ISC High Performance 2026: China's Surprise No. 1, NVIDIA's Record Dominance, and Europe's 35-System AI Supercomputer Expansion

Experts noted the result demonstrates China's ambition for self-sufficiency in computing systems rather than AI competitiveness, since GPU-accelerated systems remain better suited for AI workloads . LineShine also topped the HPCG (High-Performance Conjugate Gradient) benchmark with 22.00 HPCG-Petaflop/s and the HPL-MxP benchmark with 7.92 exaflops .

NVIDIA's Dominance: TOP500, Green500, and Grace CPU Adoption

Despite losing the No. 1 spot, NVIDIA's overall presence on the TOP500 grew to 81% — over 400 of the 500 systems use NVIDIA technology . Key metrics from NVIDIA's official blog :

• 90% of systems new to the TOP500 run on NVIDIA technology.
• 26 systems on the list now use the NVIDIA Grace CPU, up 8 from the previous list.
• 238 systems use NVIDIA GPUs; 376 systems use NVIDIA networking (mostly Quantum InfiniBand).
• Grace CPU shipments total nearly 2.5 million units .
• The top 8 systems on the Green500 energy-efficiency ranking run on NVIDIA GPUs, and 9 of the top 10 use NVIDIA technologies .
• The No. 1 system on the Green500 is KAIROS, a Bull Sequana XH3000 system using a single NVIDIA Grace Hopper GH200 Superchip, achieving 73.3 gigaflops per watt .

Europe's Record 35 New AI Supercomputers

On June 22, 2026, NVIDIA announced the largest-ever European AI supercomputing expansion: 35 new AI/HPC systems across 23 countries, built on Blackwell and Hopper architectures . Key details:

• The infrastructure will serve over 3 million researchers .
• Systems will support climate science, healthcare, clean energy, decarbonization, and quantum computing .
• The total AI computing capacity across these systems is described as delivering 800 AI exaflops, accelerating work for researchers in generative AI, climate modeling, medicine, sustainable agriculture, and energy systems .
• NVIDIA AI infrastructure now powers over 90% of Europe's AI factory buildout .

Key Named Installations

• MareNostrum 5 AI (Barcelona Supercomputing Center) — EuroHPC's first AI-dedicated facility, upgrading with NVIDIA GB300 NVL72 and GB200 NVL4 systems connected via Quantum-X800 InfiniBand. Post-upgrade AI training performance is estimated at ~20 EFLOPS, AI inference at ~33 EFLOPS .
• Blue Swan (BavariaAI) — deploying 1,000 GPUs via NVIDIA GB200 NVL4 systems and Quantum-2 InfiniBand networking at FAU Erlangen and Leibniz Supercomputing Centre (LRZ). AI training performance up to 11 EFLOPS, inference up to 22 EFLOPS, supporting Bavaria's foundation model initiative .
• IT4LIA (Czech Republic) — will be equipped with over 8,000 GPUs using NVIDIA GB200 NVL4 systems, Quantum-X800 InfiniBand networking, and AI Enterprise software, serving as an AI Factory for the region .
• NAISS's Mimer EuroHPC AI Factory (Sweden) — another system presented at the conference's Mimer AI Factory session .
• HLRS's HammerHAI (Germany) — also part of the 35-system deployment .

Vera Rubin Platform: Rack-Scale Supercomputing for Science

NVIDIA unveiled the Vera Rubin platform, a rack-scale supercomputing architecture designed to converge high-precision simulation, AI, and data analytics. It delivers approximately 7 exaflops of AI performance and 5 petaflops of native double-precision (FP64) performance . The platform is intended for the next generation of TOP500-class systems, combining FPGA-based double-precision engines with NVIDIA's full-stack CUDA-X and AI platform capabilities .

According to NVIDIA CEO Jensen Huang, the platform is designed to function as a new scientific instrument capable of accelerating research in climate modeling, drug discovery, and energy systems .

Broader Implications for Applied Research

NVIDIA and EuroHPC explicitly framed the 35-system expansion around decarbonization, clean energy, and quantum computing research . The quantum-GPU computing track was a featured theme at the conference, with European centers including CINECA, Fraunhofer FOKUS, BSC, and Jülich expanding quantum-GPU infrastructure . The systems are designed to accelerate climate modeling, healthcare, biotechnology, sustainable agriculture, and government AI services .