SANTA CLARA, CALIFORNIA / RankWire.AI / – Nvidia is set to increase the prices of many AI server configurations by more than 15% starting in early 2027, affecting systems built on the company’s Vera Rubin and Grace Blackwell platforms. The extent of each price increase varies depending on the chip generation and memory configuration, with Nvidia not announcing a blanket price hike for all server products publicly. Nonetheless, server manufacturers have communicated these new pricing adjustments to key data center clients, including major corporations.
Major data center operators such as Microsoft, Google, and Oracle have been informed of these price changes by server manufacturers, as they are significant buyers of AI infrastructure for cloud and data center applications. The rising costs of memory modules are a primary factor contributing to the overall price adjustments of these servers. AI workloads demand large quantities of high-bandwidth memory, which, along with GPUs, CPUs, networking gear, and storage, constitutes a substantial portion of server costs. The demand for AI-related deployments has maintained tight supplies across several memory categories throughout 2026.
According to TrendForce, conventional DRAM contract prices are projected to increase between 13% and 18% in the third quarter of 2026. Similarly, NAND Flash contract prices are expected to rise by 10% to 15% over the same period. The research firm further noted that server DRAM remains in short supply as suppliers prioritize capacity for AI applications. Although long-term supply agreements have helped limit some price increases, the overall cost of server memory continues to stay elevated.
Memory Costs Continue to Climb Amid Sustained Demand for AI Servers
In May, Nvidia announced that Vera Rubin had reached full production with server manufacturers and supply-chain partners around the world, with partner products utilizing Rubin expected to be available during the latter half of 2026. Vera Rubin integrates the Vera CPU, Rubin GPU, NVLink 6, ConnectX-9, BlueField-4, and Spectrum-6 networking technologies, serving as Nvidia’s successor to Grace Blackwell as its newest rack-scale architecture designed for intensive AI workloads. System builders are developing Rubin-based solutions for cloud providers, AI laboratories, and hyperscale data centers.
Meanwhile, Grace Blackwell remains a key component of Nvidia’s current data center offerings. The GB200 NVL72 configuration connects 36 Grace CPUs with 72 Blackwell GPUs in a single liquid-cooled rack, engineered to operate as a unified NVLink computing domain tailored for large-scale AI models. The reported increases in server prices differ depending on configurations, with memory capacity and chip generation playing significant roles in pricing. Consequently, the 15% figure does not reflect a uniform hike across all Nvidia server products.
Expansion of Vera Rubin Production Across Nvidia’s Server Ecosystem
As AI demand continues to absorb capacity, memory suppliers have shifted production toward high-performance and server-grade products, which TrendForce states has reduced the supply available for certain PC and consumer DRAM modules. While server shipments have remained robust through 2026, buyers have continued stockpiling data center memory, and the market analyst predicts that server DRAM supply will remain tight into 2027 as demand growth surpasses new supply. These market conditions underpin the cost increases reported by Nvidia.
Nvidia begins this pricing phase with record sales in its data center segment, having posted total revenue of $81.6 billion in its fiscal first quarter ending April 26, with data center revenue reaching an all-time high of $75.2 billion, reflecting a 92% increase year-over-year. The company forecasted second-quarter revenue at around $91 billion, plus or minus 2%, when it reported those results in May, with its fiscal second-quarter earnings to be announced on August 26.