The Jupiter AI processing system is projected to be completed this year by the European Supercomputing Joint Undertaking (EuroHPC JU) due to its modular architecture.
The EuroHPC JU of Germany and France launched the Jupiter supercomputer project in October 2023, and it is scheduled to be finished in less than a year, with the objective of being Europe’s first exascale supercomputer, meaning it can do more than one trillion computations per second.
Jupiter will employ a set of Nvidia’s GH200 super processors to train AI and is said to be the world’s most powerful AI system, capable of training AI models at a rate of roughly 90 exaflops.
Jupiter’s brain consists of 50 container modules spread across 2,300 square meters. The container blocks hold around 20 computer systems, 15 energy modules, and ten logistical modules. This strategy saves money, cuts system development time by half, and simplifies future upgrades.
Frontier, the first exascale supercomputer and considered to be the only one in the world today, was launched in 2021 by Hewlett Packard Enterprise (HPE) and is housed at the United States’ Oak Ridge National Laboratory. Lumi, owned by EuroHPC JU and housed in the CSC data center in Finland, is presently Europe’s fastest supercomputer, capable of doing 375 billion billion computations per second and ranked third in the world.
Supercomputers are massive computer systems with millions of times more computational power than regular computers, capable of addressing big global issues. They are used in virtually every discipline, including nuclear test simulation, weather forecasting, climate study, and computer encryption strength testing.
In recent years, the United States and China have engaged in a dual-code race, which is utilized for both industrial and military research. Technology corporations like as Facebook and Google have lately joined the building of supercomputers to support AI research.
