Nvidia chip shift to smartphone-style memory to double server-memory prices by end-2026 - Counterpoint

Nvidia's recent strategy to incorporate smartphone-style memory chips into its AI server architecture may lead to a significant surge in server-memory prices, potentially doubling them by the end of 2026. This alarming forecast comes from Counterpoint Research, which released its findings on Wednesday. In recent months, the global electronics supply chain has faced challenges due to a shortage of traditional memory chips. Manufacturers have been increasingly focusing their production efforts on high-performance memory chips suitable for artificial intelligence applications. However, Counterpoint warns that a new issue is emerging on the horizon. Nvidia has opted to transition its AI server memory from DDR5 chips, typically used in server environments, to LPDDR chips, which are low-power variants more commonly found in smartphones and tablets. This change is anticipated to create an unprecedented demand in the market, one that current supply chains may struggle to meet. As each AI server requires a greater number of memory chips than a standard mobile device, this shift could lead to a sudden spike in demand for LPDDR chips. Major memory manufacturers like Samsung Electronics, SK Hynix, and Micron are already grappling with supply shortages of older dynamic random-access memory products due to their strategic pivot towards high-bandwidth memory essential for AI advancements. Counterpoint highlighted that the constraints in the lower end of the memory market could soon ripple upwards, as chipmakers consider reallocating their production capacities to satisfy Nvidia's demands. The firm noted, “Nvidia’s pivot to LPDDR effectively positions them as a customer on par with major smartphone manufacturers—a transformative change for the supply chain that may struggle to accommodate this level of demand.” With the anticipated rise in server-memory prices, cloud service providers and AI developers may face increased operational costs, adding further strain to budgets that are already burdened by escalating investments in graphics processing units and power enhancements.