SK Hynix 2026 Capex to Hit Record High of KRW 40 Trillion, U.S. and Japan Expansion Timelines Uncertain
Taylor Wilson
SK Hynix confirmed 2026 capital spending at the high end of KRW 40 trillion (roughly $27 billion), betting big on AI memory manufacturing — but has made no formal commitment to new capacity in the U.S. or Japan.
Where does the $27 billion go?
2026 capex is set at the high end of KRW 40 trillion (about $27 billion), directed at expanding AI memory capacity.
Key projects: ramping M15X to volume production, building out Yongin Fab1 cleanrooms (opening early 2027), expanding advanced packaging at PMD7, and developing M17 as a new NAND production site.
This means → the bulk of the money stays in South Korea, across three lines — DRAM, NAND, and packaging — filling domestic capacity first.
How far along is the HBM roadmap?
HBM4 began shipping in Q2 2026 and enters full mass production in the second half.
Samples of the more advanced HBM4E were delivered to key customers in H1 2026, with mass production targeted for 2027.
The company is already developing HBM5 and related thermal solutions — including IHBM, a package-level cooling technology that cuts thermal resistance by more than 30%.
In plain terms = HBM — high-bandwidth memory that feeds data to AI chips — is iterating generation by generation at speed, and heat dissipation is now as hard an engineering problem as the memory itself.
How is the NAND and AI storage portfolio shaped?
321-layer NAND currently accounts for the largest share of production; the target is 50% of domestic capacity by year-end.
AI storage is not a one-product game. SK Hynix is building a portfolio across workloads: high-performance TLC enterprise SSDs, high-capacity QLC enterprise SSDs, and SLC-mode dedicated SSDs.
This means → data-center demand is fragmenting — some workloads need speed, some need capacity, some need endurance — and SK Hynix is covering them with a product matrix, not a single flagship.
SOCAMM2 products based on the 1c-nanometer process have started shipping.
Will the U.S. and Japan fabs actually get built?
The company has made no formal commitment to new capacity in the U.S. or Japan.
Corporate center president Song Hyunjong said overseas capacity decisions will be assessed on commercial viability, power and water supply, labor, local semiconductor ecosystem, and customer accessibility — with "no fixed preference for any particular location."
In plain terms = SK Hynix's stance is "we go where conditions work" — it will not rush a site decision for the sake of a political signal.
This reflects a deeper question: whether U.S. and Japan fabs materialize will be the key test of SK Hynix's real intentions for its global capacity footprint.
Content is for reference only, not financial advice.