Jupiter: Europe's Fastest Supercomputer With NVIDIA Grace Hopper Boots Up Successfully

Posted on Edited on
Europe's Jupiter supercomputer powered by NVIDIA Grace Hopper technology delivers 930 PetaFLOPS of performance, becoming the continent's fastest and world's 4th most powerful system.

Quick Report

Europe's high-performance computing sector has received a significant boost as the Jupiter supercomputer was officially powered on at the Jülich Supercomputing Center in Germany. The modular exascale system delivers a peak of nearly one ExaFLOP (930 PetaFLOPS) of FP64 performance, making it the fastest machine in Europe and the fourth most powerful supercomputer globally according to the Top500 list.

Built around NVIDIA's Grace Hopper technology, Jupiter's booster module contributes approximately 793 PetaFLOPs of the system's total computational power. Each compute node features four GH200 accelerators, resulting in a massive deployment of 23,752 accelerators spread across 5,938 nodes and housed in 125 server racks.

Energy Efficiency and AI Performance:

The GH200 accelerators powering Jupiter are noted for their impressive energy efficiency, delivering nearly 60.5 GigaFLOPS per watt. Despite this efficiency, the accelerator farm still requires double-digit megawatts of power to operate at full capacity.

While Jupiter excels at traditional high-performance computing tasks, it's primarily designed for AI workloads. In AI's preferred FP8 format, the system delivers over 90 ExaFLOPS of performance, making it an exceptionally powerful platform for machine learning and artificial intelligence research.

Investment and Future Expansion:

The reported build costs for Jupiter are estimated at approximately €500 million, but European planners are already looking ahead. To close capacity gaps and maintain competitiveness in the global supercomputing landscape, a multibillion-euro expansion is anticipated, with figures around €60 billion cited for deployments through 2030.

This ambitious expansion plan envisions approximately 13 specialized AI data centers and gigafactory-scale facilities, with the Jülich Supercomputing Center being a candidate for further development. European officials emphasize that this investment is about digital sovereignty—keeping critical computing infrastructure close to European research institutions and industry rather than depending on foreign cloud providers.

Applications and Significance:

Operationally, Jupiter will support a wide range of scientific applications, including:

  • Climate modeling and environmental simulation
  • Materials science research
  • Biological and medical research
  • AI model training and development

The system will be complemented by an inference module called Jarvis, designed for fast model serving after training on the main system.

For Europe, Jupiter represents both a significant computing capability and a strategic signal: an energy-conscious exascale platform designed to anchor AI work and catalyze further regional investment in advanced computing technologies.

Written using GitHub Copilot Claude 3.7 Sonnet in agentic mode instructed to follow current codebase style and conventions for writing articles.

Source(s)

TechPowerUp
HardwareLuxx