Nvidia has inked a multi‑year agreement with SK Hynix to design and produce next‑generation high‑bandwidth memory (HBM) for its AI accelerators.
The deal, announced during Jensen Huang’s visit to South Korea, gives SK Hynix a formal co‑development role. The partnership will deliver HBM that powers Nvidia’s upcoming Vera processor and future chips, aiming to ease the “toughest bottleneck” in the AI hardware stack.
Securing a dedicated HBM supply matters because memory bandwidth often limits AI training speed and cost. By locking in SK Hynix, Nvidia reduces reliance on spot market pricing and potential shortages, while SK Hynix gains a guaranteed revenue stream tied to a marquee customer.
The arrangement is a pragmatic response to a supply squeeze, not a marketing love‑in. Other chip makers will watch to see if similar deals become the new norm.