Overview
- NVIDIA and SK Hynix announced a multi‑year technology partnership on Monday that commits SK Hynix to jointly develop and supply custom high‑end memory for multiple NVIDIA platforms including Vera Rubin, Vera CPU, RTX Spark and Jetson Thor.
- The deal calls for SK Hynix to adopt NVIDIA software tools such as CUDA‑X, PhysicsNeMo, Omniverse and cuOpt to speed chip simulation, photolithography computations and to build wafer‑fab digital twins for more automated factory operations.
- NVIDIA CEO Jensen Huang told reporters that the Vera CPU will use SK Hynix memory and warned that high‑end memory shortages are likely to persist for several years because demand outstrips capacity across wafer, packaging and related supply steps.
- Markets reacted positively as HBM4 — the next‑generation high‑bandwidth memory — has entered volume production and NVIDIA has already tied HBM4 supply to Samsung, SK Hynix and Micron, heightening near‑term production competition and pushing storage stocks higher.
- Analysts say the pact both locks NVIDIA into a stable supply chain for AI infrastructure and spreads NVIDIA’s software and simulation know‑how to a key supplier, which could shorten validation cycles, shape pricing and sustain a multi‑year high‑demand phase for memory makers.