Nvidia Says Jensen Huang Cites $100 Billion Cost for 1GW AI Factories

Jensen Huang, chief executive officer of Nvidia, said that building an AI data center on the company’s hardware architecture is expected to cost around $100 billion once it scales to 1 gigawatt (GW) of power capacity. The estimate represents a near doubling of the current $55 billion price tag for AI factories of that magnitude, according to the statements.

The comment underscores the escalating capital expenditures required for frontier AI infrastructure as developers transition from megawatt-scale compute clusters to power plants the size of large industrial facilities. Nvidia’s architecture serves as the industry standard for training large reasoning models, which demand thousands of the company’s GPUs co-located in gigawatt-class data centers to operate efficiently.