The government will invest 1.46 trillion won to expand artificial intelligence (AI) computing infrastructure, aiming to secure 10,000 advanced graphics processing units (GPUs) within this year. Additionally, it decided to allocate 752 billion won for the demonstration project supporting the early commercialization of domestic AI semiconductors, which is 2.5 times higher than the initial budget.
The Ministry of Science and ICT announced on the 18th that it will actively promote the 'National AI Capability Enhancement Plan Follow-Up Measures' (hereinafter referred to as Follow-Up Measures). This Follow-Up Measure was prepared through joint discussions among relevant ministries at the Economic Ministers’ Meeting and the Industrial Competitiveness Enhancement Ministers' Meeting held on the 15th. With the government proposal for a supplementary budget of 1.8 trillion won for the AI sector being approved at the temporary Cabinet Meeting, the implementation of the Follow-Up Measures has intensified. The approved supplementary budget will be submitted to the National Assembly on the 22nd.
This Follow-Up Measure is based on the 'National AI Capability Enhancement Plan' announced jointly by relevant ministries at the 3rd National Artificial Intelligence Committee on February 20. Just like the direction of the enhancement plan, the Follow-Up Measures also focus on expanding AI computing infrastructure, developing next-generation AI models, and accelerating AI transformation.
The Ministry of Science and ICT noted, 'Major countries such as the United States and Europe are making a national effort to gain dominance in AI by investing hundreds of billions of dollars, and it is essential for South Korea to boldly implement national strategies to strengthen AI competitiveness.'
The U.S. government announced in January that it would invest at least $500 billion (about 730 trillion won) over four years for the AI infrastructure project 'Stargate'. The European Union (EU) also plans to invest 300 trillion won in AI, including the 'AI Gigafactory Project', as noted in February. The French government has also announced an investment of 163 trillion won for 'AI data centers'.
Earlier this year, the Chinese AI startup 'DeepSeek' astonished the world by implementing high-performance AI models with efficient algorithms. As a result, a new strategy that can overcome the existing competitive landscape centered on large-scale investments is gaining attention. In response, the government stated it would implement this Follow-Up Measure to strategically concentrate resources to elevate national AI capabilities to the next level.
The Ministry of Science and ICT mentioned regarding this Follow-Up Measure, 'The supplementary government proposal for the AI sector is prepared to practically implement the 'National AI Capability Enhancement Plan,' stating plans to proceed with immediate and effective financial input.'
◇ Securing 10,000 GPUs, the core components for AI development, within this year
Following the implementation of this Follow-Up Measure, the government plans to spend 1.46 trillion won to secure 10,000 advanced GPUs based on clustering within this year. GPUs are essential components for AI training and inference. To respond to urgent domestic AI computing demands, the government plans to invest 172.3 billion won to support the additional rental and utilization of 2,600 advanced GPUs owned by the private sector.
The scale of the demonstration project supporting the early commercialization of domestic AI semiconductors will also be expanded. Initially, 29.8 billion won was planned for this project this year, but the budget has increased to 75.2 billion won through the supplementary budget. The government plans to provide direct support to domestic AI semiconductor corporations with high potential to commercialize neural processing units (NPUs) in a timely manner. In this context, a demonstration infrastructure optimized for large-scale AI computing commercial services using domestic NPUs will also be established.
In order to strengthen tax support for advanced AI research and development (R&D) and AI infrastructure investment, the 'Tax Special Act' was amended last March. Through this, AI has been designated as a national strategic technology. Accordingly, the Ministry of Science and ICT plans to clarify the scope of strategic technologies through amendments to subordinate laws.
A policy to select and intensively support 'AI elite teams' will utilize 193.6 billion won. The intention is to select teams composed of domestic corporations to develop world-class AI models, providing concentrated support for up to three years in resources such as GPUs, data, and talent.
Support for securing and training AI talent will also be provided. The government plans to hold a 'Global AI Challenge' in the second half of the year, costing 10 billion won, where AI talents will compete. A new project will also be initiated this year for attracting top-notch AI researchers from overseas, offering up to 2 billion won annually for three years (50 billion won in total). Additionally, 30 billion won will be invested in 400 outstanding post-doctoral researchers in AI convergence fields, both domestically and internationally.
To support the growth of AI startups, the 'AI Innovation Fund', which is planned to be jointly established between the public and private sectors this year with an initial size of 90 billion won, will be expanded to 200 billion won. The government plans to increase its budget for 2025 from 45 billion won to 100 billion won. To promote the commercialization of technology in the AI sector, the budget for maximizing the utilization of research outcomes and innovative infrastructure in AI sectors within research and development special zones has been increased to 45.1 billion won.
Minister Yeo Sang-im stated, 'In the fierce global competition for AI dominance, this supplementary budget is a timely and significant decision,' and added, 'The Ministry of Science and ICT is preparing for the tasks with the urgent mindset that 'a year of delay means falling three years behind in competitiveness to lay the foundation for AI's leap to a G3 standing.'