South Korea plans to launch a 4.7 trillion won ($3.5 billion) government-backed programme to develop a frontier artificial intelligence model from March 2027, stepping up Seoul’s efforts to establish a stronger position in the global race for increasingly capable AI systems.
The Ministry of Science and ICT said it plans to select a lead developer through a competitive tender after parliament approves the 2027 budget in December. The winning team could be selected as early as February, allowing development work to begin the following month.
The programme will combine government equity investment with private-sector funding and concentrate computing chips, data and AI talent on a single project designed to produce a model with leading performance.
Register for the next Tekedia Mini-MBA.
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
The scale of the programme highlights the growing importance governments are placing on domestic AI capabilities. As frontier models require large amounts of computing power, high-quality data and specialized researchers, countries are seeking ways to pool resources rather than leaving development entirely to individual companies.
South Korea already has a substantial technology industry spanning semiconductors, consumer electronics and telecommunications. The new programme seeks to use those existing strengths to build a domestic AI capability that can compete at the frontier rather than relying entirely on models developed by U.S. and Chinese companies.
The proposed programme will bring together several resources that are increasingly difficult for individual AI developers to secure independently.
The government plans to combine public capital with private funding while concentrating computing chips, data and talent around the selected developer. That approach is intended to give the project the scale needed to compete with companies that have access to enormous computing clusters and substantial private investment.
The ministry said participation would be open to a broad range of structures, including individual companies, consortiums and special-purpose vehicles.
Startups, established technology companies, universities and researchers will therefore be able to take leading roles in the project.
The structure could also allow South Korea to spread the financial and technological burden between the state and private sector. Frontier AI development can require sustained spending on computing infrastructure, model training, data acquisition and highly specialized personnel, making it difficult for smaller companies to compete independently.
By concentrating those resources around a selected project, Seoul is effectively attempting to create a national-scale platform capable of competing with much larger international AI operations.
The programme’s competitive tender will also give the government an opportunity to select the team it believes has the strongest combination of technical expertise, computing resources, data and commercial potential.
Separate from South Korea’s Existing AI Programme
The new frontier initiative will operate separately from South Korea’s existing home-grown foundation-model programme. That programme will continue, with two final teams expected to be selected during a third-stage review in February 2027.
The decision to maintain both programmes indicates that the government is pursuing two different objectives. The existing initiative is focused on developing domestic foundation models, while the new 4.7 trillion won project is intended to concentrate resources on a model capable of competing at the global frontier.
Building a competitive AI model is not simply a matter of increasing the size of a system. Developers need access to advanced processors, large and high-quality datasets, efficient model architectures, and researchers capable of improving training and inference techniques.
South Korea’s semiconductor industry could provide an important advantage in that effort.
The country is home to Samsung Electronics and SK Hynix, two of the world’s largest memory-chip manufacturers. The government can potentially leverage the country’s broader semiconductor ecosystem as it attempts to build domestic AI computing capacity.
But that does not guarantee that South Korea will produce a globally competitive frontier model. The leading AI laboratories have accumulated years of experience, vast amounts of training data, and large-scale computing infrastructure. They also operate in markets with access to substantial pools of private capital.
The new programme is therefore an attempt to narrow a gap that has emerged quickly as frontier AI development has become increasingly capital intensive.
A National AI Race is Taking Shape
South Korea’s plans are part of a wider shift in which governments are treating advanced AI capability as an economic and strategic asset.
The United States remains home to many of the leading frontier AI companies, while Chinese developers have rapidly expanded their capabilities and built increasingly competitive models. Other countries are attempting to avoid becoming dependent on those two ecosystems.
For Seoul, the issue has extended beyond developing a chatbot or consumer AI product. A competitive domestic model could support applications across manufacturing, software, research, defense and other industries while reducing dependence on foreign AI platforms. It could also create demand for domestic semiconductor, cloud-computing and data-center infrastructure.
That makes the programme potentially relevant to more than South Korea’s AI industry. If it succeeds, the investment could create a broader domestic ecosystem in which AI developers, chipmakers, cloud providers, universities and startups benefit from a common pool of infrastructure and talent.
The government’s decision to allow consortiums and special-purpose vehicles suggests it does not want to restrict the project to a conventional technology company. Universities and research institutions could contribute scientific expertise, while startups could provide model-development capabilities and larger companies could supply computing infrastructure or capital.
The structure also leaves open the possibility that the eventual lead developer could emerge from a collaboration rather than a single established AI company.
The Challenge Will Be Turning Spending Into Capability
The size of the investment demonstrates Seoul’s willingness to commit substantial public resources to the AI race, but money alone does not guarantee a frontier model. The leading systems are the product of continuous iteration, enormous computing resources, and highly specialized research teams. Frontier AI developers must also manage the rising costs of inference as models become widely deployed.
South Korea will therefore need to translate its spending into sustained technical progress. Access to advanced chips, particularly as global demand for AI accelerators remains intense, will be critical. So will attracting and retaining researchers capable of competing with laboratories in Silicon Valley and China.
The programme’s public-private structure may help address some of those challenges by spreading costs and giving private companies an incentive to commercialize the technology. At the same time, concentrating resources behind a single national project creates its own risk. If the selected team fails to achieve the expected performance, a large amount of capital and computing capacity could be tied to an unsuccessful approach.
Keeping the separate foundation-model programme alive provides some diversification by allowing multiple teams to continue developing alternative systems.
For now, South Korea is betting that a concentrated national effort can give it a place in a global market dominated by companies with far larger technology budgets. The 4.7 trillion won programme marks a shift from supporting domestic AI development broadly to attempting to build a system capable of competing at the frontier.
Industry analysts believe that if Seoul can successfully combine its semiconductor strength, research base, private-sector capital and government support, the initiative could become an important test of whether a country outside the leading U.S. and Chinese AI ecosystems can build a globally competitive frontier model.
But the first major test will come before development even begins: selecting a lead team capable of turning the government’s unprecedented concentration of chips, data, capital, and talent into a model that can compete with the world’s most advanced AI systems.