SANTA CLARA, CALIFORNIA / RankWire.AI / – Nvidia plans to raise prices by more than 15% for many AI server configurations shipping in early 2027. The increases cover configurations using the company’s Vera Rubin and Grace Blackwell platforms. The size of each increase varies by chip generation and memory setup. Nvidia has not publicly announced a universal price increase across all server products. Server manufacturers have also relayed the new pricing information to major data center customers.
Server manufacturers have relayed the pricing information to major data center operators, including Microsoft, Google and Oracle. Those companies buy large volumes of AI infrastructure for cloud and data center services. Rising memory costs have become a major input in the reported server price changes. AI systems use substantial amounts of high-bandwidth and server memory alongside GPUs, CPUs, networking equipment and storage. Demand from AI deployments has kept several memory categories tight during 2026.
TrendForce forecast conventional DRAM contract prices would rise 13% to 18% in the third quarter of 2026. It also projected NAND Flash contract prices to increase 10% to 15% during the same period. The research firm said server DRAM remained undersupplied as suppliers prioritized capacity for AI-related applications. Long-term supply agreements have limited some price increases, but overall server memory costs remain elevated.
Memory prices rise as AI server demand stays high
Nvidia said in May that Vera Rubin had entered full production with server makers and supply-chain partners worldwide. The company said partner products using Rubin would become available during the second half of 2026. Vera Rubin combines the Vera CPU, Rubin GPU, NVLink 6, ConnectX-9, BlueField-4 and Spectrum-6 networking technologies. The platform succeeds Grace Blackwell as Nvidia’s newest rack-scale architecture for large AI workloads. System makers are building Rubin products for cloud providers, AI labs and hyperscale data centers.
Grace Blackwell remains a major part of Nvidia’s current data center lineup. The GB200 NVL72 design connects 36 Grace CPUs with 72 Blackwell GPUs in one liquid-cooled rack. Nvidia built the system to operate as a single NVLink computing domain for large AI models. The reported server price increases vary by configuration, making memory capacity and chip generation important pricing factors. The 15% figure therefore does not represent a uniform increase for every Nvidia server configuration.
Vera Rubin production expands across Nvidia server ecosystem
Memory suppliers have continued reallocating production toward server and high-performance products as AI demand absorbs capacity. TrendForce said that shift has reduced supply available for some PC and consumer DRAM products. Server shipments remained strong through 2026, while buyers continued building inventories of data center memory. The market researcher expects server DRAM supply to stay tight into 2027 as demand growth outpaces new supply. Those market conditions form the cost backdrop for the reported Nvidia pricing changes.
Nvidia enters the pricing period with record data center sales. In its fiscal first quarter ended April 26, the company reported total revenue of $81.6 billion. Data Center revenue reached a record $75.2 billion, rising 92% from a year earlier. Nvidia guided for second-quarter revenue of $91 billion, plus or minus 2%, when it reported those results in May. The company will release its fiscal second-quarter results on Aug. 26.