NVIDIA AI Updates: June 23, 2026

1. NVIDIA Vera CPU Targets Agentic Scientific AI at Los Alamos

NVIDIA announced that its Vera CPU will enable agentic scientific AI workloads at Los Alamos National Laboratory. The company framed the CPU as a foundation for autonomous, AI-driven scientific research at the national lab, pairing host compute with its accelerated platform. Source

2. JUPITER Exascale System Runs on NVIDIA Grace Hopper

NVIDIA detailed how JUPITER, Europe’s first exascale supercomputer at Germany’s Forschungszentrum Julich, runs on NVIDIA Grace Hopper Superchips and Quantum-X800 InfiniBand networking. The system supports work across brain mapping at cellular resolution, kilometer-scale global climate simulation, AI development for 6G networks, and quantum computing simulation. The announcement coincided with the ISC High Performance conference. Source

3. NVIDIA Ships New CUDA AI Software for Scientific Discovery

NVIDIA introduced new AI software, built on CUDA, aimed at accelerating scientific work spanning materials simulation and experimental astronomy. The company presented the tools as broadening the reach of AI-accelerated computing across scientific domains. Source

4. NVIDIA Details CCCL Runtime for Modern C++ on CUDA

NVIDIA published technical details on the CCCL Runtime, a modern C++ runtime for CUDA. The post described how the runtime aims to improve the developer experience for writing CUDA C++ applications. Source

5. NVIDIA Halos Brings Full-Stack Functional Safety to Robotics

NVIDIA detailed Halos for Robotics, a full-stack functional safety system for physical AI applications. The post explained how the system is designed to address safety requirements across the robotics software and hardware stack. Source