AI BriefingJUPITER Unveiled at ISC Shows the Future of Exascale Science
Key point
Europe's first exascale supercomputer JUPITER is powering groundbreaking research such as brain mapping and climate simulation through NVIDIA technology.
Details
JUPITER, built at Germany's Forschungszentrum Jülich, is Europe's first Exascale supercomputer, powered by NVIDIA Grace Hopper Superchips and NVIDIA Quantum-X800 InfiniBand networking. This system is demonstrating the practical capabilities of exascale computing by solving complex scientific challenges that were impossible with existing hardware.
Key research achievements include the following projects:
- Brain Microstructure Analysis: Using a foundation model called CytoNet, the human brain—with 86 billion neurons and 100 trillion connections—is mapped at the cellular level. Using 4,096 NVIDIA Grace Hopper Superchips, 6.5PB of data was trained in just 5 days.
- High-Resolution Climate Simulation: The ICON model is used to simulate the Earth system—including oceans, atmosphere, land, and carbon cycles—at 1km resolution. Using 20,480 NVIDIA Grace Hopper Superchips, it set a world record by simulating 146 days of real climate data in 24 hours.
Going forward, researchers aim to integrate open models such as NVIDIA Nemotron 3 120B to build AI agents that allow scientists to directly query and analyze brain data.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Google Unveils Next-Generation Weather AI 'WeatherNext 3' Trained on Real-Time Satellite Data
How Meta AI models are leading Genesis Mission's first projectAnthropic tests Jupiter-v1-p ahead of developer conference