NVIDIA's Vera Rubin NVL4 Platform Expected in Late 2026

NVIDIA announced that systems based on the Vera Rubin NVL4 platform will be available through global manufacturers in late 2026. The platform combines four Rubin GPUs and two Vera CPUs, using sixth-generation NVLink and NVLink-C2C interconnect technology, optimized for density and efficiency in modern supercomputing.

Compared to the previous Grace Hopper, Vera Rubin NVL4 delivers significant performance gains: 4x faster scientific simulation, 6x faster AI training, and 8x faster AI inference. It's compatible with liquid-cooled NVIDIA MGX modular servers, promising to boost high-performance computing and AI research.