JUPITER: Europe's First Exascale Supercomputer Is Now Running — Specs, Performance & Why It Matters

For over a decade, the global supercomputing race has been dominated by American and Chinese systems. The United States achieved the exascale milestone first with Frontier at Oak Ridge National Laboratory in 2022, followed by Aurora at Argonne. China's exascale systems, while less publicly documented, have been operational in various capacities since roughly the same period.

Now, Europe has joined this elite group. The JUPITER (Joint Undertaking Pioneer for Innovative and Major Exascale Research) supercomputer, housed at the Jülich Supercomputing Centre in Germany, has reached operational status with performance exceeding one exaflop — one quintillion floating-point operations per second. And critically, JUPITER achieves this milestone using European-designed processor technology.

The Hardware: SiPearl Rhea1 Processors

What distinguishes JUPITER from its American counterparts is its reliance on the SiPearl Rhea1 processor, designed and developed by the French semiconductor company SiPearl as part of the European Processor Initiative (EPI). This represents a deliberate strategic investment in technological sovereignty — the ability to build world-class computing infrastructure without depending on foreign chip designs for critical components.

The Rhea1 is based on the Arm Neoverse V-series architecture and has been specifically engineered for high-performance computing and artificial intelligence workloads. Key specifications include:

• Architecture: Arm Neoverse V-series, custom extended for HPC/AI workloads
• Manufacturing Process: Advanced node fabrication with high transistor density
• Memory Subsystem: High Bandwidth Memory (HBM) integration for data-intensive computations
• Accelerator Integration: Tightly coupled with NVIDIA GH200 Grace Hopper accelerators for AI and simulation workloads
• Energy Efficiency: Designed to maximize performance per watt, addressing the growing concern over supercomputer power consumption

The system integration has been performed by Eviden (the digital arm of the Atos Group, operating under the Bull brand), which has a long history of assembling Europe's most powerful computing systems.

What Can an Exascale Machine Actually Do?

Raw performance numbers — a quintillion operations per second — are impressive but abstract. The practical significance of exascale computing lies in the scientific and engineering problems it can address that were previously computationally intractable:

Climate Modeling at Unusual Resolution

Current generation climate models operate at spatial resolutions of approximately 50-100 kilometers, meaning that phenomena smaller than a medium-sized country are essentially averaged out. JUPITER enables simulations at resolutions approaching 1-5 kilometers, allowing scientists to model individual thunderstorm systems, urban heat island effects, and regional precipitation patterns with far greater accuracy. For climate adaptation planning — deciding where to build flood defenses, how to manage water resources, which agricultural practices to adopt — this improvement in resolution translates directly into better decision-making.

Drug Discovery and Molecular Simulation

Simulating the behavior of complex protein molecules and their interactions with potential drug compounds requires impressive computational resources. At exascale, researchers can model molecular dynamics at biologically relevant timescales, potentially reducing the time from initial compound identification to clinical trial readiness from years to months. The European Molecular Biology Laboratory (EMBL) and multiple pharmaceutical research consortia have already been allocated computing time on JUPITER for drug discovery projects targeting antibiotic-resistant infections and neurodegenerative diseases.

AI Model Training

Training large-scale AI models currently requires massive GPU clusters that are predominantly controlled by a handful of American technology companies. JUPITER provides European research institutions and companies with sovereign access to exascale AI training infrastructure, reducing dependence on commercial cloud providers for the development of foundation models in scientific domains including medical imaging, materials science, and natural language processing for European languages.

Engineering Simulation

Full-resolution computational fluid dynamics simulations of aircraft, automotive, and turbine designs become feasible at exascale. Rather than testing simplified models or relying on wind tunnel experiments alone, engineers can simulate the complete aerodynamic behavior of complex geometries under realistic operating conditions, accelerating design optimization and reducing physical prototyping costs.

The European Supercomputing Ecosystem

JUPITER does not exist in isolation. It represents the flagship system within the EuroHPC Joint Undertaking, a partnership between the European Union, participating countries, and private technology partners that has deployed a network of high-performance computing systems across the continent:

• LUMI (Finland) — Currently Europe's most powerful pre-exascale system
• Leonardo (Italy) — Focused on industrial and scientific applications
• MareNostrum 5 (Spain) — Barcelona Supercomputing Center's flagship
• Deucalion (Portugal) — Supporting Atlantic research programs
• Karolina (Czech Republic) — Central European computing hub

This distributed infrastructure strategy ensures that computing resources are accessible across the continent rather than concentrated in a single location, while also providing redundancy and specialized capabilities for different research domains.

The Sovereignty Question

Beyond raw computing performance, JUPITER carries profound strategic significance. In an era where access to advanced computing infrastructure increasingly determines national competitiveness in science, industry, and defense, the ability to design, manufacture, and operate sovereign computing systems represents a critical capability.

The SiPearl Rhea1 processor demonstrates that European industry can produce competitive high-performance computing silicon. Combined with the recent European Chips Act investments aimed at increasing domestic semiconductor manufacturing capacity, JUPITER works as both a practical scientific instrument and a proof of concept for European technological self-determination.

What Comes After Exascale?

Planning has already begun for the next generation of European supercomputing systems. Research programs are exploring several frontier approaches:

• Quantum-Classical Hybrid Systems: Integrating quantum processing units as accelerators within classical supercomputing architectures for specific problem classes
• Neuromorphic Computing: Brain-inspired computing architectures that offer radical energy efficiency improvements for AI workloads
• Photonic Interconnects: Using light rather than electrical signals for inter-node communication, potentially eliminating a major performance bottleneck

The exascale era has only just begun, and Europe — powered by JUPITER — is positioned as a leading participant in defining what comes next.