Post by : Bianca
Haleem
Nvidia’s recent transition in its AI server architecture has triggered fresh worries within the global electronics supply chain. The company’s choice to utilize LPDDR — the low-power memory typically used in smartphones and tablets — is anticipated to rapidly escalate server-memory prices by late 2026, as indicated by a recent report from Counterpoint.
In recent months, manufacturers globally have faced increasing difficulties due to a dwindling supply of legacy memory chips. Many chipmakers have turned their attention towards high-end memory designed for advanced AI semiconductors, resulting in a scarcity of older models. Now, Nvidia’s shift to LPDDR adds an unpredicted layer of tension.
The objective is to lower power consumption in Nvidia’s AI servers by swapping out DDR5 memory for LPDDR. However, each server necessitates a higher number of chips than a smartphone, resulting in a sudden increase in demand that existing production capacities cannot fulfill. Counterpoint highlighted that the scale of Nvidia’s requirement resembles that of a leading smartphone manufacturer, potentially altering memory suppliers’ factory capacity strategies.
Samsung Electronics, SK Hynix, and Micron are already encountering tight supply conditions after scaling back production of older dynamic random-access memory to focus on high-bandwidth memory essential for AI accelerators. The growing interest in LPDDR may exacerbate these supply constraints.
Should chipmakers prioritize capacity to support Nvidia’s requirements, shortages could expand, leading to a steep rise in prices. Counterpoint projects that server-memory prices might potentially double by the end of 2026, adding new challenges for cloud providers and AI developers already contending with soaring GPU and power-infrastructure expenses.
Nvidia is anticipated to unveil its earnings report later today, as investors keenly await insights into how this supply chain adjustment might affect its future product direction.
English
