Samsung Electronics Unveils HBM4E for the First Time at GTC... Advances AI Alliance with Nvidia

HBM4E Based on 1c DRAM and 4nm Foundry Technology

The Only Total Memory Solution Provider for the Vera Rubin Platform

Samsung Electronics will unveil its seventh-generation high-bandwidth memory, HBM4E, for the first time at NVIDIA GTC, which will be held from March 16 to 19 (local time) in San Jose, United States. The company aims to strengthen its strategic partnership with NVIDIA by emphasizing its identity as a comprehensive semiconductor company (IDM), spanning memory, logic design, foundry (semiconductor contract manufacturing), and advanced packaging.

Samsung Electronics has organized its exhibition space into three zones: ▲AI Factory ▲On-Device AI ▲Physical AI, where it will introduce its next-generation memory architectures that have been officially recognized for their technological excellence, such as GDDR7, LPDDR6, and PM9E1.

On the second day of the event, March 17, Yongho Song, Head of Samsung Electronics AI Center, will give a presentation at the special invitation of NVIDIA, highlighting the importance of NVIDIA's next-generation systems that will drive innovation in artificial intelligence (AI) infrastructure and presenting Samsung's vision for total memory solutions to support them.

Samsung Electronics has set up the "HBM4 Hero Wall" in its exhibition to spotlight its leadership in HBM technology right from the entrance path.

Leveraging the technological competitiveness accumulated through mass production of HBM4 based on its 1c DRAM process and the 4-nanometer base die design capabilities of Samsung Foundry, Samsung Electronics has unveiled its HBM4E physical chip and core die wafer for the first time. These are expected to support a speed of 16Gbps per pin and a bandwidth of 4.0TB/s.

In addition, the company showcased its HCB (Hybrid Bonding) technology, which improves thermal resistance by more than 20% compared to TCB (Thermal Compression Bonding) and supports ultra-high stacking of more than 16 layers, through a video presentation, highlighting Samsung's advanced packaging competitiveness for next-generation HBM.

Notably, Samsung Electronics used this exhibition to emphasize that it is the only company in the world capable of timely supplying all memory and storage for the NVIDIA Vera Rubin platform, underscoring its total memory solution capabilities.

Samsung Electronics also set up a dedicated "NVIDIA Gallery" to showcase ▲HBM4 for the Rubin graphics processing unit (GPU) ▲SOCAMM2 for the Vera central processing unit (CPU) ▲PM1763 for storage, highlighting the collaboration between the two companies. Samsung Electronics' SOCAMM2 is a server memory module based on LPDDR, which has completed quality verification and has begun mass production and shipment for the first time in the industry.

The Samsung Electronics PCIe Gen6-based server SSD PM1763 serves as the main storage for the Vera Rubin platform. At the booth, Samsung Electronics demonstrated NVIDIA's SCADA workload on servers equipped with the PM1763, allowing attendees to experience the industry's top-level performance firsthand, beyond just specifications. Additionally, to improve inference performance and power efficiency, Samsung Electronics plans to supply its PCIe Gen5-based server SSD PM1753 to the newly adopted CMX platform in the Vera Rubin platform.

Samsung Electronics stated, "As powerful AI systems like the Vera Rubin platform are essential for AI Factory innovation, Samsung will continue to supply high-performance memory solutions to support them," and emphasized, "Based on this cooperation, the two companies will jointly drive a paradigm shift in the global AI infrastructure."

Jensen Huang, CEO of NVIDIA, along with Jay Y. Lee, Chairman of Samsung Electronics, is greeting at the 'GeForce Gamer Festival' held on the night of the 30th at COEX Plaza in Gangnam-gu, Seoul, to celebrate the 25th anniversary of GeForce Korea. 2025.10.30 Photo by Jo Yongjun

This content was produced with the assistance of AI translation services.