very low-latency, high-bandwidth networks (such as InfiniBand) and a communication standard, MPI (Message Passing Interface), that lets processes coordinate across nodes. The interconnect is as defining as the processors.

3. Shared, scheduled access. Many users compete for one machine, so jobs are submitted to a batch scheduler that queues and allocates resources. Software environments must be portable and reproducible, which is why containerisation increasingly appears in HPC workflows.

EuroHPC has used this model to deploy a recognisable fleet: LUMI in Finland, Leonardo in Italy and MareNostrum 5 in Spain are among Europe's pre-exascale flagships, and JUPITER in Germany is Europe's first exascale system. These machines are heavily used for climate science, AI and industrial research.

Term Meaning
Node One server in the cluster (CPUs, memory, often GPUs)
FLOPS Floating-point operations per second — the speed measure
Petascale / Exascale 10¹⁵ / 10¹⁸ FLOPS performance tiers
MPI Message Passing Interface — how parallel processes communicate
Interconnect Low-latency network linking the nodes (e.g. InfiniBand)
EuroHPC JU The EU Joint Undertaking that funds and governs the machines

The EU dimension is the whole rationale. EuroHPC is explicitly about digital sovereignty — ensuring Europe owns and controls world-class compute rather than depending on others — and it connects to adjacent EU initiatives: the European Quantum Communication Infrastructure (EuroQCI) and the push to develop European processor technology. Access is allocated to researchers, public bodies and industry (including SMEs) across the Union, making top-tier computing a shared European capability rather than a national luxury.

Common points of confusion

Why it matters for EU infrastructure specialists

HPC is where infrastructure meets EU strategy, and EuroHPC is the flagship example a competition is likely to reach for. For EPSO/AD/429/26 Field 1 (ICT Infrastructure), expect questions on what the EuroHPC Joint Undertaking is and why it exists (pooling, sovereignty), on naming or recognising its supercomputers, or on core HPC concepts such as parallelism, FLOPS, exascale and MPI. Knowing both the technology and the EU policy context is exactly the blend the exam tests. Build that fluency with the full study pack: Prep for AD7 ICT Infrastructure on Prep4EU

Related guides

Ready to start preparing?

Practice MCQs in the exact EPSO format with instant feedback and explanations.

Start Learning