NVIDIA and SK hynix sign multiyear deal to co-develop memory for AI factories

NVIDIA and SK hynix said on June 7, 2026 that they have agreed to a multiyear technology partnership to develop next-generation memory for the global buildout of AI factories and to speed semiconductor design and manufacturing. The deal spans both the chips that go into NVIDIA's systems and the tools SK hynix uses to make them.

Under the deal, SK hynix will co-develop memory for NVIDIA Vera Rubin AI supercomputers, NVIDIA Vera central processors, NVIDIA RTX Spark-powered PCs and NVIDIA Jetson Thor robotic computing platforms, NVIDIA said. That breadth — from data center to robot — signals how central high-bandwidth memory has become to AI system performance, where memory bandwidth, not raw compute, is often the bottleneck.

The partnership also reaches into SK hynix's own factories. The company is using NVIDIA's CUDA-X libraries and PhysicsNeMo framework to accelerate TCAD semiconductor simulations, and building digital twins of its fabs with NVIDIA Omniverse and cuOpt to move toward autonomous manufacturing operations.

NVIDIA chief executive Jensen Huang said advanced memory is essential to AI factory performance and called SK hynix an "extraordinary partner." SK Group chairman Chey Tae-won said the two companies will co-develop next-generation memory and apply AI across chip design and manufacturing.

The announcement is light on specifics that would let outsiders gauge its scale: no volumes, dollar figures, or timelines for the co-developed memory were disclosed, and the description of the work remains largely directional.

Still, the pairing matters for both sides. NVIDIA depends on a small pool of memory suppliers to feed its accelerators, and locking in joint development with one of the largest helps secure that supply as it ships successive AI platforms.