SK Hynix is reviewing a US chip manufacturing site, Taipei Times reported this week, in what would mark the South Korean memory specialist's first publicly disclosed consideration of an American fab.

The report uses the word "mulling," pointing to an internal assessment rather than a board-approved capital project. Taipei Times did not specify a US location, technology node, product mix, capex figure, or timeline. SK Hynix has not publicly commented on the report.

What does SK Hynix produce?

SK Hynix ranks among the world's top three memory suppliers by revenue, with DRAM and NAND flash as its core product lines. The company also manufactures high-bandwidth memory, or HBM, the stacked DRAM product now central to AI accelerator supply.

DRAM accounts for the majority of SK Hynix's revenue, while NAND flash and SSDs make up a smaller, more cyclical portion. HBM volumes, still a fraction of overall DRAM output, have grown sharply on AI demand from Nvidia, AMD, and hyperscale customers operating large GPU fleets.

Its primary Korean base sits in Icheon, Gyeonggi Province, with a second DRAM packaging facility in Wuxi, China. SK Hynix acquired Intel's NAND and SSD business, including a Dalian wafer fab, in a deal that closed in 2021 for $9 billion.

How does it fit the broader US fab push?

The report lands against a multi-year build-out of US chip plants led by non-American companies:

• TSMC broke ground on its Phoenix, Arizona complex in April 2020. The project has since expanded to three fabs.
• Samsung started construction on a Taylor, Texas foundry fab in 2022, originally budgeted near $17 billion for 4 nm logic.
• Intel is building two Ohio fabs as part of a $20 billion-plus investment plan announced in 2022.
• Micron is constructing a $100 billion New York megafab as part of its domestic memory roadmap.

US policy supports this build-out through the CHIPS and Science Act, signed August 9, 2022, which authorized $52.7 billion in semiconductor manufacturing incentives. Korean chipmakers have so far lagged TSMC and Samsung in committing US wafer capacity.

What stays unconfirmed?

Taipei Times named only the broad concept:

• No US state or metropolitan area
• No capex figure or process node
• No groundbreaking or production start year
• No product allocation between DRAM, NAND, HBM, or advanced packaging

A formal SK Hynix disclosure would likely precede any site selection, incentive negotiation, or capital commitment. US Department of Commerce CHIPS Act award letters typically run several hundred pages and lock in milestone-based payouts over multi-year construction windows.

Why does memory localization matter?

Memory wafer production has concentrated in South Korea, Taiwan, and Japan, leaving the United States with limited domestic DRAM and NAND output. SK Hynix's DRAM and HBM feed directly into US-designed AI accelerators, server CPUs, and GPUs from Nvidia, AMD, and Apple.

Geographic redundancy has moved up the procurement priority list for hyperscale buyers after pandemic-era supply shocks and tightening US export controls on advanced semiconductors to China. A US fab would also let SK Hynix qualify for CHIPS Act incentives and shorten logistics for US customers.

For SK Hynix, the move would mark the first memory wafer capacity outside its established Asian network, a notable strategic step for a company that has concentrated DRAM and NAND production in Korea for three decades.

What's at stake in HBM?

HBM allocates DRAM capacity that would otherwise feed conventional DDR5 modules. SK Hynix, Samsung, and Micron are the only three HBM suppliers at scale. SK Hynix has historically led the segment, supplying HBM3 to Nvidia's H100 and H200 accelerators. Samsung and Micron qualified second-source HBM3E for Nvidia's Blackwell generation.

If SK Hynix committed US HBM capacity, it would be the first HBM wafer line outside Korea. US hyperscalers building GPU fleets would gain a domestic supply node, though HBM production involves through-silicon via stacking that complicates fab selection.