South Korea wants to move its semiconductor industry out of the Seoul suburbs and into the country's southwest, where cheaper land and a cleaner grid promise room to grow. The constraint is not the announced capital. It is the engineering, the resources, and the nine years it took to build the cluster that already exists.
On June 29, 2026, President Lee Jae Myung stood alongside Samsung Electronics chairman Lee Jae-yong and SK Hynix chairman Chey Tae-won to unveil a combined pledge of roughly 800 trillion won, about US$518 billion, for new chip plants outside the greater Seoul metropolitan area, where the country's memory-chip industry has been concentrated for decades (Paris Guardian; Washington Times). Each company committed to two new fabrication plants in the southwest, on top of their existing Gyeonggi Province complexes. Samsung's new sites sit around Gwangju and may include a military air base that is scheduled for relocation (PBS NewsHour; LA Times).
A Korea JoongAng Daily follow-up reports the combined pledges at 896 trillion won (Korea JoongAng Daily follow-up), and TechCrunch, leaning on a broader framing that bundles separate investment plans, describes the total as exceeding US$550 billion, under the headline "RAMageddon" relief for the AI memory squeeze (TechCrunch). Neither company has published a completion timeline or a capacity-per-fab figure, so the headline dollar amount reads as a multi-year ceiling, not a near-term outlay. The spread between 800 trillion won and 896 trillion won is itself a marker of how fluid the announcement still is.
The scale matters because Samsung and SK Hynix together produce roughly two-thirds of the world's memory chips, the storage hardware that AI training and inference runs depend on. Both firms have posted record profits on AI-driven demand, and a bottleneck at these two suppliers ripples outward into the cost and availability of high-bandwidth memory for AI accelerators (MarketScreener; Construction Review).
President Lee's pitch is partly a deconcentration policy. State news agency Yonhap confirmed he has committed active government support for the cluster, including help with land, permits, and workforce pipelines (Yonhap). The administration's argument for the southwest, repeated across the wire versions, is that its renewable-energy capacity gives chipmakers a cleaner grid, an advantage as AI customers increasingly press suppliers on fab carbon intensity.
That argument has not been tested at the scale being proposed. SK Hynix's Chey used the same ceremony to warn that the Gyeonggi cluster he now runs took about nine years to establish, and that the southwest will need "vast sites, along with sufficient power, water and skilled workers." The energy pitch is a government's; the timeline and the resource math are a builder's. Government officials have publicly dismissed power-and-water concerns raised against the southwest site, but a Korea JoongAng Daily opinion column flagged execution risk squarely: Korean mega projects have historically under-delivered versus initial pledges, with timeline slippage and water-and-power siting disputes the recurring failure modes (Korea JoongAng Daily opinion).
The funding announcements still leave the most consequential variables off the page. No completion date, no capacity-per-plant figure, and no water-rights or grid-interconnection schedule has been disclosed by either company or the administration. Watch, in order of falsifiability: published fab completion dates; power purchase agreements tied to specific southwest sites; and any water-rights concession or dam allocation disclosed against the Gwangju and Yongin-adjacent footprints. Until then, $518 billion is the political shape of a statement, not the engineering of a schedule.