Samsung to Begin Mass Production of HBM4 as Early as Next Week, Projects Over Threefold Increase in HBM Sales This Year

According to South Korean media reports, Samsung Electronics has decided to commence mass production of HBM4 as early as the third week of this month. These products will be used in NVIDIA’s next-generation AI computing platform, "Vera Rubin." Having already passed NVIDIA's certification tests, Samsung Electronics finalized the mass production plan after considering the launch schedule of "Vera Rubin." Industry analysts expect that NVIDIA may unveil products featuring Samsung’s HBM4 for the first time at its annual developers conference (GTC) next month.

Samsung’s HBM4 utilizes 1c DRAM technology and a 4nm process, achieving a data processing speed that exceeds the JEDEC standard of 8 Gbps, reaching up to 11.7 Gbps — a 22% improvement over the previous generation HBM3E (9.6 Gbps). The single-layer stacked memory bandwidth also reaches as high as 3 TB/s, representing a 2.4-fold increase compared to the previous generation. With 12-layer stacking technology, it can offer a maximum capacity of 36 GB. Future expansion to 16-layer stacking is expected to increase capacity to 48 GB. Another major advantage lies in its low-power design, which supports high-performance computing while significantly reducing power consumption and cooling costs for servers and data centers.

Samsung Electronics anticipates that its HBM sales this year will grow more than threefold compared to last year. To support this growth, Samsung is constructing a new 1c DRAM production line at its Pyeongtaek P4 plant. The production line is expected to be completed in the first quarter of next year, with a projected monthly capacity of 100,000 to 120,000 wafers.