Nvidia signs multi-year memory chip deal with SK Hynix - qz.com

NVIDIA and SK Hynix have entered a multi-year technology partnership to develop next-generation memory chips for AI data centers and computing platforms, while also leveraging NVIDIA's CUDA-X and Phys... Read original →

NVIDIA’s deep integration with SK Hynix signals a strategic pivot from pure compute to memory-compute co-optimization in AI infrastructure. Leveraging CUDA-X and PhysicsNeMo not only compresses 3nm DRAM development timelines but embeds EUV lithography and digital twins into the manufacturing loop, forcing equipment vendors to refine process control. While this bolsters supply chain resilience, heavy reliance on U.S.-origin EDA tools exposes SK Hynix to escalating export controls. Samsung will likely accelerate HBM4 development and deepen ties with AMD and Meta to counterbalance market share erosion. Over the next 18 months, nonlinear growth in AI data center memory bandwidth will intensify competition around CoWoS-L and near-memory computing, while geopolitical volatility may compel Korean foundries to fast-track secondary manufacturing hubs outside Taiwan, China.