Europe agog over its fastest supercomputer
Europe’s new computing powerhouse, Jupiter, is being inaugurated in Germany, promising a leap forward for climate science, medicine, energy research, and Europe’s bid to close the gap in artificial intelligence. The system is billed as Europe’s first true exascale machine—capable of at least one quintillion operations per second—and roughly comparable in raw computing punch to about a million smartphones working together.
What is the Jupiter supercomputer?
Jupiter resides at the Jülich Supercomputing Centre (JSC) in western Germany and is the continent’s first exascale system. Exascale denotes sustained performance at or above 1 exaflop—one billion billion calculations per second—unlocking simulations and AI workloads at unprecedented scale.
The machine occupies around 3,600 square meters (about 38,000 square feet), its halls lined with racks packed with processors and roughly 24,000 Nvidia accelerators—hardware currently favored for cutting-edge AI training. A consortium led by Eviden (an Atos subsidiary) and Germany’s ParTec engineered the system.
Jupiter’s development and early operation are budgeted at about 500 million euros, with funding split between the European Union and Germany. Access will be available to academic researchers across disciplines as well as companies, including for training large AI models.
According to JSC director Thomas Lippert, Jupiter marks a major step-change for European computing, delivering performance on the order of 20 times that of any other machine currently in Germany.
- Location: Jülich Supercomputing Centre, Germany
- Class: Exascale (≥1 exaflop)
- Accelerators: ~24,000 Nvidia GPUs
- Footprint: ~3,600 m² (≈38,000 sq ft)
- Budget: ~€500 million (EU and Germany)
- Average power draw: ~11 megawatts
Why Jupiter matters for Europe’s AI ambitions
Europe trails the United States and China in building and training frontier AI models. A recent Stanford University survey of “notable” models counted 40 from US-based institutions in 2024, 15 from China, and just three from Europe. Jupiter is the first system in Europe with the scale and interconnect to compete internationally for training state-of-the-art large language models (LLMs) and other foundation models.
Experts at the Barcelona Supercomputing Center note that bigger, faster systems enable larger training runs, often producing more capable models. Emmanuel Le Roux, who leads advanced computing at Eviden, calls Jupiter the largest AI machine in Europe—designed to support both scientific and industrial AI workloads.
That said, the system’s heavy reliance on Nvidia chips underlines Europe’s dependence on US semiconductor technologies, a strategic concern that European policymakers and industry are trying to address over the long term.
Beyond AI: climate, health, and energy breakthroughs
Jupiter’s performance isn’t just about AI. Its greatest scientific value may come from physics-based simulations that demand extreme scale:
- Climate modeling: Researchers aim to run higher-resolution, longer-horizon climate projections to better quantify risks of extreme events—heatwaves, heavy rainfall, droughts—and improve adaptation planning. While many models look roughly a decade ahead today, Jupiter-class resources could push credible forecasts to 30 years and, for some models, even toward century-scale scenarios.
- Neuroscience and drug discovery: More realistic simulations of brain function may offer new insights into neurodegenerative diseases such as Alzheimer’s, potentially accelerating therapeutic development.
- Energy transition: From wind farm aerodynamics and turbine blade optimization to grid stability and fusion research, large-scale simulation helps compress design cycles and reduce costs for clean technologies.
How Jupiter stacks up globally
The United States already operates multiple exascale-class supercomputers through the Department of Energy, and other regions are racing to join that club. Jupiter gives Europe its first seat at the exascale table and a national platform that can underpin homegrown AI models, bolster scientific collaborations, and support industry R&D without always relying on overseas resources.
Power draw and efficiency
Running at full tilt, exascale machines are energy-hungry. Jupiter’s average power consumption is estimated at around 11 megawatts—comparable to the needs of thousands of homes or a small industrial site. Yet the system’s designers emphasize energy efficiency: it uses the latest power-thrifty hardware, employs direct liquid cooling, and repurposes waste heat to warm nearby buildings. Operators say that makes Jupiter among the most energy-efficient of the world’s fastest systems.
The bottom line
With Jupiter, Europe gains a flagship supercomputer designed to accelerate science, sharpen industrial competitiveness, and help close the AI compute gap. Whether it’s unlocking longer-range climate insights, advancing biomedical research, or training the next generation of European AI models, Jupiter is poised to become a cornerstone of the continent’s digital and scientific infrastructure.
