Counterpoint Research: NVIDIA Corporation (NVDA.US) plans to adopt smartphone-style memory, potentially doubling server memory prices by the end of next year.

Counterpoint Research report states that NVIDIA Corporation (NVDA.US) plans to use smartphone-style memory chips in its AI servers, which could lead to a doubling of server memory prices by the end of 2026. The report adds that over the past two months, the global electronics supply chain has been affected by a shortage of traditional memory chips, as manufacturers have shifted their focus to high-end memory chips tailored for AI applications. NVIDIA Corporation is scheduled to release its third-quarter financial report after the market closes on November 19. Counterpoint points out that a new issue is brewing. NVIDIA Corporation recently decided to switch to a different type of memory chips to reduce the power consumption costs of AI servers, replacing the commonly used DDR5 (fifth-generation double data rate memory) with LPDDR (low-power double data rate memory). The report mentions that due to the significantly higher number of memory chips required for each AI server compared to smartphones, this change is expected to create a sudden demand that the industry may struggle to meet. The report also states that memory suppliers such as Samsung Electronics, SK Hynix, and Micron Technology, Inc. (MU.US) have previously reduced production capacity for high-bandwidth memory (HBM). As high-bandwidth memory is a crucial component for manufacturing advanced AI accelerators, the shortage of these manufacturers' older dynamic random-access memory products has occurred. The report indicates that as chip manufacturers consider reallocating more factory capacity to produce LPDDR to meet NVIDIA Corporation's demand, the tight supply in the low-end market may spread upwards. "The future greater risk lies in the high-end memory sector. NVIDIA Corporation's recent shift to LPDDR means it will become a customer on a scale comparable to large smartphone manufacturers - a major shift for the supply chain that may struggle to easily digest such a scale of demand," Counterpoint states. The research company forecasts that by the end of 2026, server memory chip prices will double. The increase in server memory prices may increase costs for cloud service providers and AI developers. Currently, data center budgets are already high due to record spending on graphics processing units (GPUs) and power upgrades, which may further exacerbate budget pressures.