NVIDIA and South Korea's SK hynix on June 7, 2026 sealed a multiyear technology partnership to co-develop next-generation memory aligned with NVIDIA's AI factory roadmap. The deal pulls the world's leading HBM supplier and the world's largest AI accelerator vendor into a tighter joint engineering loop, with neither side disclosing financial terms or volume commitments.
Memory Codevelopment Across The Stack
SK hynix will codevelop memory for the Vera Rubin AI supercomputer family, the Vera CPU, RTX Spark-powered PCs and the Jetson Thor robotics platform. The breadth signals NVIDIA's intent to consolidate a memory supply story that spans hyperscale AI factories, agentic personal computing and physical AI — the same hardware base used by reference humanoid designs like Isaac GR00T with Unitree H2 Plus.
AI Designs Its Own Chips
The two companies also said they will apply AI to semiconductor design and manufacturing, leaning on NVIDIA CUDA-X libraries and PhysicsNeMo to accelerate device simulation. That extends NVIDIA's playbook of co-engineering EDA tools with Cadence and dovetails with parallel chip pacts that are pulling the foundry, EDA and memory ecosystems into NVIDIA's orbit.
Market Read
Investors read the announcement as confirmation of memory tightness through 2027 — analysts at Sherwood News flagged the potential knock-on effect on PC DRAM prices — and as a wedge against rival memory supplier Samsung, whose stock fell on Google's TurboQuant compression news earlier this year. With Vera CPU silicon already sampling, the SK hynix lock-in firms up a supply story that hyperscalers are trying to copy.
Reporting based on coverage from NVIDIA, TechPowerUp and Sherwood News.