Elon Musk confirms the restart of the Dojo3 supercomputer project. Success in AI5 chip design has enabled Tesla to start porting resources over to Musk's next-gen supercomputer aimed at "space AI".
AMD and Eviden will build Alice Recoque, a €554-million exascale-class supercomputer for European AI and scientific research, powered by next-generation EPYC 'Venice' CPUs and Instinct MI430X GPUs. Deployment is expected later in the decade.
Nvidia, Oracle, and the U.S. Department of Energy will build seven ExaFLOPS-class AI supercomputers for Argonne National Laboratory — including the Oracle-built Equinox and Solstice systems with over 100,000 Blackwell GPUs delivering up to 2,200 FP4 ExaFLOPS — to power next-generation AI and scientific research.
Nvidia and HPE will build the Mission and Vision supercomputers for Los Alamos based on the Vera Rubin platform to advance national-security and open-science research using AI simulation and scientific computing.
Japan is investing over $750 million in FugakuNEXT, a zetta-scale supercomputer built by RIKEN and Fujitsu. Powered by FUJITSU-MONAKA3 CPUs and advanced accelerators, the system will integrate AI into scientific research, aiming to achieve performance 1,000× beyond current models.
The Top500 project's 65th list of performance results reveals U.S. leadership in supercomputing. The AMD-based El Capitan, Frontier, and Intel-powered Aurora take the top three spots. AI-focused systems like Microsoft's Eagle and Germany’s GH200-powered Jupiter Booster make early appearances. China submitted no new entries for this Top500 edition.