Samsung Electronics unveiled its HBM4E memory solution at NVIDIA's GTC 2026, showcasing a breakthrough in AI computing. The new HBM4E delivers 16 Gbps per pin and a staggering 4.0 terabytes-per-second bandwidth, significantly exceeding previous industry standards. This advancement is crucial for next-generation AI applications requiring massive data throughput. Samsung's HBM4E, already in mass production, leverages a cutting-edge 10nm-class DRAM process, achieving stable yields and industry-leading performance. Visitors to Samsung's booth (#1207) will see this technology alongside the existing HBM4, which offers 11.7 Gbps speeds.

Samsung's commitment to innovation extends beyond HBM4E. The company demonstrated its hybrid copper bonding (HCB) technology, a novel method enabling 16 or more layers in HBM while reducing heat resistance by over 20% compared to thermal compression bonding. This addresses critical thermal management challenges in dense AI systems. Additionally, Samsung highlighted its SOCAMM2 server memory module, an industry-first mass-produced low-power DRAM solution, and the PM1763 SSD based on PCIe 6.0 for high-speed data transfers, both integral to NVIDIA's Vera Rubin platform.

The collaboration between Samsung and NVIDIA took center stage, particularly in the 'NVIDIA Gallery.' This partnership powers Samsung's AI Factory, accelerating digital twin manufacturing using NVIDIA accelerated computing. Yong Ho Song, Samsung's AI Center VP, will detail this strategic alliance in a March 17 session titled 'Transforming Semiconductor Manufacturing with Agentic AI.' Samsung's memory solutions also cater to personal AI devices, including the PM9E3/PM9E1 NAND for NVIDIA DGX Spark and LPDDR5X/LPDDR6 DRAM for premium smartphones and wearables, offering enhanced performance and efficiency.