Samsung Allocates Over 60% of V-NAND Capacity to Nvidia for AI Storage
Samsung is reportedly allocating over 60% of its V-NAND production capacity to Nvidia to support its Context Memory Storage (CMX) technology. Additionally, Samsung is accelerating its V9 and V10 flash memory roadmaps, with V9 yield rates already exceeding 80%. This partnership addresses the critical Key-Value (KV) cache bottleneck in AI servers, which is growing rapidly due to long-context and agentic AI workloads. It solidifies Samsung's position as a key hardware supplier in Nvidia's expanding AI infrastructure ecosystem. Nvidia's CMX platform currently utilizes 576 SSDs to achieve 9,600TB of storage capacity, with demand projected to surge to over 100 million TB by 2027. Samsung is also developing V11 NAND with up to 500 layers to enable even higher density SSDs for future CMX modules.
## BACKGROUND
V-NAND (Vertical NAND) is a flash memory architecture that stacks memory cells vertically to increase storage density and capacity within the same chip area. In AI inference, the KV cache stores attention keys and values for tokens in a context window, but its linear growth creates a massive memory bottleneck that technologies like Nvidia's CMX aim to mitigate.