Overview
- The companies announced the deals during Jensen Huang’s visit to Seoul, with the main agreements unveiled on Monday and confirmed by Nvidia, SK hynix, SK Telecom, Naver and Doosan.
- Nvidia and SK hynix signed a multi‑year technology pact to co‑develop next‑generation high‑bandwidth memory such as HBM4 for Nvidia systems including the Vera Rubin accelerators.
- SK Telecom and Naver committed to phased hyperscale builds that start in 2027 — Naver beginning at 55 megawatts and targeting eventual gigawatt capacity while SK Telecom plans a multi‑gigawatt AI cloud program.
- Doosan, LG and Hyundai agreed to work with Nvidia on robotics, energy and ‘physical AI’ for factories and vehicles, and Nvidia reiterated that Vera Rubin production and HBM supplier qualifications are proceeding toward Q3 2026 deliveries.
- Jensen Huang gave an early cost benchmark of about $60 billion per one‑gigawatt AI factory, the deals did not disclose firm investment splits, and South Korea’s memory industry role was cited as central to securing supplies and sovereign AI capacity.