Bernstein: Building 1GW AI Data Center with Nvidia's Vera Rubin Costs $47 Billion

According to Bernstein, building a 1GW AI data center using Nvidia's next-generation Vera Rubin architecture, expected to deploy in 2027, would require approximately $47 billion in total capital expenditure. A single Vera Rubin NVL72 rack costs an estimated $9.1 million and would need 3,557 racks for a 1GW facility, with equipment costs reaching $32 billion and supporting infrastructure adding $15 billion. The Vera Rubin architecture delivers 2,520 FP8 petaflops of computing power per rack, more than triple the 720 petaflops of current Blackwell systems, improving cost-efficiency per unit of computation despite the 14% increase in per-rack expense.
Disclaimer: The information on this page may come from third-party sources and is for reference only. It does not represent the views or opinions of Gate and does not constitute any financial, investment, or legal advice. Virtual asset trading involves high risk. Please do not rely solely on the information on this page when making decisions. For details, see the Disclaimer.
Comment
0/400
Lifelikevip
· 13h ago
Bernstein: Building a 1GW AI Data Center using NVIDIA Vera Rubin, costing $47 billion
View OriginalReply0
StopLookingUpAtTheMountain.vip
· 13h ago
Bernstein: Building a 1GW AI data center with NVIDIA Vera Rubin, costing $47.0 billion
View OriginalReply0