Overview

• SK hynix says its next-gen HBM4 doubles the I/O width to 2,048 bits, delivers speeds above 10 Gbps, and doubles bandwidth versus HBM3E.
• The company claims more than 40% better power efficiency and up to 69% improvement in AI service performance compared with the prior generation.
• Production is set at the Incheon site using Advanced MR-MUF packaging and a mature 1b‑nm DRAM node to support stability at scale.
• Customer qualification is underway, with industry reporting expecting adoption in Nvidia’s Rubin platform and rivals AMD and others targeting 2026 accelerators.
• Shares rose about 7% on the news as Samsung and Micron advance their own HBM4 efforts; SK hynix has not disclosed stack layer counts or capacities while exceeding JEDEC’s 8 Gbps spec.