SK hynix has taken a significant step forward in the high-bandwidth memory (HBM) sector by being the first to supply HBM4 modules to NVIDIA for testing. This achievement distinguishes the Korean company from its competitors, marking it as a leader in a rapidly evolving field that combines memory and logic elements in a novel way.
The competition between Korean tech giants has been intense. With SK hynix’s move, they’ve outpaced Micron and Samsung in securing NVIDIA’s orders for HBM4, which is integral to the development of the next-generation Rubin AI GPUs. According to a Korean media report, SK hynix has already begun delivering HBM4 to NVIDIA in limited quantities, positioning themselves at the forefront of this technological advancement. This early supply is set to help NVIDIA prepare for the market launch of Rubin GPUs by the last quarter of 2025.
Micron, despite being a strong competitor, appears to be trailing due to lower yield rates in HBM4 production and challenges with production line allocations. Samsung, too, shows promise but still has unknown yield rates for HBM4. Nonetheless, Samsung has recently secured deals with AMD for its HBM3E module, signaling potential for future collaborations with NVIDIA.
SK hynix’s success in securing a premium supplier status with NVIDIA is poised to significantly boost their market share in HBM technology. This development suggests that a substantial portion of NVIDIA’s orders will likely be filled by SK hynix. However, NVIDIA plans to implement a multi-sourcing strategy, potentially including Micron and Samsung as well.
With reliable partners like SK hynix, NVIDIA is paving the way for the Rubin GPU architecture to potentially debut in September 2025. This aligns with their goal to maintain a consistent product release schedule, ensuring they meet customer demands efficiently.
