- Equipped with 8,496 high-performance GPUs to support AI-powered science and technology (AI+S&T) research
- HPE selected as final contractor after six rounds of bidding, with a contract valued at KRW 382.5 billion
- Korea to build a supercomputer expected to rank among the world’s top 10, leading the era of AI-driven innovation
- Expected to play a pivotal role in high-precision scientific computing and AI-powered R&D across Korea’s public sector
On May 12, the Ministry of Science and ICT (Minister: Yoo Sang-im, hereinafter referred to as "MSIT") announced that a contract for the construction of Korea’s sixth national supercomputer was finalized. The contract, valued at KRW 382.5 billion, including KRW 78 billion allocated for five years of maintenance, was signed between the Korea Institute of Science and Technology Information (KISTI) and Hewlett Packard Enterprise (HPE).
According to the Public Procurement Service (Administrator: Lim Ki-keun), two global supercomputer manufacturers participated in the bidding process. Following a comprehensive evaluation of technical specifications and performance, HPE was selected as the winning bidder, having offered the most favorable terms for the nation.
HPE is a globally recognized leader in large-scale supercomputing, with proprietary interconnect technology designed specifically for high-performance computing systems and an extensive track record in deploying massive infrastructures. As of November 2024, this expertise was reflected in the Top500 global supercomputer rankings, where HPE was the manufacturer behind three of the top five systems—1st-ranked El Capitan (U.S., 2.7 exaFLOPS), 2nd-ranked Frontier (U.S., 2.0 exaFLOPS), and 5th-ranked HPC6 (Italy, 607 petaFLOPS)—among a total of 106 HPE-powered systems on the list.
※ 1 petaFLOPS (PF): A computing performance level capable of executing one quadrillion (1,000 trillion) operations per second
Scheduled for completion in the first half of 2026, Supercomputer No. 6 will serve as a national flagship high-performance computing (HPC) platform. It is expected to support a wide range of applications—including large-scale, high-precision scientific and engineering simulations, as well as frontier research in artificial intelligence (AI)—in response to the rapidly evolving global R&D and industrial landscape.
The new system will be equipped with 8,496 of NVIDIA’s latest GPUs, including the GH200 model, and will deliver a peak performance of 600 petaFLOPS, along with 205 petabytes (PB) of storage and ultra-high-speed network connectivity exceeding 400 Gbps. This configuration is anticipated to place the system among the top 10 supercomputers worldwide, with inclusion in the Top500 list.
※ GH200: A next-generation chip integrating the Hopper (H100) GPU architecture with an ARM-based CPU named Grace, designed to maximize data processing efficiency as a unified CPU+GPU accelerator.
Leveraging its advanced architecture, the system will be optimized for AI training and inference, large-scale scientific and engineering simulations, and the development of AI foundation models—ultimately contributing to a substantial increase in national R&D productivity.
This initiative aligns with MSIT’s March 12 announcement of its “AI+S&T Promotion Strategy,” which seeks to drive a paradigm shift in R&D by accelerating AI adoption across Korea’s scientific and technological sectors, many of which remain at an early stage of AI integration.
The strategy calls for expanding the use of specialized AI models across eight key domains—including next-generation materials, innovative pharmaceuticals, and ultra-fine semiconductors—to transform traditional R&D practices and position AI as a key driver of scientific discovery.
※ The eight AI+S&T fields for specialized model development: next-generation materials, innovative pharmaceuticals, ultra-fine semiconductors, space exploration, quantum computing, next-generation batteries, eco-friendly technologies, and future energy.
Currently, the public sector in Korea faces a severe shortage of shared high-performance GPU infrastructure. This forces researchers to independently acquire costly GPUs or rely on overseas cloud services, which not only result in high expenditures but also raise concerns over potential data security breaches.
To address this issue, MSIT will ensure the rapid deployment of Supercomputer No. 6 to support demands for extreme-scale scientific computing, big data analytics, and AI+S&T research—particularly for medium to large-scale AI model development.*
“Medium to large-scale AI development” refers to projects that utilize approximately 50 to 200 GPUs and are characterized by a scope that allows for completion of AI training or scientific computing tasks within a three-month period.
Recognizing that the backlog of GPU demand across research and industrial sectors cannot be addressed immediately, MSIT is working with fiscal authorities to establish customized public infrastructure for AI+S&T.
This initiative will enable domain-specific GPU access for specialized research institutions, including government-funded entities.
* Examples of domain-specific GPU applications include:
-(Materials) AI-driven multi-modal technologies for material innovation
-(Pharmaceuticals) AI-based compound synthesis and drug property prediction
-(Energy) AI-assisted optimization of fuel cell manufacturing processes
-(Smart Manufacturing) Autonomous semiconductor component manufacturing powered by AI
Kim Seong-soo, Director-General of R&D Policy at MSIT, remarked,
“The timely signing of the contract for Supercomputer No. 6 comes in response to the surging demand for GPUs in Korea. We expect it will empower researchers and industry leaders to overcome persistent challenges and drive groundbreaking scientific innovation.”
For further information, please contact the Public Relations Division (Phone: +82-44-202-4034, E-mail: msitmedia@korea.kr) of the Ministry of Science and ICT.
※ Please refer to the attached PDF.