The first stage of the competition recently concluded with the surprising elimination of Motif Technologies, a small startup that produced the highest-performing open-source model, and Naver Cloud, which was disqualified for using foreign software components.
This massive capital expenditure, or investment in physical assets, aims to build 18.4 gigawatts of data center capacity by 2035 to ensure the nation’s infrastructure can support large-scale AI development.
By establishing its own models and facilities, South Korea intends to avoid potential restrictions or service delays from American frontier model labs like OpenAI or Anthropic.
This move positions the country as a global leader in the sovereign AI movement, treating AI infrastructure as a national priority similar to its successful broadband expansion in the 1990s.
While the project seeks self-reliance, it has paradoxically strengthened ties with the American chipmaker Nvidia.
Because South Korea currently lacks a domestic producer of high-end AI chips (accelerators) that can compete with international standards, its new data centers will rely heavily on Nvidia’s hardware.
Sources indicate that Nvidia is using these national projects to diversify its customer base and secure long-term deals for high-bandwidth memory from Korean suppliers like SK Hynix.
This creates a technical mechanism where South Korea provides the memory chips and infrastructure, while Nvidia provides the essential processors to power the country's independent AI ambitions.