AI/ML
SK Hynix and Nvidia have a multi-year, strategic co-operation agreement to co-develop memory technology for Nvidia GPU servers, edge RTX Spark workstations and Jetson robots, with SK Hynix using Nvidia SW technology for semiconductor factory digital twin simulation and optimising manufacturing with robots.
They say they are partnering to to advance next-generation memory for the global AI factory buildout. SK Hynix supplies high-bandwidth memory (HBM) to Nvidia for use with its GPUs such as the Blackwell systems and coming Vera Rubin accelerators with HBM4 technology. It is one of three suppliers along with Samsung and Micron. Neither Micron nor Samsung have a similar strategic agreement with Nvidia.
SK Group chairman Chey Tae-Won said: “SK Hynix and Nvidia have been building toward this for years, and this partnership reflects the depth of that collaboration. Together, we are co-developing the next generation of memory for AI factories and applying AI to how we design and manufacture semiconductors — work that will shape the future of AI infrastructure.”
This deal shifts its Nvidia relationship from an SK Hynix supply contractor to a broader alliance between the SK Group and Nvidia. They will build future AI factories together and SK Hynix will share its memory technology roadmaps with Nvidia. SK Group will launch a gigawatt-class AI factory in Korea next year.
SK Hynix says that, through this partnership, it will diversify into new AI infrastructure, personal AI and physical AI markets Nvidia is creating, co-developing memory not just for for Vera Rubin AI supercomputers, but also for Vera CPUs, RTX Spark powered PCs, and Jetson Thor robotic computing platforms.
Nvidia founder and CEO Jensen Huang said: “AI factories are the engines of the next industrial revolution, and advanced memory is essential to their performance. SK Hynix has been an extraordinary partner to Nvidia, playing a central role in delivering advanced memory technologies for Nvidia AI computing platforms. Together, we will co-develop the next generation of memory for AI factories and support the accelerating global expansion of AI infrastructure — from frontier model training to agentic and physical AI.”
At a press event he said: “The age of physical AIN has finally arrived and no country is better prepared for robotics than Korea."
There could well be a long-term agreement (LTA) for SK Hynix memory chip supply as part of this new partnership. The two says their agreement supports supply for advanced memory, addressing the extended development cycles, advanced fabrication and capital investments to sustain the global buildout of AI factories. That suggests to us that money is flowing from Nvidia to SK Hynix somehow but it doesn’t actually specify volume SK Hynix chip supply to Nvidia.
SK Hynix is using Nvidia’s CUDA-X libraries and AI to speed semiconductor simulation, including technology computer-aided design and computational lithography workflows. It’s using CUDA-X and the PhysicsNeMo framework to deliver core workload acceleration across its in-house simulation codes and AI physics workflows.
It says that, by extending these tools to the semiconductor electronic design automation (EDA) and simulation ecosystems, this initiative paves the way for three-way collaborations among chipmakers, Nvidia and EDA software vendors.
SK Hynix is also developing fab digital twins as a foundation for autonomous fab operations. This involves Nvidia Omniverse libraries and OpenUSD pipelines, to build 3D factory scenes for visualizing, simulating and optimizing complex semiconductor manufacturing environments. The digital twins can support operational optimization, including the movement of autonomous mobile robots and other fab assets, using the open source, GPU-accelerated Nvidia cuOpt decision optimization engine and Metropolis platform.
Nvidia and SK Hynix are exploring ways to connect digital twins with existing legacy software and agentic AI workflows, enabling AI systems to reason over fab data, automate tasks and improve manufacturing decision-making.