Nvidia raising AI server prices over 15% as memory shortage bites

A global memory shortage affecting DRAM makers Samsung, SK hynix, and Micron simultaneously is driving the increase, not the GPUs themselves. Nvidia's Vera Rubin and Grace Blackwell systems are hit hardest because they require huge amounts of memory, such as a single Vera Rubin NVL72 rack using 20.7TB of HBM4 plus 54TB of LPDDR5X. Price notices for these increases, expected to start early next year and reportedly reaching around 17%, were relayed through server-assembly middlemen rather than communicated directly by Nvidia. Track how memory shortages reshape AI infrastructure costs and vendor pricing on daily.dev.

TrendForce expects DRAM supply to stay constrained through 2027, as AI server production continues pulling manufacturing capacity toward HBM and server memory at the expense of other memory production. This sustained tightness is why Nvidia server price increases are expected to persist rather than being a short-term spike, affecting both enterprise AI infrastructure and consumer hardware pricing. Follow the memory supply crunch as it keeps affecting AI infrastructure budgets on daily.dev.

A 17% increase could add at least $5 billion to the cost of a single 1-gigawatt data center build, based on Nvidia server costs alone. That figure excludes power, cooling, networking, physical buildings, and financing, meaning the true total cost increase for a large-scale AI data center project is even higher once those additional expenses are factored in. Compare real AI infrastructure cost projections as pricing shifts on daily.dev.