Samsung Electronics unveiled a new generation of artificial intelligence (AI) memory technology on Tuesday. The world’s largest memory chip manufacturer presented its V10 bonding V-NAND or BV-NAND prototype at the ‘Future of Memory and Storage’ (FMS) conference held in Santa Clara, California. This chip features over 400 layers and a new wafer-bonding architecture, resulting in increased storage density and improved performance.
The company stated that its BV-NAND technology increases memory density by approximately 58% compared to the previous V9 NAND. Additionally, it enhances read, write, and input/output performance to meet the growing demand for AI systems requiring higher capacity and more power-efficient storage. As the scope of AI has expanded from training large language models to answering user queries, the demand for high-capacity NAND memory has also risen. NAND flash memory is used in devices like smartphones to store data such as photos, videos, and applications.
Samsung also showcased concept models for the industry’s first zHBM and zNAND-O architectures. Through these, the company presented its vision for next-generation 3D memory, in which memory cells are stacked vertically to store more data.
Typically, High-Bandwidth Memory (HBM) is integrated with AI processors using advanced packaging technology; however, Samsung’s zHBM concept proposes vertically stacking memory atop the AI accelerator to reduce data-transfer distances, thereby potentially improving bandwidth and energy efficiency.
Samsung stated that its wafer-bonding technology is expected to deliver more than 10 times the memory density of conventional HBM5 designs, along with a threefold improvement in energy efficiency and a reduction of over 50% in thermal resistance.
Samsung has expressed confidence in sustained AI-driven demand, stating that long-term customer agreements could account for 60% to 70% of its future memory sales. Meanwhile, analysts at Citi note that recent supply-chain data indicates DRAM and NAND inventory levels remain significantly lower than historical norms. This suggests that memory supplies remain tight, despite concerns regarding weak demand in certain end markets.
