Hangang supercomputer enters trial service in South Korea

Hangang is designed for AI-based scientific research alongside large-scale computation and simulation.[1] Its intended uses include discovering drug candidates with AI and developing new materials.[1] KISTI lists biotechnology, materials and chemistry, earth sciences, and semiconductor research among its priority areas.[4] That range positions the system as shared infrastructure for research across several scientific and technical disciplines, rather than a resource intended for a single field. The announced uses describe the system’s purpose; the evidence does not report completed research outcomes from Hangang.

KISTI began recruiting researchers to use Hangang in September, and the system has been in a trial run since then.[3] KISTI is currently running a trial service, while official service is scheduled to begin on the 14th of next month.[1][2] The government plans to allocate around 15% of Hangang’s total computing resources to industrial innovation and around 55% to general R&D by universities and other academic, industrial and research institutions.[1] It plans to use about 30% of the resources remaining after the industrial-innovation allocation for national strategic R&D, including K-Moonshot.[1]

The government plans to provide small and medium-sized enterprises and startups with technical support, consulting and computing resources.[1] Hangang is intended to help smaller firms carry out complex design work and simulations without purchasing their own expensive computing equipment.[1] For companies considering research or engineering workloads, those plans make access arrangements and technical assistance important details to check as the service moves from trial to official operation. The evidence describes planned support and intended use, but does not specify application procedures, eligibility rules or the amount of computing time available to individual firms.

Hangang was funded with 370 billion won in government investment.[1] According to KISTI and the Ministry of Science and ICT, it has 8,496 GPUs and 9,936 CPUs.[2] Its theoretical peak performance is 614 petaflops, and it has 205 petabytes of storage.[3] Compared with its predecessor Nurion, Hangang offers 24 times the theoretical computing performance, 10 times the storage capacity and four times the network performance.[3] The ministry expects Hangang to place in or near the Top500’s top 10; the ranking is due to be released at the SC26 conference in Chicago this month.[3]