South Korean President Lee Jae Myung stood in Gwangju on June 29 and announced a national industrial program worth more than 1,000 trillion won, roughly $1 trillion at current exchange rates, spread across the next nine years. If you have been keeping score of what “national AI strategy” is starting to mean when a country writes the full checkbook, this is now the number to beat.
The pieces. Samsung Electronics and SK Hynix, together with their supplier ecosystem, commit 800 trillion won, about $518 billion, to build two new chip fabrication sites each in the southwest province of South Jeolla. SK Group, GS Group, and Naver commit 550 trillion won, about $356 billion, for AI data centers targeting 18.4 gigawatts of aggregate capacity by 2035. A separate 81 trillion won cluster for advanced chip packaging goes up in the Chungcheong region near Seoul. Central government seed money sits at 5 to 20 trillion won on top.
Industry Minister Kim Jung-kwan delivered the fab and cluster details. Science Minister Bae Kyung-hoon covered the data center piece. President Lee’s framing was a “triple axis” of semiconductors, physical AI, and data centers, presented as the play for global AI dominance.
Two things stand out. First, this is roughly the entire annual GDP of Poland, or Taiwan, or Saudi Arabia, being pointed at a single sector by a single mid-sized country. That is not a subsidy package. That is a national posture. Second, the composition is telling. Fabs and packaging are the parts that make actual chips. The data center gigawatts are the parts that consume them. South Korea is not betting on one layer of the AI stack, it is trying to own two of the three physical layers under it and count on Samsung and SK Hynix to keep the third.
Whether that math works depends heavily on whether NVIDIA still needs Korean HBM in 2035, which is a much less obvious question than it was two years ago. It also depends on whether the Chinese fabs that Beijing is pouring similar money into arrive at trailing-edge parity fast enough to eat the low end. But the direction of travel is clear. The nation-state layer of the AI infrastructure buildout has arrived, and the numbers on the scoreboard now start with a T.