AI factory

An AI factory is industry shorthand — popularized by Nvidia — for a purpose-built data center engineered end-to-end to turn electricity and data into AI training and inference output, rather than to host general-purpose enterprise workloads. Unlike a traditional colocation hall, an AI factory is designed around dense GPU clusters — often tens of thousands of accelerators tied together over NVLink and high-speed InfiniBand or Ethernet fabrics — with rack densities of 100-140+ kW that push direct-to-chip or immersion cooling from an option to a default requirement. Power procurement, frequently gigawatt-scale and sometimes behind-the-meter, is treated as a first-order design constraint alongside compute, since GPU clusters can swing load by tens of megawatts in seconds during training synchronization. The term describes the physical facility and its engineering, distinct from 'neocloud,' which describes the business model of companies that sell access to that capacity.