NVIDIA technologies now power 81 percent of the world's 500 fastest supercomputers, with momentum accelerating as nine o
NVIDIA technologies power more than 400 of the world's 500 fastest supercomputers, representing 81 percent of the TOP500 according to the latest rankings released this week at the ISC High Performance conference in Hamburg, Germany. This marks a gain of 17 systems from the previous list, with particularly strong momentum in new deployments: nearly nine of every 10 systems new to the ranking are built on NVIDIA technologies.
This percentage reflects a deliberate preference for machines built for AI, simulation and science together. The effect is compounding. NVIDIA systems across the TOP500 now deliver more than 2 times the AI training throughput and nearly 3 times the AI inference throughput of every other platform combined.
GPU and networking adoption have each reached new highs. NVIDIA GPUs are accelerating a record 238 systems while NVIDIA networking is connecting a record 376 systems. The vast majority run on NVIDIA Quantum InfiniBand, which serves as the backbone of large-scale AI and high-performance computing, with the rest using Ethernet. Underlying this trend is a larger shift: accelerated computing is becoming the foundation for systems handling the world's most demanding work across AI and science.
NVIDIA's reach now spans the full system through GPU, networking and increasingly CPU. NVIDIA Grace CPU adoption has reached 26 systems, up from 18 in the previous list, with nearly 2.5 million Grace CPUs shipped. Grace-based machines occupy top positions in both rankings, with JUPITER at number 5 and Alps at number 10 on the TOP500, and KAIROS at number 1 on the Green500. Each pairs an NVIDIA GPU with the Grace CPU in a single NVIDIA Grace Hopper Superchip, enabling the two to share memory with minimal overhead in a design built for the memory-intensive demands of modern AI.
NVIDIA recently announced the Vera CPU, which builds on Grace's success by taking CPU performance and energy efficiency to new levels for the most demanding AI workloads in modern data centers, where agents move from answering basic questions to taking actions, running code, using tools and evaluating results.
NVIDIA swept the Green500 ranking of the most energy-efficient supercomputers. The top eight all run on NVIDIA GPUs and nine of the top 10 use NVIDIA technologies. KAIROS, an NVIDIA Grace Hopper system at France's University of Toulouse, leads the list at 73.3 gigaflops per watt, with Grace Hopper systems taking the top four spots across France, Germany and the United Kingdom.
A record 35 NVIDIA AI HPC supercomputers are in development across Europe, equipping more than 3 million researchers with next-generation infrastructure for continental AI, accelerated science and industrial innovation. Among these is JUPITER, Europe's fastest supercomputer and its first to reach exascale, located at the Jülich Supercomputing Centre in Germany. JUPITER is mapping the human brain at cellular scale, simulating Earth's climate and advancing the AI behind next-generation 6G networks.
The newest arrivals to the list run on the NVIDIA Blackwell architecture, with B200 and GB200 systems entering the rankings across Asia, Europe and the United States, with the first GB200 systems debuting in Japan. The buildout is global, ranging from a new AI factory in South Africa to national AI systems in Saudi Arabia, Singapore and Vietnam. The world's AI buildout is running on NVIDIA.