Nvidia (NASDAQ: NVDA) announced at the ISC High Performance 2026 conference in Hamburg, Germany, that a record 35 AI supercomputers built on its infrastructure are in development across 23 European countries, serving more than 3 million researchers.

The systems are based on Nvidia's Blackwell and Hopper platforms, with 800 AI exaflops of capacity deployed or announced since last year. Projects include the Barcelona Supercomputing Center's MareNostrum5 AI upgrade, BavariaAI's Blue Swan, Italy's IT4LIA, Germany's HammerHAI and Sweden's Mimer EuroHPC AI Factory.

Key system specifications include IT4LIA, which will deploy over 8,000 GPUs delivering 82 exaflops of AI training and 164 exaflops of AI inference performance. The Barcelona Supercomputing Center's AI Factory will use Nvidia GB300 NVL72 systems to deliver up to approximately 20 exaflops of AI training and 33 exaflops of AI inference performance. BavariaAI's Blue Swan will bring 1,000 GPUs delivering up to 11 exaflops of AI training performance.

Siemens Energy is using Nvidia technologies, including Nvidia Omniverse libraries and CUDA-X, to simulate gas turbine burners designed to run on up to 100% hydrogen. The workflow cuts simulation times by up to 77%, according to the announcement.

On quantum computing, researchers at the Jülich Supercomputing Centre, working with an Nvidia team, simulated a universal 50-qubit quantum computer using the JUPITER supercomputer, which is powered by Nvidia GH200 Grace Hopper Superchips. Nvidia stated this set a world record for full simulation of a universal quantum computer at that qubit count.

CINECA, Fraunhofer and the Barcelona Supercomputing Center are among institutions adopting Nvidia's CUDA-Q platform to integrate quantum processors with GPU supercomputing infrastructure.

The announcements were made based on a press release issued by Nvidia at the ISC High Performance 2026 conference on June 22, 2026.