NVIDIA and SK hynix Announce Multiyear Technology Partnership to Supply AI Memory and Deploy Autonomous Factory Digital Twins
NVIDIA and SK hynix form multiyear technology partnership to ensure future supply of advanced memory for worldwide AI infrastructure. The alliance addresses increasing capital requirements and lengthy manufacturing times in modern chip production according to the official joint announcement. It is a joint road map for co design and manufacturing of advanced memories suitable for large scale datacenters and autonomic computing platforms.
Hardware and future silicon development are central elements of this technological cooperation. Jensen Huang founder and CEO of NVIDIA pointed out that
"memories of a higher caliber are crucial to the operations of AI factories."
The duo will jointly co design advanced memories for growing global infrastructure that will power everything from training the largest frontier models to the new agentic and physical AI systems.
"This reflects our deepening bond and future of AI infrastructure"
Chey Tae won Chairman of SK Group said.
SK hynix will broaden its portfolio of memory products for multiple new hardware systems developed by NVIDIA. As part of this cooperation the companies will co develop tailored memories for both the Vera Rubin AI supercomputers and stand alone Vera CPUs. Memory technology will be scaled down for personal and edge computing systems like RTX Spark processor driven desktop PCs and the Jetson Thor robotics platform for humanoid autonomic machines.
The companies will embed machine learning in the chip design process in addition to co development of hardware. SK hynix will use NVIDIA CUDA X software libraries for its internal technology computer aided design flows and computational lithography. In this way the memory maker will use NVIDIA PhysicsNeMo framework to conduct faster, complex physics simulations. This paves the path for 3 way collaborations between chip makers, software companies and design automation specialists.
The third pillar of the cooperation will use automation of the physical process inside semiconductor factories. SK hynix is building digital twins of its factories utilizing NVIDIA Omniverse libraries and OpenUSD workflows. Highly precise 3 D virtual representations enable engineering staff to visualize, simulate and optimize complex production processes before the operations are launched in a real manufacturing floor.
The digital twins will run the NVIDIA cuOpt engine for orchestration of transport of autonomous mobile robots and factory assets within such facilities. This system uses graphics processing unit accelerated path calculation to compute efficient material transfer paths in real time. Researchers will also look into how 3 D virtual models could be combined with agentic AI workflows so that autonomous machines could analyze manufacturing data in real time.
