Nvidia Informs Customers of AI Server Price Hikes Exceeding 15% Starting Next Year Amid Surging Memory Chip Costs

en.Wedoany.com Reported - On August 22, due to surging memory chip costs, Nvidia has notified several of its largest customers, including Microsoft, Google, and Oracle, that prices for servers equipped with its AI chips will rise broadly, with increases exceeding 15% in most cases. The price hike will take effect for systems shipping early next year, affecting models powered by its flagship Vera Rubin and Grace Blackwell chips.

The direct cause of this price increase is the sustained tight supply of high-bandwidth memory (HBM) and server DRAM, which has driven up the manufacturing costs of AI servers. Nvidia had already raised prices by up to 15% on server chips such as the H200 and B200 in early 2026, and this further increase reflects the ongoing pressure of rising memory chip costs.

As the absolute dominant player in the global AI chip market, Nvidia is neither able to stabilize prices nor absorb the continuously rising costs on its own. This situation highlights that amid surging demand for AI infrastructure, memory chip manufacturers—Samsung Electronics, SK Hynix, and Micron Technology—have gained considerable bargaining power. According to South Korean media outlet Business Korea, Samsung Electronics and SK Hynix are experiencing strong demand and tight supply for HBM and server DRAM, which is conducive to maintaining high memory prices and profits.

Market analysts believe that this price increase may place additional pressure on AI data center construction plans. Following the news, Nasdaq 100 futures fell 0.2% in early trading on Monday, with concerns growing over runaway AI infrastructure costs.