Nvidia AI servers face price hikes tied to memory costs

Nvidia’s largest customers have been notified that servers containing its AI chips will rise in price by more than 15% in many cases from early next year, as memory costs climb. The increases apply to systems built around Vera Rubin and Grace Blackwell chips, with the final impact varying by chip generation and memory configuration.

The warnings have come through contract server assemblers that supply operators including Microsoft, Google and Oracle. Nvidia, which has a gross margin of about 75%, has not responded to requests for comment and is not absorbing the added memory expense, according to the reported customer notifications.

Pressure is concentrated in DRAM, where Samsung, SK hynix and Micron produce most global supply. Output is rising but has not caught up with demand from AI infrastructure. Other parts of the market are already seeing the effect, with Nvidia and AMD raising gaming card prices, and Amazon Web Services putting GPU prices up 20%.

European AI infrastructure plans face additional exposure. The EU has committed around €20bn to AI gigafactories, while a French consortium has bid $10bn for one site, with budgets built around last year’s hardware prices. Operators such as Nebius are expanding Nvidia capacity in the same market as data centre projects already contend with delays, labour shortages, tighter capital markets and local opposition.