Nvidia Corp is raising server prices by more than 15% for customers due to a sharp increase in memory costs, according to industry reports. The price adjustments, expected to apply to systems shipped in early 2025, vary by chip generation and memory configuration. Nvidia has also raised prices for its gaming-oriented PC graphics cards, Bloomberg reported.
The increases come as Nvidia’s flagship AI accelerator chips, including the Vera Rubin and Grace Blackwell processors, rely heavily on DRAM capacity. The company charges tens of thousands of dollars per chip, with gross margins reported at 75%. Contract server builders have notified major data center operators—including Microsoft, Google, Oracle, Amazon, and Meta—of the upcoming cost hikes through private communications.
Memory chipmakers Samsung Electronics, SK Hynix, and Micron dominate global DRAM production and have gained significant pricing power amid surging AI infrastructure demand. While cloud providers are developing in-house chip programs, they remain dependent on Nvidia for data center expansions due to limited alternative supply sources.
The price increases add to challenges facing AI data center build-outs, which are already constrained by project delays, labor shortages, tighter capital markets, and local opposition. Apple and Qualcomm have also raised prices amid ongoing chip shortages, underscoring broader supply chain pressures in the semiconductor sector.
