Nvidia and SK Hynix Sign Multiyear AI Memory Partnership

Nvidia and SK Hynix signed a multiyear technology partnership on June 8, 2026, to co-develop next-generation memory for AI factories, personal AI, and physical AI markets. The agreement follows a series of high-level meetings in Seoul between Jensen Huang, CEO of Nvidia, and SK Group leadership, including Chairman Chey Tae-won and SK Hynix CEO Kwak Noh-jung.

Under the deal, SK Hynix will produce specialized high-performance memory for Nvidia's Vera CPUs, Vera Rubin AI supercomputers, Jetson Thor robotic platforms, and RTX Spark-powered PCs. As part of the collaboration, SK Hynix will integrate Nvidia's CUDA-X libraries and PhysicsNeMo framework into its chip design workflows and utilize Nvidia Omniverse to create digital twins for autonomous factory operations.

During the visit, Huang warned that global memory shortages are structural rather than cyclical and are likely to persist for several years, affecting the entire supply chain from wafers to packaging. The partnership aims to secure supply visibility for Nvidia, as SK Hynix currently provides an estimated 50% to 70% of the high-bandwidth memory requirements for Nvidia's Blackwell accelerators. Following the announcement, Nvidia shares rose 1.37 percent in overnight Nasdaq trading, while SK Hynix shares fell 7.68 percent on the KSE.