SK Hynix scouts global locations for new facilities as AI memory demand surges

SK Hynix, the company that makes the memory chips powering most of the world’s AI infrastructure, is running out of room to build them. CEO Kwak Noh-Jung says the company is actively exploring global locations for new business opportunities, driven by a simple problem: demand for high-bandwidth memory is growing faster than the company can expand its existing facilities.

The South Korean chipmaker currently commands 56.4% of the HBM market revenue share as of Q1 2026, making it the dominant supplier of the memory technology that companies like Nvidia rely on for their AI accelerators.

A multi-continent buildout

On August 27, 2026, the company broke ground on a $4 billion advanced packaging facility in West Lafayette, Indiana. The site represents SK Hynix’s first major US investment focused on HBM production, with cleanroom operations targeted for October 2028 and mass production slated for the second half of 2029.

The CEO’s criteria for new sites include stable power, abundant water, available land, and a supportive ecosystem.

Beyond the US, SK Group Chairman Chey Tae-won has flagged Japan as another potential location for semiconductor investments, specifically Miyagi Prefecture. No final decisions have been made on overseas fab plans beyond the Indiana project.

The domestic side of the equation is even more staggering. SK Hynix has laid out a total investment plan of KRW 1,100 trillion, roughly $712 billion, for production facilities across South Korea. That breaks down to KRW 600 trillion earmarked for Yongin, KRW 100 trillion for Cheongju, and KRW 400 trillion for an entirely new cluster in southwestern Korea.

The competitive landscape

SK Hynix is the second-largest supplier of both DRAM and NAND flash memory by revenue worldwide, sitting behind Samsung. But in the HBM segment specifically, it has pulled ahead decisively.

With mass production at the Indiana facility not expected until late 2029, SK Hynix is making a three-year forward bet on AI memory demand.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.