Nvidia may soon charge for its AI software the way it already charges for its chips. Chief executive Jensen Huang told an audience at GTC Berlin on October 22 that running Nvidia’s full software stack for AI agents costs about $2,000 per GPU per year.
Software licensing will become an increasingly major part of the company’s business, Huang said, as Nvidia looks beyond one-off hardware sales toward recurring revenue from the data centers it calls AI factories.
The figure is an example, not a confirmed price list. Nvidia has not said which products the fee would cover, when it might take effect, or how it would relate to its existing AI Enterprise suite, which lists at $4,500 per GPU for a year.
The arithmetic shows why the idea matters: at $2,000 per GPU, five million paid licenses would produce about $10 billion a year. That is illustrative math from commentators, not company guidance, and Nvidia has not disclosed how many GPUs would qualify.
The comments followed the October 21 launch of the OpenShell Reference Architecture, a secure runtime for agents built on Vera CPUs and BlueField-4 networking chips running Sentry software, which can quarantine an agent that steps outside its limits.
For developers, the shift underlines how much modern AI work now runs on managed agent infrastructure. Coding assistants such as Cursor, Codeium and Cline all depend on GPU-backed systems, so any per-GPU software fee could eventually feed through into the cost of building with AI.
Source: TechSpot
