Direct:
Nvidia (NASDAQ: NVDA) is deepening a multi-generation dependency with SK Hynix (KRX: 000660) that now runs from data center silicon down to PCs and robots. Jensen Huang and SK Group chairman Chey Tae-won announced the partnership on June 7 at SK's Seorin Building in Seoul, framed around memory for AI factories. SK Hynix is estimated to supply 50 to 70% of Nvidia's HBM4 already, and this extends that relationship into Nvidia's CPUs, PCs, and robotics platforms while bringing Nvidia's software into SK Hynix's chip design and fab work.
SK Hynix will codevelop memory for the Vera Rubin supercomputers, Vera CPUs, RTX Spark PCs, and Jetson Thor robotics platforms. It will use CUDA-X and PhysicsNeMo for chip design and semiconductor simulation, and Omniverse, OpenUSD, and cuOpt for digital twins of its own fabs. That gives Nvidia a hand in both the memory specifications and parts of SK Hynix's engineering workflow.
Jensen Huang used the visit to warn that the memory shortage could last for years. Vera Rubin is in full production, with deliveries scheduled to begin in the third quarter, and HBM4 is the supply everyone is watching. Tying its largest HBM4 supplier into a multi-year codevelopment plan is Nvidia protecting its roadmap against a shortage its own demand is driving.
The announcement is still heavy on roadmap intent. It discloses no volumes, pricing, capacity reservations, or purchase commitments. The estimated HBM4 share and Vera Rubin's production schedule are the firmest facts in it. Huang's June 8 meeting with Samsung (KRX: 005930) shows Nvidia is keeping more than one HBM4 supplier in play, with Micron (NASDAQ: MU) also qualified.
For the consumer angle, this lands on DRAM and leaves NAND untouched, so SSD pricing is not directly in play. The RTX Spark and personal AI lines are the thread to watch, because they tie client DRAM development to Nvidia's platforms. Nothing here discloses a retail allocation shift, but more codevelopment aimed at AI factories and Nvidia-spec PCs is capacity competing with the DIMM channel, and Huang has now put a multi-year horizon on the shortage.
Drafted with AI assistance against parallel reporting.
Sources:
- DigiTimes, “Nvidia and SK Hynix announce multiyear memory partnership covering AI servers, PCs, and robotics” (June 8, 2026, paywalled stub: headline and lead only)
- SK hynix Newsroom / NVIDIA Newsroom, “Multiyear Technology Partnership to Advance Memory for AI Factories” (June 7, 2026): scope across Vera Rubin, Vera CPUs, RTX Spark PCs, Jetson Thor; CUDA-X, PhysicsNeMo, Omniverse, OpenUSD, cuOpt; AI infrastructure, personal AI, physical AI
- Korea Herald / Korea Times (June 8, 2026): Seorin Building announcement, Chey Tae-won meeting
- 24/7 Wall St. / Yahoo Finance, “Nvidia and SK hynix to Partner as Jensen Huang Warns Memory Shortage Could 'Last for Years'” (June 7, 2026): shortage warning
- TrendForce, “SK hynix to Supply Memory for NVIDIA Vera CPU as Partnership Deepens; Samsung Meeting on June 8 in Spotlight” (June 8, 2026): Vera CPU memory, ~50-70% HBM4 share context, Samsung meeting, multi-supplier picture
- Tom's Hardware / TechTimes, “Nvidia Vera Rubin Enters Full Production: Samsung, SK Hynix, Micron Named HBM4 Suppliers” (June 2, 2026): full production, Q3 deliveries, Micron qualified
Reddit: https://ift.tt/qiakwyO