NVIDIA AI Enterprise

NVIDIA AI Enterprise is NVIDIA's licensed, supported software platform for building and running production AI on NVIDIA GPUs. It has two layers that release on independent schedules. The application layer holds NIM microservices, NeMo tools and AI frameworks. The infrastructure layer holds GPU drivers, vGPU partitioning, Kubernetes operators and Run:ai workload orchestration.

Licensing is per GPU. Every GPU installed in a server or workstation that hosts the software needs a license, including each GPU on a multi-GPU board. Each H100 PCIe, H100 NVL and H200 NVL card includes a five-year subscription, activated with the card's serial number. That term starts 90 days after the board ships to the server maker. Hopper-based DGX systems include it in the DGX software bundle, while Blackwell-based DGX systems need licenses bought separately. A production branch is supported for 9 months, with a new one every 6 months. A long-term support branch is supported for 3 years, with a new one every 30 months.

Count licenses per installed GPU, not per server, and check whether the card already carries a subscription. Regulated deployments that must hold one software version for up to 3 years belong on a long-term support branch, not a production branch.

Sources

Last verified October 7, 2026.