Nvidia Explores Annual Licensing Model for AI Software
Jensen Huang indicated that Nvidia is contemplating an annual licensing fee of approximately $2,000 per GPU for its software that operates AI agents.
However, the company has yet to clarify when this fee might be instituted, what it will encompass, or whether this figure will serve as a baseline charge.
This potential pricing strategy was a topic of discussion during Huang’s presentation at GTC Berlin on October 22, 2026, where Nvidia unveiled plans to establish a robust AI infrastructure business that integrates chips, software, and security solutions for AI agents.
The Register reported that the annual license cost could be pegged at around $2,000 for each GPU.
Software licensing is set to become an increasingly significant aspect of our organization’s operations. The licensing strategy will entail, for instance, an estimated charge of $2,000 per GPU annually for accessing the software that Nvidia provides for agent management within these AI infrastructures.
– Jensen Huang
A projection by FourWeekMBA posits that if Nvidia were to secure five million paid licenses at the proposed rate, the annual revenue could surge to $10 billion.
Over a five-year horizon, if all customers opted for renewal, this could translate to a staggering $50 billion.
Another scenario anticipates the shipment of 10 million eligible enterprise GPUs within a single year, with half potentially incorporating the paid software.
This scenario also suggests a recurring revenue potential of $10 billion annually, amounting to $50 billion over five years, contingent on subscription continuance. It’s crucial to note that these figures are projections rather than confirmed sales or licensing agreements.
Nvidia has refrained from divulging the number of GPUs that might fall under the proposed fee, the standardity of the $2,000 price point, or the specific software included in the licensing agreement.
Notably, Nvidia’s AI Enterprise platform is currently priced at $4,500 per GPU for a one-year subscription.
The firm has not substantiated that these two pricing structures pertain to identical products or terms, thereby indicating that the lesser figure should not be interpreted as an established price reduction.
This pricing discourse emerges as Nvidia is actively enhancing its infrastructure for the deployment and governance of AI agents within corporate environments.
As these agents possess the capability to utilize tools, access data, and execute tasks, organizations are confronted with significant security and management dilemmas when scaling such operations.
On October 21, 2026, Nvidia unveiled the OpenShell Reference Architecture, integrating Vera Rubin computing infrastructure, Vera CPUs, and BlueField-4 data processing units propelled by Nvidia Sentry.
This innovative framework aims to delineate the operational boundaries of agents, governing their access to tools and resources. Sentry independently oversees agent activities, capable of isolating any that breach designated boundaries.
This architecture aspires to embed security controls within the fundamental infrastructure rather than merely relying on safeguards situated within an AI model or application.
Earlier in September, Nvidia rolled out its Open Agent Safety Platform. The newly introduced reference architecture builds upon this initiative, extending its relevance to enterprise infrastructure and edge computing systems.
Collaboratively, Anthropic is integrating OpenShell along with BlueField hardware into Claude Managed Agents, thus equipping businesses with enhanced oversight on agent accessibility within execution sandboxes.
Similarly, Salesforce has developed a Slack Admin Bridge for OpenShell, empowering administrators to scrutinize agent access requests, grant or deny permissions, and monitor operations through Slack.
The proposed licensing fee aligns with Huang’s broader assertion that Nvidia is transitioning beyond its core GPU business into diverse realms including networking, systems, enterprise software, and AI infrastructure. In a recent appearance on the BG2Pod podcast, he articulated:
In the not-too-distant future, the perception may shift such that we are no longer classified merely as a chip company.
– Jensen Huang
Huang highlighted Nvidia’s expansive engagement in sectors such as healthcare, automotive, telecommunications, and life sciences, where the company provides comprehensive technology solutions.
As of now, Nvidia has yet to declare when the proposed $2,000-per-GPU fee may take effect or how this will integrate into its pre-existing enterprise software framework.
Source link: Mezha.net.
