June 18, 2026 /SemiMedia/ — SK hynix said it has supplied samples of its 12-layer HBM4E to major customers, marking the next step in the company’s roadmap for high-performance DRAM used in artificial intelligence systems.

The company said the sample delivery reflects its accumulated experience in HBM development and mass production. SK hynix added that it will work closely with key customers to ensure the product enters mass production on schedule.

Compared with HBM4, the new HBM4E product delivers major improvements in performance and power efficiency. SK hynix said the device supports a pin speed of up to 16Gbps and improves energy efficiency by more than 20%, increasing data processing capability for AI training and inference workloads.

The product also uses a next-generation interface and design optimizations to reduce data transmission latency, helping maintain stable operation in high-bandwidth environments. SK hynix expects the technology to improve processing efficiency in next-generation AI data centers and large-scale computing systems.

The 12-layer HBM4E uses SK hynix’s advanced MR-MUF process to achieve 48GB capacity while improving structural stability. Compared with HBM4, thermal resistance has been reduced by about 17%, supporting more stable operation in high-performance computing environments.

MR-MUF, or mass reflow molded underfill, is a packaging process in which liquid protective material is filled between stacked semiconductor chips and then cured to protect inter-chip circuitry and improve package reliability.

SK hynix has built mass production and supply experience across HBM3, HBM3E and HBM4 products. As AI chip platforms demand higher memory bandwidth, larger capacity, better power efficiency and improved thermal performance, HBM4E is expected to become an important technology for future AI accelerators and data center systems.

SK hynix Chief Development Officer Ahn Hyun said the company will extend its technology leadership and mass production capability to HBM4E, while strengthening cooperation with partners to support continued AI innovation.