Nvidia to Raise Prices on Vera Rubin and Blackwell Server Systems by About 15–17%

Rising memory costs from tight high‑bandwidth memory (HBM) supply are forcing the company to pass large component price increases onto data‑center customers.

Overview

Multiple outlets reported Monday that Nvidia plans to raise server-system prices roughly 15 to 17 percent for units using Vera Rubin and Grace Blackwell GPUs that ship in early 2027.
The reported hikes are driven mainly by sharply higher memory costs as suppliers shift production toward HBM for AI workloads, tightening DRAM supply and lifting prices for server memory.
Contract server manufacturers have begun notifying major cloud customers, including Microsoft, Google and Oracle, about the forthcoming price adjustments.
Industry notes warn the increase could add billions to large data‑center projects — one report cited at least a $5 billion rise for a one‑gigawatt facility — and could speed hyperscalers' moves to buy custom chips or secure long‑term memory deals.
Nvidia has not publicly responded to requests for comment and the reports arrive ahead of the company's upcoming fiscal quarterly earnings, which could shape customer procurement choices.