NVIDIA announced on its official website on June 7 that it has entered into a multi-year technology collaboration agreement with SK Hynix, focusing on joint development of next-generation memory and expansion of supply. The specific areas of cooperation include jointly developing memory for NVIDIA Vera Rubin AI supercomputers, Vera CPU, RTX Spark PC, and the Jetson Thor robotics computing platform.
Full-stack applications from CUDA-X to Omniverse
The agreement confirms the NVIDIA technology tools both parties will use:
NVIDIA CUDA-X library: accelerates semiconductor simulation, including TCAD workflows and computational lithography
NVIDIA PhysicsNeMo framework: used for SK Hynix internal simulation code and AI physics workflows
NVIDIA Omniverse library + OpenUSD pipeline: builds 3D factory scenes for digital twin visualization and simulation
NVIDIA cuOpt: GPU-accelerated decision optimization engine, supporting autonomous mobile robots and wafer fab asset management
NVIDIA Metropolis platform: autonomous mobile robot management
Digital twins for the wafer fab: a technical path to autonomous operations
The agreement covers SK Hynix’s efforts to advance digital twin technology for wafer fabs, as a foundation for autonomous wafer fab operations. SK Hynix will use Omniverse, OpenUSD, and cuOpt to build 3D factory scenes that support visualization, simulation, and optimization of complex semiconductor manufacturing environments, and explore how to connect digital twins with existing traditional software and AI workflows so that AI systems can perform inference on wafer fab data, automate tasks, and improve manufacturing decisions.
FAQ
Which specific products does this collaboration between NVIDIA and SK Hynix cover?
Based on NVIDIA’s official website announcement, the collaboration covers memory co-development of four named platforms: the Vera Rubin AI supercomputer, Vera CPU, RTX Spark PC, and the Jetson Thor robotics computing platform. SK Hynix will use this to expand into three new market categories: AI infrastructure, personal AI, and physical AI.
What are the multi-year term and the specific scale of this cooperation agreement?
The announcement states that the agreement is “multi-year,” but NVIDIA’s official website announcement does not disclose specific collaboration years, funding scale, or financial details such as order volume.
What are SK Hynix’s specific applications in the field of digital twins for wafer fabs?
According to the announcement, SK Hynix will use NVIDIA Omniverse libraries, the OpenUSD pipeline, and the cuOpt engine to build a digital twin platform supporting fully autonomous wafer fab operations, including factory scene visualization, manufacturing process optimization, and autonomous mobile robot management.
