South Korea plans to launch a 4.7 trillion won ($3.5 billion) programme from March 2027 to develop a frontier artificial intelligence model capable of competing with the world’s leading systems, according to the Ministry of Science and ICT and Reuters reporting published on October 6, 2026.
The government plans to select a lead developer through a competitive process after the National Assembly approves the 2027 budget in December. Reuters reported that the winning team could be chosen as early as February 2027, with development beginning the following month.
The project is expected to combine government equity investment with private capital while concentrating computing power, data and AI talent on a single high-end development effort. Participation is expected to be open to individual companies, consortiums and special-purpose vehicles, including startups, universities and research teams.
Importantly, the new frontier AI programme is not being presented as a replacement for South Korea’s existing sovereign foundation-model project. The government says the current programme will continue, while the new initiative will focus on developing a model with leading-edge performance.
What Is South Korea’s Frontier AI Model Plan?
The new programme is South Korea’s attempt to move from supporting a broad domestic AI ecosystem toward concentrating resources on a model designed to compete near the frontier of global performance.
Reuters reported that the Ministry of Science and ICT intends to provide 4.7 trillion won in public backing, equivalent to about $3.5 billion at current exchange rates. Local reporting says the structure may involve government equity investment combined with private funding rather than a conventional grant-only programme.
The government’s aim is not simply to produce another Korean-language model. Officials have framed the project as a high-performance AI effort capable of narrowing the gap with the most advanced systems developed in the United States and China.
That distinction matters because frontier AI development requires much larger amounts of computing power, specialist talent, training data and capital than ordinary application-focused AI projects.
When Will the Programme Start?
The government is targeting March 2027 for the start of development.
Before that can happen, the 2027 budget must pass the National Assembly. Reuters reported that the ministry plans to begin the selection process after budget approval in December, with a lead developer potentially chosen by February.
That timetable remains a government plan rather than a completed procurement process. The final schedule could change depending on parliamentary approval, tender rules and the structure ultimately chosen for the project.
The ministry has indicated that it wants to move quickly because model performance is advancing rapidly and delays of even a few months can matter in frontier AI.
How the Government Plans to Fund the Project
South Korea’s approach is expected to mix public and private capital.
Rather than funding the entire project as a direct government subsidy, officials have discussed an equity-based structure in which the state invests alongside private participants.
Local reporting has also discussed the possibility of using a special-purpose company or another joint structure, although the ministry has previously cautioned that some details had not yet been finalised.
That makes the final financial structure one of the most important issues to watch. Equity investment could give the public sector a direct financial interest in the project’s success, but it also raises questions about ownership, governance, intellectual property and how returns would be shared.
Computing Power Will Be Central to the Strategy
Frontier AI models are extremely compute-intensive. Training a top-tier model requires large clusters of advanced accelerators, high-speed networking, storage systems and reliable power infrastructure.
South Korea has already been increasing state support for AI computing. The Ministry of Science and ICT’s 2026 investment plan included additional advanced GPUs, a National AI Computing Center and new work on AI network technologies.
The new frontier programme is expected to concentrate computing resources rather than distribute them evenly across many teams.
Local reporting has described plans to focus access to advanced accelerators, data and talent on the selected developer so that the programme can pursue maximum model performance rather than simply broad participation.
Why South Korea Is Making a Bigger AI Bet Now
The timing reflects growing concern that the global AI race is accelerating faster than earlier policy frameworks anticipated.
Leading U.S. companies continue to invest heavily in larger models and AI infrastructure, while Chinese developers have improved open models and reduced the cost of deploying capable AI systems.
South Korean policymakers increasingly view AI as a national competitiveness issue rather than only a technology-sector priority.
The government has repeatedly said it wants Korea to become one of the world’s leading AI powers and reduce excessive dependence on foreign foundation models.
The Existing Sovereign AI Programme Will Continue
One of the biggest points of confusion around the new initiative has been whether South Korea is abandoning its existing sovereign foundation-model programme.
The science ministry says it is not.
The current programme has supported several domestic AI teams and is designed to build Korean-controlled foundation models that can be used in industry, the public sector and local services.
An official ministry update said LG AI Research, SK Telecom and Upstage advanced from the first phase of the sovereign model competition, while the government planned to select additional participants and continue the programme through further evaluation.
Why the Government Is Splitting AI Policy Into Two Tracks
The emerging strategy separates two goals that are difficult to achieve with a single programme.
The first goal is frontier performance: building an AI model capable of competing with the strongest international systems.
The second goal is practical diffusion: ensuring domestic AI models are deployed across companies, public institutions and sector-specific applications.
Trying to optimise one programme for both objectives can create tension. Frontier development may require concentrating billions of dollars and scarce GPUs on one or a small number of teams, while industrial diffusion benefits from supporting a wider range of companies and specialised models.
South Korea’s new structure is intended to pursue both goals separately but in parallel.
Who Could Participate?
The government says participation will not be limited to a single type of company.
Reuters reported that individual firms, consortiums and special-purpose entities may be eligible, with startups, universities and researchers also expected to play roles.
This is important because frontier AI development increasingly depends on teams that combine model research, semiconductor access, software engineering, data expertise and large-scale infrastructure.
South Korea has several large technology groups capable of contributing resources, but the final lead developer has not been selected and no company should be described as the winner before the official tender process.
South Korea Already Has Several Major Domestic AI Teams
The country’s existing sovereign AI programme has produced several large domestic models.
According to the Ministry of Science and ICT, SK Telecom has been developing a 519-billion-parameter model, LG AI Research has worked on a 236-billion-parameter frontier-level model, and Upstage has developed a 100-billion-parameter foundation model.
NAVER Cloud and NC AI have also participated in the programme with multimodal and industry-focused models.
Those projects provide South Korea with an existing pool of engineering teams, training experience and model infrastructure that could feed into future frontier AI work, even though participation in the new programme has not yet been finalised.
What Does ‘Frontier AI’ Actually Mean?
Frontier AI generally refers to the most capable generation of general-purpose artificial intelligence models available at a given time.
These systems typically perform across many tasks rather than being built for one narrow function. They may handle text, code, images, reasoning, tool use and increasingly multimodal inputs.
The term does not have one universal technical threshold, and governments and companies can use it differently.
In South Korea’s case, officials are using the phrase to describe a model intended to compete with leading global systems rather than simply meet domestic application needs.
Why Chinese AI Models Are Part of the Competitive Pressure
South Korean reporting has specifically pointed to advances by Chinese open-source and open-weight models as part of the pressure behind the new programme.
Chinese developers have become increasingly competitive on model performance while often making models available at lower deployment costs than some leading proprietary systems.
That has changed the AI competition from a simple U.S.-versus-China infrastructure race into a wider battle over model quality, openness, cost and developer adoption.
For South Korea, the risk is that local companies become dependent on foreign systems both for frontier capability and for widely used open models.
Technology Impact: More Than Just One Model
The programme could influence South Korea’s wider technology ecosystem even if the final model itself does not become the world’s most powerful.
Large-scale AI development requires semiconductor capacity, high-speed networking, data centres, cloud infrastructure, model evaluation tools and specialist software.
Government investment can therefore create demand across multiple technology sectors.
It could also give local AI researchers access to training infrastructure that would otherwise be difficult to secure because advanced computing resources are expensive and globally constrained.
Business Impact for Korean Technology Companies
For Korean technology companies, the programme could create both opportunities and competitive pressure.
Companies selected to participate could gain access to government-backed computing, data and capital. Suppliers of chips, networking equipment, cloud services and data-centre infrastructure could also benefit indirectly.
At the same time, concentrating resources on a small number of teams means some firms may receive much more support than others.
That makes the design of the selection process commercially important, especially for startups that may have strong research but less access to capital than large conglomerates.
Could Startups Really Compete for the Lead Role?
The government has said it wants startups and research teams to have meaningful opportunities to participate.
In practice, leading a frontier AI project requires enormous financial and technical capacity.
A startup may therefore be more likely to participate through a consortium or special-purpose vehicle that brings together capital, compute providers, researchers and larger industrial partners.
The final tender rules will determine whether the programme genuinely creates space for smaller companies or mainly favours organisations that already control large infrastructure.
National Security and AI Sovereignty
South Korea’s AI policy is also shaped by concerns about technological sovereignty.
Relying entirely on foreign AI models can create strategic risks around data, security, language coverage, industrial dependence and access during geopolitical disputes.
A domestic frontier model would not remove South Korea’s dependence on imported semiconductor technology or global software ecosystems, but it could give the country more control over a critical layer of AI capability.
The government’s existing sovereign model programme was explicitly designed in part to reduce technological, cultural, economic and security dependence on global models.
The $3.5 Billion Figure Needs Context
The 4.7 trillion won figure is large, but it should not be interpreted as money that has already been fully spent or transferred to a developer.
The programme is planned for the 2027 budget process and will require parliamentary approval.
The final amount deployed, the public-private funding ratio and the timing of investment may change as the project is structured.
For that reason, Mukhbir is describing the figure as the planned size of the programme rather than a completed investment.
How This Compares With Korea’s Broader AI Spending
The frontier AI initiative sits within a wider increase in Korean AI investment.
Official ministry plans for 2026 included 4.46 trillion won for nationwide AI transformation priorities, up from 3.44 trillion won in 2025.
That broader funding covers infrastructure, GPUs, AI semiconductors, networking, data and talent rather than only one model.
The frontier programme therefore represents a more concentrated high-performance bet inside a much larger national AI strategy.
What It Could Mean for Consumers
Consumers are unlikely to see an immediate new chatbot in March 2027 simply because development begins.
Frontier models take time to train, evaluate and deploy safely.
If the project succeeds, Korean consumers could eventually gain access to more capable domestic AI products with stronger Korean-language performance and closer integration with local services.
Businesses and public agencies may also gain alternatives to relying exclusively on foreign model providers.
South Asia and Pakistan Relevance
There is no direct Pakistan-specific component in South Korea’s programme.
The broader relevance to South Asia is strategic. Governments and technology companies across the region are also debating how much computing capacity, sovereign AI capability and domestic model development they should fund.
South Korea’s approach provides an example of a state using public capital, infrastructure and competitive selection to support frontier AI rather than relying only on private-sector investment.
For Pakistan, any lesson is policy-oriented rather than immediate: the project shows how countries with advanced technology sectors are treating AI compute, talent and models as strategic infrastructure.
What Happens Next?
The next major milestone is the National Assembly’s decision on the 2027 budget.
If funding is approved, the Ministry of Science and ICT plans to open a competitive selection process for the organisation or consortium that will lead the frontier programme.
A winner could be selected as early as February 2027, followed by a planned March start.
At the same time, South Korea’s existing sovereign foundation-model programme will continue through its next evaluation stage, making 2027 a key year for both tracks of the country’s AI strategy.
Important Questions Still Unanswered
The final governance structure has not yet been fully disclosed.
It remains unclear exactly how much private capital will accompany the 4.7 trillion won government commitment, how ownership will be divided, who will control intellectual property and how commercial returns will be shared.
The exact computing allocation, model architecture, safety requirements and evaluation benchmarks are also not final.
Those details will determine whether the programme becomes a competitive national AI platform or simply another large technology spending programme.
FAQs
How much is South Korea planning to spend on the frontier AI programme?
The government plans a 4.7 trillion won programme, equivalent to about $3.5 billion, subject to the 2027 budget process.
When will South Korea start developing the frontier AI model?
The Ministry of Science and ICT is targeting March 2027, after the budget is approved and a lead developer is selected.
Has the lead developer already been chosen?
No. Reuters reported that the government plans a competitive tender, with a winner potentially selected as early as February 2027.
Will the new programme replace South Korea’s sovereign AI project?
No. The ministry says the existing sovereign foundation-model programme will continue separately while the frontier initiative focuses on top-end model performance.
Who can participate in the programme?
The government expects to allow participation by individual companies, consortiums or special-purpose entities, with startups, universities and researchers also able to play roles.
What resources will the government provide?
The plan combines public equity investment with private capital and is expected to concentrate computing chips, data and AI talent on the project.
Why is South Korea investing so heavily in AI?
The government wants to improve national competitiveness, reduce dependence on foreign models and compete more directly with leading AI developers in the United States and China.
Does the programme directly affect Pakistan?
No direct Pakistan project has been announced. Its relevance is mainly as an example of national AI strategy, infrastructure investment and sovereign-model policy.
About The Author
