Nvidia has notified major customers that server prices featuring its AI chips will increase by over 15% starting next year due to rising global memory component costs.

Key Points

• Server manufacturers have informed clients, including Microsoft, Google, and Oracle, that hardware costs will rise significantly beginning in early 2025.
• The price hikes impact systems utilizing Nvidia’s latest hardware, specifically the Grace Blackwell and Vera Rubin chip architectures.
• Supply constraints for high-bandwidth DRAM from manufacturers like Samsung, SK hynix, and Micron are the primary drivers behind the increased production costs.
• Other major tech firms, including Amazon Web Services and Qualcomm, have already implemented price increases for their respective AI and hardware services.
• European AI infrastructure projects, which rely on multi-billion euro investments, face potential budget shortfalls due to these unexpected hardware price adjustments.

Why it Matters

These price increases signal that the high cost of AI infrastructure is shifting from chip designers to the end users and operators of data centers. This trend threatens to disrupt the financial planning of large-scale AI projects and could slow the pace of global data center expansion.