South Korea Plans $3.5 Billion Frontier AI Model Development Program

South Korea has announced a 4.7 trillion won ($3.5 billion) initiative to develop a frontier artificial intelligence model, with the program scheduled to launch in March 2027. The Ministry of Science and ICT stated that the project aims to bolster the nation's competitiveness in the global AI race by concentrating computing resources, data, and talent. This new initiative will operate independently of the government's existing foundation model program, which is currently in its final stages of developer selection.
Program Structure and Implementation
The government plans to utilize a mix of state equity investment and private funding to support the development of the model. A lead developer for the project will be selected through a competitive tender process, with the selection expected to occur between February and March 2027, following parliamentary approval of the 2027 budget in December. The ministry indicated that participation will be open to a wide range of entities, including individual companies, consortiums, and special-purpose vehicles, with an emphasis on enabling startups, universities, and researchers to take leading roles.
Strategic Objectives
The initiative is part of Seoul's broader strategy to build cutting-edge artificial intelligence capabilities. By creating a dedicated frontier model project, the government seeks to ensure that South Korea remains at the forefront of AI innovation. The existing foundation model program, which will select two final teams in February 2027, will continue alongside this new effort. The government is also considering the use of a special-purpose company to manage the project, reflecting the scale and strategic importance of the investment.
What the Korean funding announcement does, and does not, fund
The ₩4.7 trillion figure is a proposed programme envelope, not evidence that one model or developer has already received the money. The sequence matters: parliamentary budget approval, tender rules, selection of a lead organisation and actual deployment are separate milestones. South Korea is also pursuing a different existing foundation-model effort, so treating the two programmes as one would inflate the scope of this announcement. A useful measure of success will be whether the new project makes compute and high-quality datasets available to researchers and domestic companies, rather than simply producing another nationally branded model.
