Hangang, South Korea’s sixth national supercomputer, was commissioned on 2 November with 8,496 GPUs including Nvidia’s GH200, a theoretical peak 24 times that of the 2018 Nurion it replaces, with full service due to start on 14 December.

Hangang represents a substantial upgrade in South Korea's computational infrastructure, primarily through its hardware composition. The system includes 8,496 GPUs, among them Nvidia's GH200 integrated CPU-GPU accelerator chips. The Ministry of Science and ICT reports a theoretical computing performance 24 times that of Nurion, alongside 10 times the storage and four times the network performance. These are theoretical comparisons, not measured results on specific research workloads.

This hardware configuration directly expands the nation's research capabilities. While Nurion supported CPU-based large-scale computational science and simulation, Hangang is specifically designed to add AI-based scientific research using its newer GPUs. The Ministry stated that Hangang will serve as a core base for supporting the combination of AI with science and technology, aiming to facilitate new scientific discoveries beyond what was primarily possible with the older system's architecture.

The new supercomputer is intended for national strategic research and development, including projects like K-Moonshot, which aims to double research productivity by 2030. It will also back exploratory research at universities and government-funded institutes, alongside technology development at companies. While commissioned and running a trial service, its full practical impact on these diverse research areas will emerge once official service commences on December 14.