South Korea's Ministry of Science and ICT will start on March 16 a call for cloud service providers (CSP) to supply a graphics processing unit (GPU) leasing programme, and a separate call to recruit industry users for more than 2,000 additional government GPUs that can be used, to strengthen domestic industry-academia-research artificial intelligence (AI) research and development capabilities.

The government is pushing to build an AI expressway as a key national agenda item. It drew up a 2025 supplementary budget plan to secure 13,000 advanced GPUs. This year, it is also pursuing advanced GPU procurement worth 2.08 trillion won.

To bridge gaps stemming from the time required to secure advanced GPUs and to help invigorate the private GPU services market, the ministry is promoting a GPU leasing programme that supports industry, academia and research with GPU resources held by private CSPs.

The GPU leasing programme call comprises 2 projects: a high-performance computing support project and an AI research computing support project. The target is CSPs that can provide GPU services in South Korea.

The call for the high-performance computing support project will run until next month on April 16. The ministry will select a CSP to supply at least 1,060 GPUs to industry. It plans to choose an operator with flexible GPU allocation and resource operation and management capabilities to efficiently provide small-scale AI computing resources to industry, including small and medium-sized companies and startups, with services offered as up to 2 servers or in unit quantities of GPUs.

For this, it will comprehensively evaluate AI computing resource ownership and deployment plans and security capabilities, among other factors. The selected supplier must provide GPU-based high-performance computing resources and development environments optimised for AI development.

The call for the AI research computing support project will run until next month on April 6. The ministry will select a CSP to supply at least 960 GPUs to AI researchers in academia and research institutions.

The ministry plans to select an operator that can provide large-scale computing resources and an optimal research and development environment to support development of core AI technologies. It will comprehensively evaluate the adequacy of plans to support research computing resources, the status of computing resources held, and the level of research support environment to be provided. The selected supplier must provide allocation of computing resources and a research environment optimised for research and development related to development of extra-large AI, such as large language models.

The ministry will also run in parallel a second 2026 user call for an advanced GPU utilisation support programme to allocate more than 2,000 government GPUs available for additional use among GPUs secured through last year's supplementary budget. This user call will run until March 30, focusing on short-term demand from industry, including small and medium-sized companies and startups. It plans to swiftly proceed with user selection procedures, including a suitability review and evaluation, and supply the GPUs from early April.

Choi Dong-won (최동원), director general for AI Infrastructure Policy at the ministry, said, "As AI computing infrastructure capabilities directly translate into national AI competitiveness, we will concentrate government capabilities to expand domestic AI computing infrastructure and strengthen competitiveness."