Seoul Economic Daily reports that Nvidia has asked Samsung, SK hynix and Micron to build the base die of its NVHBM memory on TSMC's 3nm process. Samsung's Device Solutions division reportedly agreed after an internal review. The report cites industry sources and is not an official announcement.
NVHBM, unveiled in August and based on HBM4E, moves the memory controller into the base die. Nvidia claims up to 30% more bandwidth, 15% lower power and 25% more compute die area than standard HBM4E. A single process keeps one Nvidia design consistent across suppliers. Industry estimates put TSMC's 3nm yields above 90%, against roughly 60% for Samsung's 2nm node.
Samsung loses its turnkey pitch for this custom part, but keeps its own 4nm base dies for standard HBM4/HBM4E and custom HBM for customers such as AMD. It is reportedly validating a 2nm backup. The base die is now real logic, so NVHBM will compete with AI accelerators and flagship phone chips for the same 3nm wafers.