A concept describing a modern data center designed not just for storage or running web services, but specifically for the industrial-scale production of intelligence (tokens). In an AI Factory, raw data enters, and refined intelligence (in the form of trained models or generated inference) exits. It emphasizes the shift from traditional general-purpose computing to specialized, high-performance infrastructure (like clusters of thousands of H100 GPUs) optimized for massive matrix operations. This represents a fundamental shift in how computing infrastructure is architected, moving from CPU-centric servers to GPU-accelerated supercomputers.
Popularized by Nvidia CEO Jensen Huang to describe the new industrial revolution driven by accelerated computing.
Adopted by hyperscalers (Azure, AWS) and sovereigns building national AI infrastructure to characterize their massive GPU investments.