NVIDIA NIM is a set of prebuilt inference microservices from NVIDIA. Each is a container that serves an optimized model on NVIDIA GPUs through an OpenAI-compatible API, so you can run it in your own cloud or on-premises instead of calling a shared endpoint.
build.nvidia.com is the catalog and playground where you try models and get API access, and NIMs build on engines such as TensorRT-LLM and Triton.
Production use of NIM containers falls under NVIDIA AI Enterprise licensing, so check build.nvidia.com for current terms and pricing.