Samsung Supplies Over 60% of V-NAND Capacity to Nvidia

Samsung Electronics is significantly increasing its supply of V-NAND flash memory to Nvidia, allocating more than 60% of its production capacity to the U.S. technology firm. The expanded collaboration, as reported by South Korean media sedaily, is driven by the escalating needs of artificial intelligence infrastructure.

The surge in AI server deployment necessitates larger capacities for temporary context data, often referred to as Key-Value cache. Nvidia's Context Memory Storage (CMX) solution aims to alleviate this bottleneck by incorporating substantial amounts of NAND flash. Each Nvidia server can be equipped with up to 576 SSDs, totaling 9600 terabytes of storage, powered by Samsung's V-NAND technology.

Industry analysts project a substantial rise in demand for NAND flash for CMX applications. Demand is expected to reach 35 million terabytes by 2026 and exceed 100 million terabytes in 2027. With an estimated annual NAND production of 250 million terabytes, Samsung is positioned as a primary supplier for Nvidia's ambitious AI initiatives.

Samsung is advancing its V-NAND technology roadmap. Its V9 generation, featuring 286 layers, currently accounts for 60% of its V-NAND output, boasting a yield rate above 80%. The company is also developing the V10 and V11 generations, with V11 potentially reaching 500 stacked layers, promising further increases in storage density and capacity to meet future demands.