Samsung is reportedly developing custom HBM logic dies using 2 nm foundry nodes, according to a ZDNet South Korea report. The company's System LSI division is tailoring these dies for specific customer needs, with a portfolio spanning from 4 nm to 2 nm. This move targets the next generation of AI accelerators requiring extreme memory bandwidth.

Samsung's current sixth-gen HBM4 memory is built on a 4 nm process. The new 2 nm logic dies are expected to arrive after HBM4E (seventh gen), possibly post-2027. An insider noted the custom SoC team is leading the design to meet strong enterprise AI demand, which is anticipated to surge with these advanced modules.

The shift to 2 nm is a significant step in the memory industry's race to support AI workloads. Samsung's custom approach could give it an edge over competitors like SK Hynix and Micron. For consumers and data centers, this means future AI hardware will gain more efficient, high-speed memory access, though widespread availability is still years away.