SANTA CLARA, CALIFORNIA / RankWire.AI / – Nvidia is planning to increase prices by over 15% on many AI server configurations scheduled for release in early 2027. These hikes apply to systems built on the Vera Rubin and Grace Blackwell platforms. The magnitude of each price adjustment depends on the chip generation and memory configuration involved. Nvidia has not issued a blanket announcement for a universal price hike across all server lines, but several server manufacturers have already communicated the new pricing details to major data center clients.
Major data center operators such as Microsoft, Google, and Oracle have been informed of the updated pricing by server manufacturers. These companies purchase large quantities of AI infrastructure for cloud services and data centers. The rising costs associated with memory components have become a key factor influencing the reported increases in server prices. AI deployments demand significant amounts of high-bandwidth and server memory, in addition to GPUs, CPUs, networking hardware, and storage systems. The demand from AI applications has kept several memory categories in short supply throughout 2026.
According to TrendForce, contract prices for conventional DRAM are projected to rise by 13% to 18% in the third quarter of 2026. Simultaneously, NAND Flash contract prices are expected to grow by 10% to 15% during the same period. The research firm highlighted that server DRAM remains undersupplied as suppliers prioritize capacity for AI-related needs. While some long-term supply agreements have restrained the extent of price increases, overall server memory costs continue to stay elevated.
Memory costs escalate amid sustained AI server demand
In May, Nvidia announced that Vera Rubin had entered full production with server manufacturers and supply-chain partners around the globe. The company indicated that products based on Rubin would start appearing in the second half of 2026. Vera Rubin integrates the Vera CPU, Rubin GPU, NVLink 6, ConnectX-9, BlueField-4, and Spectrum-6 networking technologies. It succeeds Grace Blackwell as Nvidia’s latest rack-scale architecture designed for large AI workloads. Manufacturers are developing Rubin-based solutions for cloud providers, AI research labs, and hyperscale data centers.
Grace Blackwell remains a core component of Nvidia’s current data center lineup. The GB200 NVL72 design links 36 Grace CPUs with 72 Blackwell GPUs within a single liquid-cooled rack. This configuration is intended to operate as one NVLink-enabled computing domain capable of handling extensive AI models. Price increases vary depending on the configuration, with memory capacity and chip generation being critical factors. Consequently, the 15% figure does not apply uniformly across all Nvidia server offerings.
Expansion of Vera Rubin production across Nvidia’s server portfolio
Suppliers of memory have shifted more production toward server and high-performance products as AI demand continues to absorb capacity. TrendForce noted that this shift has reduced the supply available for some PC and consumer DRAM products. Despite strong server shipment numbers through 2026 and ongoing inventory accumulation by buyers, the market forecast indicates that server DRAM shortages will persist into 2027, with demand outpacing new supply. These conditions significantly influence the pricing environment for Nvidia’s recent price adjustments.
Nvidia enters this pricing phase with record-breaking data center sales. In its fiscal first quarter ending April 26, the company reported total revenue of $81.6 billion, with Data Center revenue hitting a record $75.2 billion—up 92% year-over-year. Nvidia projected second-quarter revenue of approximately $91 billion, plus or minus 2%, during its May earnings report. The company is scheduled to release its fiscal second-quarter results on Aug. 26.