Nvidia and SK hynix deepen partnership as Jensen Huang warns memory shortage could persist for several years

Nvidia CEO Jensen Huang met SK hynix leadership in Seoul and stated the global memory shortage "is going to persist for several years," signaling a structural constraint rather than cyclical tightness. A formal expansion of cooperation between the two companies was teased as a possible announcement.

SK hynix is the leading supplier of HBM3E and HBM4 (high-bandwidth memory used in GPU accelerators) to Nvidia, with estimates suggesting it could provide roughly 50% to 70% of Nvidia's HBM4 requirements. An expanded agreement likely includes multi-year volume commitments, accelerated production ramps for HBM4 and HBM4E, yield optimization partnerships, and pricing frameworks tied to capacity expansion.

The partnership is evolving beyond chip procurement into industrial co-design, including joint optimization of AI data centers for GPU and memory architecture, custom memory configurations for AI workloads, and use of Nvidia Omniverse for semiconductor plant simulation. SK hynix has outlined plans to roughly double memory capacity over a five-year horizon.