Key information
NVIDIA CEO Jensen Huang recently highlighted the increasing importance of software licensing for the company's AI business model. During a Q&A session at GTC Berlin, Huang suggested that running the software required to power agents in AI factories could cost approximately $2,000 per GPU annually. This figure was presented as an example rather than a formal pricing announcement.
The $2,000 estimate creates ambiguity regarding existing offerings, as NVIDIA currently lists its AI Enterprise suite at $4,500 per GPU for a one-year subscription including support. Because the company has not clarified if Huang's figure represents a new bundle, a future model, or a price reduction, analysts caution against interpreting it as a definitive change in current costs.
At the same event, NVIDIA introduced the OpenShell Reference Architecture, an open-source runtime designed to sandbox AI agents. This system integrates Vera Rubin inference hardware, Vera CPUs, and BlueField-4 DPUs running NVIDIA Sentry for safety monitoring. Huang emphasized this architecture as a key component for managing AI agent security.
The shift toward recurring software revenue remains a strategic priority for NVIDIA. By licensing software at the suggested rate, the company could generate significant income, with every million GPUs potentially yielding $2 billion in annual revenue. No specific timeline for the implementation of this licensing model has been provided by the company.
