The government announced a 'national artificial intelligence (AI) capability enhancement plan' that includes investing over 1 trillion won within the year to secure 10,000 graphics processing units (GPUs). However, there are criticisms from within and outside the industry that, in a situation where the technological gap with global top tiers like the United States and China is widening, the government is engaging in 'showy investment.' In particular, there are numerous points made that simply securing 10,000 GPUs without measures to secure AI talent cannot be a fundamental plan for nurturing AI.
According to the industry on the 22nd, over 80% of the government proposal for a supplementary budget of 1.8 trillion won for the AI sector, which was approved at the Cabinet meeting on the 18th, is allocated for the purpose of securing GPUs. Specifically, 1.46 trillion won will be invested in securing 10,000 GPUs within the year. The 10,000 GPUs purchased by the government are expected to be provided to domestic corporations and academia for developing AI foundation models.
Additionally, 193.6 billion won will be used for the AI model development project called 'World Best LLM Project,' 75.2 billion won for supporting the early commercialization demonstration project of AI semiconductors, and 55 billion won for establishing an AI innovation startup fund.
However, the industry notes that compared to the investment levels pursued by global AI powerhouses like the United States and China, the proposed budgets are grossly inadequate. The U.S. government announced plans to invest $500 billion (730 trillion won) in building AI data centers over four years through a private joint investment called Stargate this January. The Chinese government plans to invest 10 trillion yuan (about 1,946 trillion won) to secure AI industry competitiveness by 2030.
In this situation, the investment scale announced by our government is only 1.8 trillion won, and there are criticisms that it is overly fixated on securing GPUs. Kim Kyung-won, a chair professor at Sejong University's Business Administration department, noted, 'Given that the AI duopoly of the U.S. and China has already solidified, and it is difficult to catch up with the technological gap or investment scale, securing 10,000 GPUs cannot be a fundamental solution.' He emphasized that this is merely a measure for show without effectiveness.
Above all, the most serious issue is the government's lack of interest in securing AI talent. Of this supplementary budget for AI, only 45 billion won related to securing AI talent accounts for 2.4% of the supplementary budget. The Ministry of Science and ICT has allocated 10 billion won for hosting global AI competitions, 5 billion won for supporting the recruitment of foreign AI researchers to Korea, and 30 billion won for improving the treatment and research investment for 400 postdoctoral researchers.
Kim Yong-seok, head of the Semiconductor Education Center at Gachon University, stated, 'Even if there are hardware (GPUs), without AI talent to properly utilize them, they become useless.' He emphasized that nurturing talent for developing AI foundation models and compilers is particularly important.
According to the Paulson Institute at the University of Chicago, among the top 20% AI researchers worldwide, the largest percentage of their origins is 47% from China, followed by the United States (18%) and Europe (17%). Only 2% are from South Korea.