South Korea’s national supercomputer project is facing delays as the country struggles to secure critical AI chips, a key resource for advancing science and technology.
As major global tech companies and governments compete to secure AI chips, these resources are increasingly viewed as crucial for national technological competitiveness. Therefore, in S. Korea, experts are urging that AI be designated a national strategic priority to encourage targeted investment.
According to the Ministry of Science and ICT, the government had originally planned to launch the nation’s “Supercomputer No. 6″ next year. However, the timeline has now been pushed to early 2026.
This supercomputer, budgeted at around 2.9 trillion won ($2 billion), was to be powered by GPUs (graphics processing units) optimized for AI computations. Yet, multiple rounds of bidding ended without any successful candidates last year. In the meantime, the current supercomputer, No. 5, is running at up to 90% capacity, leaving some researchers reliant on overseas supercomputing facilities.
“The spike in demand for AI chips after the ChatGPT boom has created a supply crisis and driven up costs, further impacted by currency fluctuations,” said Hong Tae-young, Director of KISTI’s Supercomputing Infrastructure Center. Even after a 53% budget increase to approximately $3.1 billion, securing the necessary technology remains uncertain.
The global race for securing the latest AI chips highlights how AI competition has essentially become a battle of financial power.
NVIDIA’s H100, one of the most widely used AI chips from the U.S., is currently priced at around $35,500 each. Also as major companies like OpenAI, Microsoft, and Meta, as well as various nations, compete for the latest Blackwell chips, prices are expected to rise further. Nations such as Saudi Arabia and the UAE have reportedly entered the race for AI chips, underscoring the issue’s geopolitical significance.
In pursuit of global AI leadership, countries worldwide are making unprecedented investments. According to Stanford University’s “AI Index Report 2024,” the United States invested $67.2 billion in private AI research last year, nine times China’s $7.7 billion. S. Korea ranked ninth globally with $1.39 billion.
On Wall Street, financial firms have recognized the growing demand for AI investments, introducing a new form of collateralized loans backed by AI chips. Blackstone, Pimco, Carlyle, and BlackRock have provided an estimated $11 billion in loans secured by AI chips to cloud companies since last year, according to the Financial Times.
In South Korea, industry leaders are concerned about falling behind in the race. During SK’s “AI Summit” on Nov. 5, some of AI executives from major S. Korean companies expressed these concerns.
Park Sung-hyun, CEO of Rebellions Inc., emphasized that “affordable and reliable access to GPUs is critical for both industry and research, yet securing this requires national support beyond what individual companies can manage.” Jeong Shin-ah, CEO of Kakao, added that it’s becoming increasingly difficult for any single company to invest the billions of dollars needed for GPUs. “Securing GPUs and internet data centers should be seen as a national priority,” she said.
