SK hynix ships HBM4E samples, joining Samsung in next-gen AI memory race
SK hynix announced it has shipped samples of its 12-layer HBM4E memory for AI applications to major customers, entering the validation phase three weeks after Samsung Electronics made a similar claim. The new memory delivers 48GB capacity, 16 Gbps speeds, and improved energy efficiency and thermal performance compared to HBM4, positioning SK hynix as a key supplier for next-gen AI accelerators like Nvidia’s Rubin Ultra platform.
SK hynix announced on Thursday it has begun shipping samples of its 12-layer HBM4E memory to major customers, marking a shift from product roadmaps to customer validation for AI chipmakers. This move follows Samsung Electronics’ May 29 claim of shipping the world’s first HBM4E samples, though SK hynix did not claim a first-mover advantage. Instead, the company emphasized its proven track record in HBM3, HBM3E, and HBM4, alongside its 58% market share in high-bandwidth memory, according to Counterpoint Research. The HBM4E samples deliver 48GB capacity and pin speeds of up to 16 gigabits per second, with over 20% improved energy efficiency compared to HBM4. SK hynix used its Advanced MR-MUF packaging process, which reduces thermal resistance by 17% and enhances structural stability and heat dissipation. These improvements address critical challenges in AI training and inference, where faster memory generates more heat in dense computing environments. The competition between SK hynix and Samsung is significant, as HBM4E is expected to support Nvidia’s upcoming Rubin Ultra accelerator platform, slated for 2025. Sample validation is a critical step for AI chipmakers to integrate memory into their designs before mass production. SK hynix did not disclose specific customers or a mass-production timeline but stated it will work with partners to achieve production in a timely manner. Ahn Hyun, SK hynix’s president and chief development officer, highlighted the company’s goal to strengthen its AI leadership with HBM4E, positioning itself as a full-stack AI memory provider. The move underscores the intensifying race in next-generation memory solutions for AI workloads, where efficiency and performance are key differentiators. SK hynix’s entry into the validation phase follows its established dominance in the HBM market, trailing only Samsung and Micron in next-gen production efforts. The company’s focus on supply capability and thermal management aligns with growing demands for high-performance memory in AI infrastructure.
This content was automatically generated and/or translated by AI. It may contain inaccuracies. Please refer to the original sources for verification.