SK Hynix approved roughly $38.1 billion, 54 trillion Korean won, in new capital spending to build two additional memory-chip fabs, a Yongin facility focused on HBM and next-generation DRAM, and a Cheongju fab, directly citing sustained AI-driven demand for high-bandwidth memory and enterprise storage as the reason. This lands the same week other reporting indicated Samsung, SK Hynix, and Micron have already sold out their entire 2027 DRAM and HBM production capacity, and it is the clearest evidence yet that memory manufacturers view the current AI memory shortage not as a temporary spike to ride out with existing capacity, but as a structural, multi-year demand shift worth committing tens of billions of dollars in new fixed capacity to meet. That is a meaningfully different signal than a price hike or an allocation cut: building a new fab is a multi-year bet, and the timelines here underline just how far out this capacity crunch extends, construction on the Yongin fab does not even begin until July 2027, with its first cleanroom not coming online until June 2029, meaning the new supply this investment produces will not meaningfully ease today's shortage for roughly three years. For any team planning infrastructure budgets, hardware refresh cycles, or product bills-of-materials that depend on memory pricing, the realistic planning assumption implied by this investment timeline is that DRAM and HBM tightness persists through at least 2027 and likely into 2028, not that relief is coming next year; new fab announcements are a sign the shortage is being taken seriously by suppliers, not a sign it is about to end.