South Korean tech giants pledge over $900B for AI infrastructure and memory fabs

South Korea’s two largest memory chip companies, Samsung and SK Hynix, announced plans to invest $518 billion to build four new memory fabrication plants in the southwestern region of the country, alongside a $52 billion investment for an HBM (high bandwidth memory) packaging hub in the central region. The commitments were disclosed as part of a broader national investment plan unveiled at a presidential briefing, with the chairmen of both companies in attendance.

The national plan also includes a $356 billion commitment from SK Group, Naver, and GS Energy to construct AI data centers across South Korea through 2035. SK Group separately outlined a medium- to long-term roadmap totaling 2,100 trillion won (~$1.4 trillion), with 1,100 trillion won allocated to expanding semiconductor production capacity and 1,000 trillion won directed toward AI data centers nationwide.

South Korean President Jae Myung Lee framed the investments as critical to the country’s next industrial era, emphasizing semiconductors, physical AI, and AI data centers as the "triple axis" of national strategy. He noted that existing facilities in Yongin and Pyeongtaek, near Seoul, have "already reached their limits," and urged accelerated investment in the southwest to decentralize production.

Samsung separately detailed a 2,655 trillion won (~$1.7 trillion) investment over the next decade, including 425 trillion won for the Honam region, with Gwangju selected as the site for a new semiconductor fab and Haenam for an AI data center. SK Hynix, a core affiliate of SK Group, is central to the chip expansion push, while SK Telecom will lead the development of 15 gigawatts of AI data center capacity.

The scale of commitments—exceeding $900 billion in total—reflects South Korea’s ambition to become an "irreplaceable" industrial power in the AI era. However, the article notes that deep-tech industries like semiconductors and AI operate on timelines that may outpace demand cycles, raising the risk of oversupply if market conditions shift before new capacity comes online.