Particle.news
Download on the App Store
6 ARTICLES
·
4h ago

SK Group Pledges to Ease AI Memory Bottleneck With New Fabs and Deeper Partnerships

The company set near-term capacity milestones to redirect the AI race toward efficiency.

SK Group Chairman Chey Tae-won speaks to reporters at the venue of the SK AI Summit in southern Seoul on Nov. 3, 2025. (Yonhap)
Image
SK Telecom Co. CEO Jung Jai-hun delivers the keynote speech at the annual SK AI Summit 2025 in southern Seoul on Nov. 3, 2025, in this photo provided by the company. (PHOTO NOT FOR SALE) (Yonhap)
Kwak Noh-jung, chief executive officer of SK hynx Inc., delivers a speech during the SK AI Summit in southern Seoul on Nov. 3, 2025. (Yonhap)

Overview

  • Chairman Chey Tae-won said SK will accelerate output to meet AI-driven demand, with SK hynix’s M15X fab in Cheongju ready to begin operations in 2026.
  • The Yongin semiconductor cluster is targeted to reach capacity equivalent to 24 Cheongju fabs by 2027 to bolster memory supply.
  • SK hynix CEO Kwak Noh-jung unveiled an HBM roadmap that starts with 16-layer HBM4/HBM4E in 2026 and advances to HBM5/HBM5E between 2029 and 2031, as the company moves to become a full-stack AI memory creator.
  • SK is tightening collaboration with OpenAI, Nvidia and AWS, with OpenAI’s Stargate project reported to seek roughly 900,000 HBM wafers per month and both sides exploring AI data centers in Korea.
  • Nvidia was cited as planning up to 260,000 GPUs in South Korea, including 50,000 for SK, as the companies discuss a manufacturing AI cloud leveraging the Omniverse platform.