SANTA CLARA, CALIFORNIA / RankWire.AI / – Nvidia is preparing price increases of more than 15% for many AI server configurations scheduled to ship in early 2027. The changes affect systems built around Vera Rubin and Grace Blackwell technology. Final increases differ according to chip generation, memory capacity and system design. Nvidia has not announced a single companywide increase covering every server configuration. Manufacturers that assemble AI systems have communicated revised pricing to large data center customers.
Microsoft, Google and Oracle rank among the major cloud operators buying large volumes of accelerated computing equipment. Their data centers use AI servers for model training, inference and cloud services. Memory has become one of the largest cost pressures across these systems during 2026. Modern AI servers combine GPUs with high-bandwidth memory, server DRAM, storage and high-speed networking. Strong demand for those components has kept supplies tight across several parts of the memory market.
TrendForce projected conventional DRAM contract prices would increase 13% to 18% during the third quarter of 2026. It also forecast NAND Flash contract prices to rise 10% to 15% over the same period. Server DRAM remains particularly constrained as memory producers allocate more capacity to AI and data center products. Higher memory prices have increased the cost of building advanced computing systems. Those increases form a key part of the pricing backdrop for next-generation AI servers.
Memory costs add pressure across AI infrastructure
Vera Rubin entered full production with server manufacturers and supply-chain partners in 2026, according to Nvidia. Partner systems using the platform are scheduled to become available during the second half of the year. Rubin combines the Vera CPU and Rubin GPU with NVLink 6 and several networking technologies. The platform targets large-scale artificial intelligence workloads in cloud and hyperscale data centers. It follows Grace Blackwell as the company’s newest rack-scale computing architecture.
Grace Blackwell continues to serve as a core platform in current AI data center deployments. The GB200 NVL72 system connects 36 Grace CPUs with 72 Blackwell GPUs inside a liquid-cooled rack. Nvidia designed the platform to operate as one large NVLink computing domain. Pricing changes tied to these systems vary by hardware configuration rather than following one fixed percentage. Memory capacity, processor generation and rack design all affect the final cost of each server configuration.
Server demand keeps memory supply conditions tight
Memory manufacturers have shifted more production toward server and high-performance products as artificial intelligence demand absorbs capacity. TrendForce has said this transition reduced supply available for some PC and consumer memory categories. Data center operators also continued buying large volumes of server memory through 2026. The research firm expects server DRAM availability to remain tight into 2027 as demand grows faster than new supply. That environment continues to influence component costs across AI infrastructure.
Nvidia enters the pricing period after another quarter of record data center revenue. The company reported fiscal first-quarter revenue of $81.6 billion for the period ended April 26, 2026. Data Center revenue reached $75.2 billion, up 92% from the same quarter a year earlier. Nvidia also guided for second-quarter revenue of $91 billion, plus or minus 2%. The company is scheduled to report its fiscal second-quarter results on Aug. 26, providing its latest financial update.