SANTA CLARA, CALIFORNIA / RankWire.AI / – Nvidia is planning to implement price increases exceeding 15% on many AI server configurations set for shipment in early 2027. These adjustments impact systems based on Vera Rubin and Grace Blackwell technologies. The final price hikes vary depending on factors such as chip generation, memory capacity, and system configuration. Nvidia has not issued a uniform companywide price increase applicable to all server models. Instead, hardware manufacturers have communicated updated pricing to large data center clients for their AI system assemblies.
Microsoft, Google, and Oracle are among the leading cloud service providers purchasing significant quantities of accelerated computing hardware. Their data centers rely on AI servers for tasks including model training, inference, and cloud-based services. Throughout 2026, memory costs have become a major expense for these systems. Modern AI servers typically combine GPUs with high-bandwidth memory, server DRAM, storage solutions, and high-speed networking. The high demand for these components has kept supply tight across various parts of the memory market.
TrendForce predicted that traditional DRAM contract prices would increase by 13% to 18% during the third quarter of 2026. It also forecasted NAND Flash contract prices to rise 10% to 15% over the same period. Server DRAM remains especially constrained as memory producers dedicate more capacity to AI and data center applications. The rising memory prices have driven up the costs associated with building advanced computing systems, forming a key element of the pricing landscape for next-generation AI servers.
Memory price hikes intensify pressures on AI infrastructure costs
According to Nvidia, Vera Rubin reached full production with server manufacturers and supply-chain partners in 2026. Systems incorporating the platform are expected to become available in the second half of the year. Rubin integrates the Vera CPU and Rubin GPU with NVLink 6 and multiple networking technologies. Designed for large-scale AI workloads in cloud and hyperscale data centers, the platform succeeds Grace Blackwell as Nvidia’s latest rack-scale computing architecture.
Grace Blackwell remains a foundational platform in current AI data center deployments. The GB200 NVL72 system pairs 36 Grace CPUs with 72 Blackwell GPUs inside a liquid-cooled rack. Nvidia engineered this setup to operate as a unified NVLink computing domain. Pricing adjustments linked to these systems depend on hardware configurations such as memory capacity, processor generation, and rack design, rather than a fixed percentage across all models.
Demand for servers sustains tight memory supply conditions
As artificial intelligence demand shifts more production toward server and high-performance products, memory manufacturers have reallocated capacity accordingly. TrendForce noted that this transition has led to reduced supply availability in certain PC and consumer memory segments. Additionally, data center operators continued purchasing large volumes of server memory throughout 2026. The research firm predicts that server DRAM supply will remain tight into 2027 as demand continues to grow faster than new production. These conditions continue to influence component costs across AI infrastructure.
Nvidia is entering this pricing phase after posting another quarter of record revenue from data centers. The company announced fiscal first-quarter revenue of $81.6 billion for the period ending April 26, 2026. Data Center revenue reached $75.2 billion, a 92% increase compared to the same quarter the previous year. Nvidia has also projected second-quarter revenue of $91 billion, plus or minus 2%. Its fiscal second-quarter results are due to be reported on August 26, marking its latest financial update.