Samsung Electronics has begun mass production and commercial shipment of Samsung HBM4, marking a major milestone in advanced memory technology. The launch positions the company at the forefront of high-bandwidth memory innovation designed to support the rapidly growing demands of AI computing.
Samsung HBM4 Sets New Performance Standards
Built on Samsung’s sixth-generation 10nm-class DRAM process (1c), HBM4 delivers stable yields and strong performance from the start of production without requiring additional redesigns.
Samsung adopted advanced nodes, including the 1c DRAM and 4nm logic process, to unlock greater performance potential. According to Sang Joon Hwang, Executive Vice President and Head of Memory Development at Samsung Electronics, this strategy enables the company to meet escalating customer demands for higher performance.
HBM4 achieves a processing speed of 11.7Gbps, approximately 46% higher than the industry standard of 8Gbps and a significant improvement over HBM3E. Performance can scale up to 13Gbps, helping reduce data bottlenecks as AI models grow increasingly complex.
Built for Efficiency and Higher Capacity
Samsung HBM4 also delivers major bandwidth gains, increasing total memory bandwidth per stack by 2.7x to reach up to 3.3TB/s.
Using 12-layer stacking technology, Samsung offers capacities ranging from 24GB to 36GB, with plans for 16-layer stacking that could expand capacity to 48GB in the future.
To address power and thermal challenges caused by increased data I/Os, Samsung integrated low-power design innovations into the core die. As a result, HBM4 provides:
-
40% better power efficiency through low-voltage TSV technology and PDN optimization
-
10% improved thermal resistance
-
30% enhanced heat dissipation compared to HBM3E
These improvements help data centers maximize GPU throughput while managing total cost of ownership.
Strengthening Production and Supply Chain
Samsung is advancing its HBM roadmap with extensive manufacturing resources and one of the industry’s largest DRAM production capacities. Its Design Technology Co-Optimization (DTCO) approach between Foundry and Memory divisions ensures high quality and yield while reducing production cycles.
The company is also expanding technical collaborations with global GPU manufacturers and hyperscalers developing next-generation ASICs.
Future Outlook for Samsung HBM4
Samsung expects its HBM sales to more than triple by 2026 and is actively scaling production capacity to meet rising demand. Looking ahead, the company plans to begin sampling HBM4E in the second half of 2026, followed by custom HBM samples in 2027.
With strong performance, efficiency, and scalability, Samsung HBM4 is set to play a critical role in powering future AI infrastructure and high-performance computing environments.
Read Also: BorgWarner Enters Data Center Market with Turbine Power Solution
