Nvidia discusses a proposed $2,000 annual fee per GPU for agent software

Tom’s Hardware reported that Nvidia was discussing an annual fee of $2,000 per GPU for software that “runs agents on AI factories.”[1] Notebookcheck separately reported that Nvidia was considering a $2,000 annual software fee per GPU for running AI agents.[2] The reports describe a proposal, rather than a confirmed price with a stated start date.[1] Notebookcheck said the licensing model could generate billions of dollars in recurring revenue, but did not establish that revenue as a result already achieved.[2] For teams assessing budgets, the key point is that the proposed charge is expressed per GPU and per year.

Tom’s Hardware said the potential overlap between the proposed fee and Nvidia AI Enterprise was unclear.[1] The report also said Jensen Huang had not revealed when the proposed pricing could take effect.[1] Buyers should treat the $2,000 figure as a proposed price, not as a published effective rate.

Nvidia’s licensing guide lists a one-year self-managed AI Enterprise subscription at $4,500 per GPU, a two-year subscription at $9,000, and a three-year subscription at $13,500.[3] The guide also lists a five-years-for-the-price-of-four multi-year discount.[3] These published prices give buyers a reference point, but do not resolve whether AI Enterprise and the proposed agent software would cover overlapping needs.[1][3] Nvidia lists a perpetual AI Enterprise license at $22,500 per GPU, including five years of support.[3] Its licensing guide says a license is required for every GPU installed in a server or workstation hosting the software.[4]

Nvidia says AI Enterprise can be purchased as a subscription, through cloud consumption, or as a perpetual license with required five-year support.[4] Subscriptions must be renewed to remain active, while Nvidia lists cloud-hosted production pricing at $1 per GPU per hour, in addition to cloud-provider instance costs.[4][3] The guide limits education and Inception discounts to qualified institutions and program members.[3] It also says each H200 NVL Tensor Core GPU includes a five-year AI Enterprise subscription.[4]