Nvidia has informed cloud providers that select AI server configurations will see a 15% price increase effective in 2027 due to escalating component and packaging costs.

Nvidia has reportedly notified major cloud customers that select next-generation server systems will see price increases exceeding 15% beginning in early 2027. The adjustment is expected to impact Vera Rubin and Grace Blackwell architectures. Notifications have been distributed via Nvidia’s server partners, prompting hyperscalers such as Microsoft, Google, and Oracle to anticipate higher procurement costs.

The primary catalyst is surging memory costs. High-bandwidth memory (HBM) and other DRAM prices are climbing rapidly, with some analyst estimates projecting that memory components could account for 62% of a Vera Rubin system’s total cost. Contract pricing for DRAM is forecast to rise between 58% and 63% in the second quarter of 2026, following a 90% to 95% surge in the first quarter. Compounding the pressure is constrained supply; analysts describe the market as tight, and SK Hynix reportedly exhausted its entire 2026 production capacity by October of the preceding year.

For organizations operating at hyperscale, these developments warrant close monitoring. Nvidia previously implemented similar price adjustments across its GeForce lineup due to comparable memory cost pressures. Competing solutions, including AMD accelerators and custom silicon, remain subject to the same constrained memory supply chain. The reported pricing update will apply to specific Nvidia server configurations manufactured by partner facilities and scheduled for shipment from early 2027 onward.