The government will push a total 2.08 trillion won project to secure advanced graphics processing units (GPU) together with private cloud corporations. The move aims to build large-scale computing infrastructure needed for an independent AI foundation model project and for national-level AI projects.
The Ministry of Science and ICT said on the 8th that it selected three cloud service corporations—Naver Cloud, Samsung SDS, and Elice—to participate in the "advanced GPU acquisition, build-out, and operations support" project.
The project is considered a core foundation of the government's so-called "AI highway" build-out. The Ministry of Science and ICT plans to secure a total of 9,704 advanced GPUs in cooperation with the selected corporations and use them to support demand for AI research and development and service development in the private and public institutional sectors.
The GPUs to be introduced will consist of 2,016 units of the next-generation product "Verarubin" and 7,688 units of the B300. By corporation, Naver Cloud will deploy 1,008 Verarubin units and 3,112 B300 units; Samsung SDS will deploy 1,008 Verarubin units and 2,016 B300 units; and Elice will deploy 2,560 B300 units.
The government initially aimed to secure 15,000 units based on the B200, but by combining Verarubin and the B300, it will be able to achieve performance equivalent to about 19,000 B200 units. The government said that although the total unit count has decreased, in performance terms it is about 30% higher than the initial target.
Of the secured volume, 2,016 Verarubin units and 4,360 B300 units will be allocated for government use. These GPUs will be used for the independent AI foundation model project, national AI projects, and the development and advancement of AI models and services across industry, academia, and research.
The remaining 3,328 B300 units will be operated independently by the cloud service corporations participating in the project. Reflecting their nature as operating costs, each corporation plans to use them to provide cloud-based GPU services and strengthen the competitiveness of its own AI models and services.
The Ministry of Science and ICT plans to place purchase orders for GPUs within this month. After that, cloud corporations that complete delivery and build-out will sequentially launch B300-based services within the year. Verarubin will be rolled out in phases starting in the first half of 2027, in line with the product release schedule.
Bae Kyung-hoon, Deputy Prime Minister and Minister of the Ministry of Science and ICT, said, "Advanced GPUs such as the Verarubin we are securing this time will be a key driver that elevates the speed of AI research and development and technical capabilities," adding, "We will secure world-class AI infrastructure capabilities to support AI innovation and growth by domestic corporations and research institutes."