Particle.news
Download on the App Store
6 ARTICLES
·
2d ago

Merz Inaugurates Jupiter, Europe’s First Exascale Supercomputer

EU-backed capacity at Jülich opens access for AI training across research and industry.

German Chancellor Friedrich Merz and Research, Technology and Space Minister Dorothee Baer attend the ceremonial launch of Jupiter, Nvidia's new high-performance computer, marking the start of operations for the first European supercomputer of the exascale class, at Juelich research centre, in Juelich, Germany, September 5, 2025. REUTERS/Jana Rodenbusch
German Chancellor Friedrich Merz reacts as he attends at the ceremonial launch of Jupiter, Nvidia's new high-performance computer, marking the start of operations for the first European supercomputer of the exascale class, at Juelich research centre, in Juelich, Germany, September 5, 2025. REUTERS/Jana Rodenbusch
German Research, Technology and Space Minister Dorothee Baer attends the ceremonial launch of Jupiter, Nvidia's new high-performance computer, marking the start of operations for the first European supercomputer of the exascale class, at Juelich research centre, in Juelich, Germany, September 5, 2025. REUTERS/Jana Rodenbusch
German Chancellor Friedrich Merz speaks at the ceremonial launch of Jupiter, Nvidia's new high-performance computer, marking the start of operations for the first European supercomputer of the exascale class, at Juelich research centre, in Juelich, Germany, September 5, 2025. REUTERS/Jana Rodenbusch

Overview

  • Officials said the system surpassed the exascale threshold at the inauguration, with the GPU Booster online now and a CPU cluster using SiPearl’s Rhea1 expected in 2026.
  • Jupiter occupies a site about half a football field in size and integrates roughly 24,000 Nvidia accelerators linked by Quantum-2 InfiniBand networking.
  • It placed fourth on the June 2025 TOP500 ranking and was described as the most energy-efficient system among the top five.
  • The €500 million project is jointly funded by EuroHPC and German authorities and was built by Eviden (Atos) and ParTec.
  • Operators plan to allocate compute time to researchers and companies for training AI models and for large-scale simulations in climate, medical and energy-transition research.