South Korea launches 2.08-trillion-won AI computing drive, invites global GPU providers.
South Korea’s government announced a major push to expand AI computing resources, unveiling the 2026 AI Computing Resource Utilization Foundation Project. The Ministry of Science and ICT (MSIT) and the Information Technology Promotion Agency (NIPA) said the project has a total budget of about 2.08 trillion won to build advanced GPU infrastructure and a unified operating environment, with the aim of enabling domestic universities, industry, and research to accelerate AI development.
The plan follows last year’s supplementary budget, which secured about 1.4 trillion won to acquire 13,000 GPUs. Officials say the new program will again leverage both government funds and private sector technology to mount GPU capacity and provide stable computing services to Korea’s AI ecosystem.
A key change this year is the broadened eligibility for participation. While last year’s call limited applicants to companies with principal operations in Korea, this year the requirement centers on whether a bidder can provide and operate GPU-as-a-Service (GPUaaS) domestically. That effectively opens the field to foreign firms, including global cloud providers such as AWS, alongside domestic incumbents, intensifying competition for the business.
Bidders will be evaluated largely on cost effectiveness and the ability to scale. The government requires at least one cluster comprising 256 servers, totaling 2,048 GPUs. Proposals that plan larger clusters are expected to receive higher scores. The equipment should be centered on the latest GPUs, including Blackwell-class models, with additional points for proposals that incorporate next-generation architectures such as Vera Rubin.
The selected contractor must purchase the GPUs and deploy the servers within the year and will be responsible for five years of operation and maintenance through 2031. All GPUs, servers, and related equipment funded through the project will be owned by NIPA, the implementing agency.
Security credentials are part of the bid criteria. Proposers must demonstrate current or planned information security management system (ISMS) and Cloud Security Assurance Program (CSAP) certifications, reflecting Korea’s emphasis on secure AI infrastructure.
Applications open next month, with a submission deadline on the 13th through NIPA’s project-management system (NXT). A two-stage process will include an evaluation in March, on-site data center inspections in April, and the final award announcement planned for May, followed by contract negotiations with the winning party.
Why this matters beyond Korea: the plan signals a substantial, government-led commitment to AI infrastructure that could shape regional technology dominance, influence GPU demand and pricing, and affect global cloud competition. For U.S. readers, the outcome could affect partnerships and competition between Korean and American cloud providers in Asia, the resilience and localization of AI supply chains, and the pace at which Korean universities and firms can collaborate with global tech players on advanced AI research and deployment. The policy also highlights ongoing debates about national strategies for critical AI infrastructure, data security, and sovereign control of key computational resources.