Jensen Huang Accelerates Again: NVIDIA and SK hynix Together for Next-Generation AI Platforms

NVIDIA and SK hynix have announced a multi-year technological agreement aimed at strengthening the development of future generations of memories intended for AI infrastructure. The agreement expands an already established collaboration between the two companies and aims to support the growing demand for advanced components necessary for the global spread of so-called AI Factory.

According to the two companies' communication, the project foresees a close integration between NVIDIA's infrastructure roadmap and the future memory technologies developed by SK hynix. The goal is to ensure continuity in the availability of strategic components in a context characterized by increasingly longer development cycles, complex production processes, and particularly high industrial investments.

The collaboration will not be limited to platforms for training large AI models but will involve several emerging segments. SK hynix will contribute to the development of memories aimed at future AI supercomputers based on the Vera Rubin architecture, the Vera CPUs, PCs equipped with RTX Spark solutions, and Jetson Thor robotic platforms. These are markets that NVIDIA identifies as part of the expansion of artificial intelligence into the realms of personal AI and physical AI.

The agreement also includes extensive use of artificial intelligence in semiconductor design and manufacturing processes. SK hynix will utilize NVIDIA's CUDA-X libraries and the PhysicsNeMo framework to accelerate simulations related to Technology Computer-Aided Design (TCAD), computational lithography, and other internally developed simulation codes. The intention is to reduce the time and cost of research and development activities while simultaneously increasing the accuracy of the analyses necessary for designing the most advanced chips.

The two companies also see opportunities for extended collaboration with the Electronic Design Automation (EDA) tool sector, creating an ecosystem that involves semiconductor manufacturers, software developers, and NVIDIA itself to further improve simulation and design flows.

Another pillar of the agreement concerns the digitalization of production facilities. SK hynix is working on creating digital twins of its semiconductor factories using technologies like NVIDIA Omniverse, OpenUSD, and the cuOpt optimization engine. By creating virtual replicas of the plants, it will be possible to simulate, monitor, and optimize complex production processes, including the management of autonomous robots and other resources present on production lines.

The companies are also evaluating the integration of digital twins with legacy software and agent-based AI systems, with the goal of automating operational tasks, analyzing data from the facilities, and supporting increasingly autonomous decision-making processes within the factories.