Samsung Debuts HBM4E at GTC 2026, Unveils Full AI Memory Lineup

At the NVIDIA GTC 2026 conference, Samsung comprehensively showcased its technological lineup in the AI computing field, exhibiting a full range of products from high-performance memory to low-power solutions for personal devices. 

The highlight of Samsung's exhibition at the conference was its sixth-generation HBM4 memory, which is now in mass production and specifically designed for NVIDIA Vera Rubin platform. It is said that the HBM4 delivers a stable processing speed of 11.7 Gbps, surpassing the industry standard of 8 Gbps, and can be further enhanced up to 13 Gbps. By utilizing the most advanced sixth-generation 10nm-class DRAM process technology, Samsung has achieved stable yield rates and leading performance.

Samsung unveiled its next-generation HBM4E memory for the first time. This product achieves a per-pin transfer speed of up to 16Gbps and a bandwidth of up to 4.0 TB/s. Additionally, Samsung demonstrated its Hybrid Copper Bonding (HCB) technology. This new technology will enable next-generation HBM to achieve 16 or more stacked layers while reducing thermal resistance by over 20% compared to thermocompression bonding.

In the personal devices sector, Samsung demonstrated memory solutions optimized for on-device AI workloads, including the Samsung PM9E3 and PM9E1 NAND products designed for the NVIDIA DGX Spark.

Samsung also exhibited its LPDDR5X and LPDDR6 DRAM solutions tailored for premium smartphones, tablets, and wearables. Among these, the LPDDR5X achieves a per-pin speed of up to 25 Gbps while reducing power consumption by 15%. Based on this, LPDDR6 further increases bandwidth to 30-35 Gbps per pin and introduces advanced power management features such as adaptive voltage scaling and dynamic refresh control, providing performance support for next-generation on-device AI workloads.