State of AI Report Compute Index - sources list
H100 chart
- MareNostrum 5 (ES), 1,120 accelerated nodes with 4 Hopper GPUs each, therefore 4,480 H100s, link
- Ruhr University Bochum, 56 H100s link
- RWTH Aachen University: 52*4 H100s, therefore 208 H100, link
- Jean Zay (FR): upgraded their A100s, so now they have 1,456 H100 and 416 A100 and 1832 V100, http://www.idris.fr/annonces/idris-extension-jean-zay-h100.html
- CSCS Alps (CH): https://www.cscs.ch/computers/alps, 10,752 H100
- EPFL (CH): 336 H100 https://www.epfl.ch/research/facilities/scitas/hardware/kuma/
- Hessian: 280 H100 - https://top500.org/system/180340/
- JUPITER: 2k gh200 now, https://www.fz-juelich.de/en/ias/jsc/jupiter/tech, more to come link
- KAUST (Saudi): Shaheen III Phase-2 2,800 H100 https://www.hpc.kaust.edu.sa/sites/default/files/2025-02/2025-02-04-1-Shaheen_III_Intro_Hardware_Environment.pdf
- Israel-1 (ISR): NVIDIA R&D cluster with 2,048 H100 https://www.new-techeurope.com/2024/11/24/nvidias-israel-1-ranks-no-34-in-the-top500-list-of-the-worlds-most-powerful-supercomputers/
- MSFT Eagle (USA): 14,400 H100 https://top500.org/lists/top500/2025/06/
- JUPITER BOOSTER (DE): 24,000 GH200 https://www.fz-juelich.de/en/jupiter
- NVIDIA Eos (USA): 4,608 H100 https://blogs.nvidia.com/blog/eos/#:~:text=Eos%20Supercomputer%20Fuels%20Innovation,to%20autonomous%20machines%20and%20beyond.
A100 chart