SK Hynix has begun shipping 12‑stack HBM4E samples to key customers, advancing validation toward mass production. The new memory achieves 16Gbps per pin , over 20% better power efficiency , and lower latency compared to HBM4 – tailored for AI training and inference workloads.
The 12‑layer stack uses SK Hynix’s proprietary MR‑MUF (Mass Reflow Molded Underfill) technology, delivering 48GB capacity per stack and a 17% reduction in thermal resistance , improving stability under sustained high‑load AI compute.
With a mature production track record across HBM3, HBM3E, and HBM4, SK Hynix positions HBM4E as a critical enabler for next‑gen AI infrastructure.
ICgoodFind: SK Hynix’s 12‑stack HBM4E raises the bar for bandwidth and thermal performance, reinforcing its lead in AI memory.
