Nvidia server prices reportedly increase by 15%, with some customers already receiving price hike notifications

Some of NVIDIA’s largest customers have been informed that, due to skyrocketing memory chip costs, the price of servers equipped with NVIDIA AI chips will increase by more than 15% in most cases.

On August 22, Bloomberg cited sources saying that this price hike will take effect for systems shipped from early next year, impacting systems featuring flagship Vera Rubin and Grace Blackwell chips.

Factories manufacturing servers for major companies such as Microsoft, Google, and Oracle have already issued price increase notifications. The specific price adjustment will depend on the generation of NVIDIA chips and memory configurations.

The price rise will not only drive up the cost of building AI data centers, but also adds a new market focus for NVIDIA’s upcoming Q2 earnings announcement scheduled for August 26.

The fundamental driver behind this price surge is the structural shortage of DRAM memory supply.

The performance of NVIDIA’s AI accelerator processors relies heavily on the amount of DRAM (dynamic random access memory) provided, and the global DRAM production capacity is almost entirely divided among Samsung, SK hynix, and Micron.

Although these three companies continue to expand production, the growth rate still trails far behind the explosive demand for AI infrastructure, causing prices for this commodity chip to soar.

According to Bloomberg News, memory chip suppliers have gained unprecedented bargaining power from the supply-demand imbalance.

NVIDIA’s inability to absorb rising costs or hold prices steady precisely underscores the strong position of Samsung, SK hynix, and Micron in the current AI wave. Recently, both Apple and Qualcomm said they have been forced to raise product prices due to chip shortages.

NVIDIA is one of the most profitable companies in the semiconductor industry, with a gross margin as high as 75%.

Relying on the production constraints at TSMC, prices for its AI accelerator chips can easily reach tens of thousands of dollars.

Originally derived from PC gaming graphics cards, this product’s price has soared due to persistent strong demand and a lack of viable competitors. NVIDIA recently also increased the price of its gaming-oriented PC graphics cards.

The reactions of major clients such as Amazon, Microsoft, Google, and Meta to this round of price hikes will be crucial.

All these companies are pursuing their own chip development plans to reduce dependence on NVIDIA, but for now, data center construction remains highly reliant on NVIDIA purchases.

According to Bloomberg News, whether these clients can achieve greater independence also depends on whether they can secure sufficient memory from Samsung, SK hynix, and Micron.

Rising prices will further increase the complexity of large-scale AI data center expansion. Project delays, labor shortages, tightening capital markets, and community resistance have already disrupted numerous construction plans; this latest surge in cost pressures adds further complications.

As the world’s most valuable publicly listed company, NVIDIA will announce quarterly earnings next week, and its latest developments remain a crucial benchmark for the tech industry and the many investors backing AI infrastructure.