SANTA CLARA, CALIFORNIA / RankWire.AI / – Nvidia is planning to raise prices by more than 15% for numerous AI server configurations scheduled for shipment in early 2027. These adjustments impact systems built with Vera Rubin and Grace Blackwell technology. The final price hikes vary depending on chip generation, memory capacity, and system design. Nvidia has not announced a uniform companywide increase that applies to all server models. Instead, manufacturers assembling AI systems have communicated updated pricing to large data center clients.
Microsoft, Google, and Oracle are among the key cloud service providers purchasing substantial quantities of accelerated computing hardware. Their data centers rely on AI servers for tasks such as model training, inference, and cloud offerings. Throughout 2026, memory costs have become one of the most significant financial pressures affecting these systems. Modern AI servers integrate GPUs with high-bandwidth memory, server DRAM, storage, and high-speed networking. The strong demand for these components has kept supply tight across multiple areas of the memory market.
According to TrendForce, contract prices for conventional DRAM were projected to increase between 13% and 18% during the third quarter of 2026. The same forecast indicated NAND Flash contract prices could rise by 10% to 15% over that period. Server DRAM, in particular, remains constrained as memory manufacturers allocate more capacity toward AI and data center products. The rising memory costs have increased the expense of building advanced computing systems, which forms a significant part of the pricing environment for next-generation AI servers.
Growing memory expenses influence AI infrastructure pricing
In 2026, Nvidia reported that Vera Rubin entered full production with server manufacturers and supply-chain partners. Systems incorporating the platform are scheduled to become available in the latter half of the year. Rubin combines the Vera CPU and Rubin GPU with NVLink 6 and various networking technologies. Designed for large-scale AI workloads in cloud and hyperscale data centers, it follows Grace Blackwell as Nvidia’s latest rack-scale computing platform.
Meanwhile, Grace Blackwell continues to serve as a foundational architecture in current AI data center deployments. The GB200 NVL72 system links 36 Grace CPUs with 72 Blackwell GPUs inside a liquid-cooled rack. Nvidia designed this platform to function as a single, large NVLink computing domain. Price adjustments related to these systems differ based on hardware configuration rather than following a fixed percentage. Variations in memory capacity, processor generation, and rack design all influence the final cost of each server setup.
Memory supply tightness persists due to rising server demand
As AI demand continues to grow, memory producers have shifted a larger share of their output toward server and high-performance applications. TrendForce has noted this transition has reduced supply for some PC and consumer memory categories. Data center operators have also maintained high-volume purchases of server memory throughout 2026. The research firm expects server DRAM availability to stay limited into 2027 as demand outpaces supply growth. This environment remains a key factor influencing component prices across AI infrastructure.
Following another quarter of record data center revenue, Nvidia is entering this pricing phase. The company reported fiscal first-quarter revenue of $81.6 billion for the period ending April 26, 2026. Data Center revenue reached $75.2 billion, marking a 92% increase compared to the same quarter last year. Nvidia also forecasted second-quarter revenue of $91 billion, with a margin of plus or minus 2%. The company plans to release its fiscal second-quarter results on Aug. 26, offering its latest financial outlook.