Overview
- Chair Chey Tae-won said SK will ramp output with the M15X fab in Cheongju starting operations next year and the Yongin cluster targeted for 2027, projects reported at over 140 trillion won combined.
- SK hynix detailed an HBM roadmap to supply HBM4 and HBM4E from 2026 and introduce HBM5 and HBM5E between 2029 and 2031, while shifting to a “full-stack AI memory creator” model with closer co-design.
- SK Telecom plans to scale its AWS-partnered Ulsan AI data center from a planned 100‑megawatt, 60,000‑GPU facility toward a 1‑gigawatt target, with expansion into Vietnam, Malaysia and Singapore also envisioned.
- SK formalized alliances with OpenAI, Nvidia and AWS, including an October MOU with OpenAI for a large data center and a reported OpenAI request for up to 900,000 HBM wafers per month for the Stargate project.
- Nvidia discussed deploying up to 260,000 GPUs in South Korea with 50,000 for SK, and the companies announced a deal for 50,000 Blackwell processors and cooperation on a manufacturing AI cloud using Omniverse.