ZDNet reports Samsung may use 2nm node for NVIDIA HBM base dies
This digest was compiled by AI from multiple sources — links to the originals are below.
ZDNet reports Samsung may repurpose a production line at its under-construction R&D facility in Giheung to make HBM base dies using a 2nm process. The plans could be aimed at NVIDIA's AI chips and remain subject to change, with the facility expected to open in about two years.
Reported 2nm Plans
ZDNet reported, citing industry sources, that Samsung may repurpose the NRD-K Line 2 at its under-construction R&D facility in Giheung to manufacture HBM base dies using a 2nm process. The base dies could be destined for NVIDIA's AI chips. Since the facility's opening is roughly two years away, the plans remain subject to revision. Samsung earlier announced it would use 4nm technology for HBM4 memory chips.
HBM Base Die Evolution
HBM memory chips stack multiple DRAM chips and require a base die. The shift to HBM3 and HBM3E saw manufacturers adopt logic nodes ranging from 20nm to 12nm for the base die. HBM4, used in NVIDIA's Rubin chips, introduced wider buses and more through-silicon vias. TSMC and Samsung Foundry then took over manufacturing of the HBM base die.
Foundry Partnerships
SK hynix has teamed up with TSMC for its HBM chips, with TSMC offering mature N12 and advanced N5 node options. Micron has also partnered with TSMC for HBM4 memory chips. Samsung earlier said it would use 4nm manufacturing for HBM4, while ZDNet reports a possible 2nm node for base dies.
What's Next
Samsung has not confirmed the report, and the Giheung facility is not expected to open for about two years. It remains unclear whether the potential 2nm base die plan would replace or extend Samsung's announced 4nm HBM4 roadmap.
1 source
ZDNet reports Samsung may use 2nm node for NVIDIA HBM base dies
