SK, Nvidia chiefs meet again as AI memory race intensifies - theinvestor.co.kr

Recent meetings between NVIDIA CEO Jensen Huang and SK Group Chairman Chey Tae-won underscore the deepening collaboration in the AI memory sector, as demand for high bandwidth memory (HBM) surges ahea... Read original →

The third Huang-Chey meeting this year signals a strategic pivot beyond procurement—it’s a race to co-define the AI memory stack. With HBM4 set to bottleneck NVIDIA’s Vera Rubin platform, SK hynix leverages its TSV and hybrid bonding edge to transition from vendor to architecture partner. This pressures Samsung to accelerate 3nm GAA integration with HBM4 and may push Micron to reroute capacity through packaging facilities in Taiwan, China. U.S. export controls on advanced tools have already raised HBM yield management costs by ~18%, while Japan-South Korea material restrictions inflate supply chain risk premiums. Within 18 months, NVIDIA is likely to lock in over 70% of SK hynix’s HBM4 output via prepayments and co-development deals, starving AMD and custom AI chipmakers of critical bandwidth and cementing a GPU-HBM vertical alliance that redefines hardware moats in AI.