NVIDIA uses the term to describe a full-stack facility combining advanced chips, packaging, memory, and networking with land, power, and shell (LPS) infrastructure, all built to run its DSX platform for training and inference. The framing treats compute output as a production line — energy and data in, tokens and model intelligence out — the way a factory converts raw inputs into goods. NVIDIA's August 2026 guarantee of LPS costs at the PORTS-Pike site in Ohio, built for OpenAI, is one of the largest concrete examples of the concept.