SK Hynix Approved $39 Billion for Two New Fabs — and the Two Are on Completely Different Clocks

SK hynix approved a combined investment of approximately 54 trillion won — roughly $39 billion — across two separate fab projects, announced August 7, 2026. The larger allocation, 35.2 trillion won, funds a second-phase fab at the company's Yongin semiconductor cluster, referred to internally as Yongin Y2. The smaller allocation, 19.1 trillion won, funds a new NAND fab, M17, at SK hynix's Cheongju site.

The two projects are on different construction clocks. Yongin Y2 breaks ground in July 2027, with its first cleanroom scheduled to open in June 2029. Cheongju M17 starts construction earlier, in February 2027, with its first cleanroom opening in December 2028 — roughly six months ahead of Yongin Y2's. Staggering two large fab builds by about half a year is a sequencing choice: it lets SK hynix stage engineering, construction labor, and equipment procurement across sites rather than demanding peak resources at both locations simultaneously, which matters at this scale given how constrained specialized fab-construction capacity has become industry-wide during the current AI buildout.

The two fabs also serve different parts of SK hynix's memory business. Yongin Y2, as part of the broader Yongin DRAM-focused cluster, points toward continued DRAM and HBM capacity expansion — the product category central to SK hynix's position supplying high-bandwidth memory for Nvidia and other AI accelerator makers. Cheongju M17 is explicitly a NAND fab, expanding capacity in the storage-memory side of the business rather than the DRAM/HBM side. Approving major capacity in both categories in the same announcement signals the company sees AI-driven demand pressure across its full memory portfolio, not narrowly concentrated in HBM.

SK hynix framed the investment as a strategic decision to seize opportunities in line with the market's growth speed and to contribute to the stability of the global AI semiconductor supply chain — language that frames this explicitly as AI-driven demand response rather than routine capacity replacement on a normal upgrade cycle.