Version 2

NVIDIA's ISC High Performance 2026 presence in Hamburg (June 22–26) has produced announcements across five fronts: production science results from JUPITER, Europe's exascale system at Jülich [1]; a three-system Los Alamos National Laboratory procurement (Mission, Vision, Veritas) targeting 2027 [3]; NAIRR pilot outcomes across 700+ U.S. research projects [4]; new scientific software including cuPhoton and CERN's DAQIRI [5]; and, on June 23, TOP500/Green500 data claiming 81% of the fastest 500 systems run NVIDIA technology [2] alongside a BioNeMo Agent Toolkit said to compress drug discovery from 12 years to 12 months [6]. All announcements originate from NVIDIA's own blog.

NVIDIA is using ISC 2026 to present exascale compute as delivering production science today, not future capability, while TOP500 figures reinforce a dominant position across commercial and national-lab procurement. The competitive question is whether AMD's recent supercomputer wins represent a meaningful alternative or a marginal share of a market NVIDIA now claims to dominate by volume.

NVIDIA opened ISC High Performance 2026 on June 22 with four coordinated blog posts covering European exascale science, U.S. national-lab procurement, federal research access, and scientific software tools. On June 23 the company added TOP500/Green500 results and a new biomedical AI platform, extending the ISC narrative into a second day before the conference closes June 26. All announcements are self-published by NVIDIA.

The centerpiece of NVIDIA's European story is JUPITER, operated by Jülich Supercomputing Centre and built on Grace Hopper Superchips. Three headline results are cited: CytoNet, a brain microarchitecture foundation model trained on 6.5 petabytes of data from 21 post-mortem brains, completed in under five days on 4,096 Superchips; the ICON climate model won the Gordon Bell Prize for Climate Modelling by simulating a fully coupled Earth system at 1-kilometer global resolution; and Jülich set a quantum simulation record by fully simulating a 50-qubit universal quantum computer, exploiting coherent CPU-GPU memory in GH200 chips [1]. NVIDIA also reported a record 35 AI HPC supercomputers in development across Europe and NVIDIA networking connecting 376 of the 500 fastest systems [2].

At Los Alamos, NVIDIA and HPE announced three systems — Mission, Vision, and Veritas — using Vera CPUs, Rubin GPUs, and Quantum-X800 InfiniBand, with 2027 operational targets [3]. NVIDIA claims the Vera CPU delivers 7x better performance on LANL's URSA agentic AI workloads compared to the x86-based Crossroads system and 3x better on Monte Carlo simulations with 4x more memory per core. The NAIRR pilot has supported over 700 U.S. research projects, producing the Walrus foundation model for fluid simulations, the MIST molecular benchmark suite, and Boston University's BEACON infectious disease pipeline [4]. Software announcements include cuPhoton (14,900x faster astronomical image loading for Rubin Observatory's LSST on GB200 NVL72), CERN's A-GHOST DAQIRI system for real-time AI on previously discarded collision data, and the BioNeMo Agent Toolkit [5][6].

No independent benchmark verification appears in the current item set. Tom's Hardware noted that the LANL procurement follows recent AMD wins in the supercomputer market [7], a competitive dimension absent from NVIDIA's own framing. The TOP500 figures — 81% of systems, 90% of new entrants, 2x AI training throughput over all other platforms combined — come from NVIDIA's interpretation of the TOP500 list, not from the TOP500 organization itself [2].