Samsung Yongin Chip Factory Moves to 2029 as AI Memory Demand Pulls Timeline Forward
A one to two year acceleration from the 2030-2031 original target. South Korea is also adding a new semiconductor cluster in Gwangju.
Samsung Yongin chip factory, a major AI memory fab under construction south of Seoul, will start operations in 2029 - one to two years ahead of the 2030-2031 target Samsung set when the project launched. Reuters reported the confirmation today, citing a statement from a Samsung spokesperson. AI memory demand drove the decision.
Memory chips are inside every AI training cluster and inference server built today. HBM - the stacked DRAM format inside Nvidia's H100, H200, and Blackwell GPUs - ships in limited quantities that cloud providers and hyperscalers compete to lock in, with Samsung and SK Hynix as the only two suppliers currently at volume scale. Pulling the Yongin fab forward by two years means more AI memory capacity comes online before the demand wave that AI infrastructure buildouts are driving toward.
South Korea Plans to Double Memory Chip Output Within Five Years
Samsung's move follows a government-level push. Last month, Samsung and SK Hynix each pledged hundreds of billions of dollars in domestic production after President Lee Jae Myung called for measures to address regional economic inequality across South Korea. Yongin is the anchor of that plan, but not the whole of it.
Seoul's second piece is Gwangju. South Korea intends to build an additional chip cluster in Gwangju city alongside the Yongin expansion, pairing a new geographic hub with existing Samsung and SK Hynix sites. Between the two locations and SK Hynix's own expansion commitments, the government targets a doubling of South Korea's total memory chip production capacity within five years.
SK Hynix Currently Leads on HBM - Yongin Gives Samsung a Production Platform to Compete
SK Hynix leads. That company holds the dominant share of HBM3e supply to Nvidia, covering the memory format inside Blackwell and H200 GPUs - the hardware developers rent on AWS, Azure, and Google Cloud when running large AI workloads. Samsung has been qualifying its own HBM3e with major GPU partners, and getting the Yongin chip factory operational in 2029 instead of 2031 gives Samsung a dedicated production site two years earlier in a competition where SK Hynix currently holds the strongest position.
Samsung has published no investment figure for the Yongin chip factory. SK Hynix's parallel expansion commitments and the new Gwangju cluster also carry no announced cost estimates, leaving the full price tag of South Korea's five-year capacity doubling unquantified.
