Nvidia Tops 400 Systems On TOP500 Supercomputer List
Nvidia says its technology now powers more than 400 of the world’s 500 fastest supercomputers, with Grace CPUs, GPUs and networking expanding across AI and science systems.
Nvidia Expands Its TOP500 Footprint
Nvidia says its technology now powers more than 400 of the world’s 500 fastest supercomputers, equal to 81% of the latest TOP500 ranking released at the ISC High Performance conference in Hamburg.
The company also says nearly nine of every 10 systems new to the list use Nvidia technologies.
The figures give the AI infrastructure market a hardware-level signal.
Nvidia is not only selling GPUs into cloud AI clusters; its processors and networking are also taking a larger share of national and scientific supercomputing systems.
Nvidia reported that TOP500 systems using its technology produce more than 2x the AI training throughput and nearly 3x the AI inference throughput of all other platforms combined.
Those comparisons are vendor-reported, but they show how Nvidia is framing supercomputing as an AI training and inference market, not only a traditional simulation market.
The announcement also links supercomputing to national AI capacity.
Nvidia pointed to deployments in Europe, Japan, South Africa, Saudi Arabia, Singapore and Vietnam, which moves the story beyond one benchmark list and into the location of future AI factories and research systems.
GPUs And Networking Set The System Shape
GPU and networking adoption both reached new highs in the ranking.
Nvidia GPUs accelerate 238 systems, while Nvidia networking connects 376 TOP500 systems.
Most of those networked systems use Nvidia Quantum InfiniBand, with the rest on Ethernet.
The networking count matters because large AI and high-performance computing systems depend on how quickly processors, memory and storage can communicate.
The source-backed numbers show that Nvidia's position reaches beyond chips into the fabric that ties clusters together.
Nvidia's CPU story is also growing.
Grace CPU adoption reached 26 systems, up eight from the previous list, and the company said nearly 2.5 million Grace CPUs have shipped.
Grace-based machines include JUPITER at No. 5 and Alps at No. 10 on the TOP500, while KAIROS ranks No. 1 on the Green500 efficiency list.
That CPU adoption is still small beside Nvidia's GPU and networking counts, but it is strategically important.
Grace gives Nvidia a path to sell more of the full system stack into AI and high-performance computing sites rather than leaving the host processor to another vendor.
Efficiency And Europe Shape The Buildout
The Green500 ranking gives the power-efficiency side of the AI infrastructure story.
The first eight systems in that list use Nvidia GPUs, and nine among the first 10 systems use Nvidia technologies.
KAIROS, a Grace Hopper system at France’s University of Toulouse, leads the Green500 at 73.3 gigaflops per watt.
Europe is a major deployment region in the announcement.
Nvidia said a record 35 Nvidia AI HPC supercomputers are in development across Europe, with infrastructure intended to support more than 3 million researchers.
JUPITER at the Jülich Supercomputing Centre in Germany is described as Europe’s fastest supercomputer and its first to reach exascale.
New Blackwell systems are also entering the ranking.
Nvidia said B200 and GB200 systems appeared across Asia, Europe and the U.S., with the first GB200 systems debuting in Japan.
The company also pointed to a new AI factory in South Africa and national AI systems in Saudi Arabia, Singapore and Vietnam.
Nvidia did not publish customer-level economics for the new systems, leaving the operating evidence at ranking share, throughput claims, deployment geography and efficiency scores.
