Samsung Electronics began shipping the first 12‑layer HBM4E samples to global customers, extending its HBM roadmap after mass‑producing HBM4 earlier this year. The new stack runs at a stable 14 Gbps pin speed, scalable to 16 Gbps, delivering up to 3.6 TB/s per stack. This jump exceeds HBM4 performance by more than 20 %. The samples target hyperscale AI servers, showing Samsung’s push for premium HBM market.

The 12‑layer HBM4E ships with 48 GB capacity, a 30 % increase over its predecessor, and Samsung plans 32 GB (8‑layer) and 64 GB (16‑layer) variants to match customer needs. Built on Samsung’s 6th‑gen 10 nm DRAM process and a 4 nm logic base die, the design improves power efficiency by 16 % and cuts thermal resistance over 14 % while maintaining yield and reliability.

Executive Vice President Sang Joon Hwang says Samsung’s advanced manufacturing and early infrastructure investments let it keep pace with AI‑driven demand. Positive feedback on HBM4’s 11.7 Gbps SiP tests has already secured a stable supply, and the company aims to begin mass production of HBM4E as customers finalize designs. The rollout gives data‑center builders a higher‑bandwidth, lower‑power option for large language models.