SK hynix puts 99.5% of its 1,105 trillion won fab program inside South Korea

Add the four named sites and the program comes to 1,105.4 trillion won ^[13]. Of that, 1,100 trillion won is inside South Korea, or 99.5 percent ^[14]. The Indiana fab, at roughly 5.4 trillion won ^[7], is 0.49 percent of the total ^[15], and the Honam cluster is about 74 times its size ^[18].

Yongin carries two price tags in the same sentence, 600 trillion won and $430 billion ^[4], an implied 1,395 won to the dollar, which puts the whole program near $790 billion ^[16]. No company spends that in a year. Yongin is set to begin operating in February next year ^[4]. The account of the forum does not include a spending timetable for the Honam cluster or the Cheongju fab ^[20]. These are site totals across build-out phases.

Kwak said demand for AI memory will remain high as the artificial intelligence industry keeps growing ^[19]. "Without memory, the AI industry cannot exist," Kwak said ^[9]. On the sites themselves: "Through the new global production bases and research foundations we are preparing in step with that demand, we will firmly defend our current technological edge and overwhelming market dominance," he said ^[3]. He said it in Santa Clara, at an event SK hynix has hosted since 2012 for global talent and attended by Big Tech customers and prospective hires ^[11][2].

The capital is not moving toward the customers who buy the memory. Indiana is the only committed number outside Korea ^[7], and additional fabs in the United States and Japan are reported to be under study ^[8].

If the cluster figures include supplier, utility and tenant spending across two decades, SK hynix's own outlay is a fraction of 1,105.4 trillion won ^[13] and a single Indiana fab is not the right comparison. If AI memory demand holds the way Kwak described it ^[19], adding phases at Yongin and Cheongju is cheaper per wafer than starting again in a new country. And if tariffs or customer localisation force American capacity, 5.4 trillion won ^[7] is a deposit and the follow-on number is large.

In my view the verifiable commitment here is Yongin's February start ^[4]; the rest is optionality. The panel that followed the keynote agreed that competition will turn on connecting technologies and selling system-level solutions ^[12]. That is an argument about products. The company is also developing high-bandwidth flash, or HBF, after HBM ^[10]. A Japan or second US fab announced in the tens of trillions of won would show the Korean weighting was a matter of sequencing; a published phase schedule for the 400 trillion won Honam cluster ^[5] would show whether that money belongs to this decade.