President Lee Jae Myung stood with the heads of Samsung Electronics and SK Hynix on June 29, 2026, and announced that South Korea's two memory-chip giants would invest 800 trillion won — roughly $518 billion — with suppliers to build four new fabs in the country's southwest. "We must secure the core elements of AI faster than any other country," Lee said, framing the plan around a "triple axis" of semiconductors, physical AI and data centers. Add in a parallel data-center push from SK Group, GS Group and Naver, and the headline figure Seoul is using for the whole package tops $576 billion — with Science Minister Bae Kyung-hoon sketching an even longer horizon, targeting 1,000 trillion won and 18.4 gigawatts of AI data-center capacity by 2035.
Samsung's share is 425 trillion won for two memory fabs and a national AI computing center, sited in Gwangju, on land that includes a military airbase now slated for relocation. Local governments in Gwangju and South Jeolla province are chipping in an additional 5 to 20 trillion won of their own. A separate 81 trillion won packaging cluster is planned further north, near Seoul in the Chungcheong region. Investors did not greet the news warmly: Samsung's shares fell 4.86 percent the day of the announcement and SK Hynix dropped 1.68 percent, a reaction that reads less as skepticism about AI demand than as concern over the sheer scale of capital being committed at once. Regional construction and electrical-equipment stocks tied to Gwangju and South Jeolla moved the opposite direction, some hitting daily limit-up.
A Supplier Base That Isn't There Yet
The complication sits underneath the ceremony. Ministry of Trade, Industry and Energy data submitted to lawmaker Koo Ja-keun and first reported by Chosun Ilbo found that Honam — the region encompassing Gwangju and Jeolla — has the weakest base of semiconductor materials, components and equipment suppliers of any of the candidate cluster regions the government considered. Chipmaking is not a single factory; it is an ecosystem of hundreds of specialized suppliers clustered within trucking distance of the fab, the way Samsung's existing Pyeongtaek and Hwaseong sites in Gyeonggi province are surrounded by decades of accumulated supply chain. Building that ecosystem from a standing start in Honam is a multi-year problem layered on top of the years-long timeline before the fabs themselves produce anything.
Eighty-five percent of Honam voters backed Lee in the last presidential election. The opposition says that number, not the supplier data, is what actually decided where the fabs would go.
That gap has handed Lee's opponents a ready-made line of attack: that the Gwangju site was chosen for regional politics rather than industrial logic. Honam is the traditional stronghold of Lee's Democratic Party, and critics argue the announcement functions as an act of political patronage dressed up as chip strategy. Lee pushed back directly over the weekend before the announcement, posting on X to reject the favoritism charge, though the specific content of those posts wasn't independently retrieved for this piece. The dispute won't resolve quickly — permitting, land transfer for the airbase relocation, and supplier recruitment will all take years to play out, giving the political argument as long a runway as the industrial one.
Analysts are more focused on a different risk: whether Korea is building into a memory glut. Morningstar's Jing Jie Yu has warned that accelerating capital spending over the next decade "increases the risk of an oversupply longer term," a caution echoed elsewhere in the market-research community. The counterargument is timing: most of the newly announced capacity is not expected online before 2030, and near-term-announced capacity is unlikely to ship before 2027 at the earliest, meaning any oversupply is a multi-year-out risk rather than an immediate one. Deutsche Bank's own research reportedly expects DRAM markets to stay tight beyond 2028. In the meantime, SK Hynix's bet on high-bandwidth memory — the specialized chip stacks that feed AI accelerators — has already made it more valuable than Samsung by some measures, sharpening a competitive dynamic the new fabs are explicitly designed to extend.
Seoul's framing is that this is industrial policy in service of regional rebalancing as much as AI competitiveness — narrowing the gap between a Seoul-centered economy and everywhere else was a Lee campaign pledge before it was a chip strategy. Whether Gwangju becomes a genuine second hub for Korean semiconductors, or a very expensive test of how much political will can substitute for an industrial base that took Gyeonggi province forty years to build, is a question the 2028 supplier data will answer more honestly than this week's stock prices did.