Huang's Korea visit is less a diplomatic courtesy than a supply chain necessity. Vera Rubin in full production means HBM4 must scale in lockstep — and both Samsung and SK Hynix are headquartered in Seoul. By announcing the Korea R&D center at this exact moment, NVIDIA is embedding itself institutionally into the ecosystem that controls the memory bandwidth Vera Rubin requires. Early adopters named at GTC Taipei — Anthropic, OpenAI, and SpaceXAI — are counting on that supply chain holding through a fall delivery window that gives NVIDIA no margin for component delays.
Loading story