At NVIDIA GTC 2026 on March 17, Samsung Electronics showcased its comprehensive AI computing portfolio, highlighting its position as the only semiconductor company offering a complete solution spanning memory, logic, foundry, and advanced packaging.

HBM4E takes center stage—Samsung's next-gen memory delivers 16 Gbps per pin speed with 4.0 TB/s bandwidth, designed for next-generation AI accelerators and hyperscale data centers. The company also introduced Hybrid Copper Bonding (HCB) technology enabling 16+ layer HBM stacks with 20% lower thermal resistance than thermal compression bonding.

Samsung's sixth-generation HBM4 memory has entered mass production for NVIDIA's Vera Rubin platform, achieving 11.7 Gbps processing speed (exceeding the 8 Gbps industry standard) with potential to reach 13 Gbps.

In personal devices, Samsung showcased LPDDR5X (25 Gbps per pin, 15% lower power) and LPDDR6 (30-35 Gbps per pin) with adaptive voltage scaling and dynamic refresh control.

The company also unveiled PM1763 SSD with PCIe 6.0 interface, delivering 2x performance and 1.6x better efficiency at 25W power limit, featuring completed SI signal integrity analysis.

Samsung's "NVIDIA Gallery" at the event highlighted collaborations including SOCAMM2 server modules (industry's first production-ready) and PM1753 SSDs for inference workloads.

ICgoodFind: Samsung's GTC showcase confirms its technological leadership, driving next-gen AI computing across memory, storage, and advanced packaging