Samsung Says Nvidia Partnership Extends Beyond HBM to Foundry Collaboration - thelec.net

Samsung Electronics has extended its strategic partnership with NVIDIA beyond high bandwidth memory (HBM) into foundry collaboration, signaling a long-term vision in the semiconductor industry. At a p... Read original →

Samsung’s deepening foundry collaboration with NVIDIA marks a strategic shift from transactional HBM supply to co-engineered AI silicon stacks. Technically, joint development on 4nm/8nm nodes accelerates SOCAMM adoption and pressures SK hynix to fast-track CoWoS-compatible HBM packaging. Geopolitically, this alliance mitigates U.S.-Korea export control risks but raises IP leakage concerns for third parties like Groq. TSMC will likely counter with tighter CoWoS-HBM integration, while SK hynix may pivot toward Intel and foundries in Taiwan, China to diversify dependencies. Within 18 months, Samsung’s HBM4E yield rates will determine its leverage in NVIDIA’s AI training chip allocation; if extended to sub-3nm nodes, this partnership could directly erode TSMC’s dominance in AI SoCs.