Nvidia may start charging $2,000 per GPU each year for software that runs AI agents on its chips. CEO Jensen Huang talked about this possible fee at GTC Berlin on October 22, as the company aims to grow its software business. Nvidia has not yet said which products this pricing would apply to or when it could begin.
“Software licensing will become a major part of our company’s business. The way that we license will be, for example, it’s about $2,000 per GPU per year to run all the software that Nvidia provides that runs agents on these AI factories,” Jensen Huang said.
The possible revenue could be large. If five million licenses were sold at $2,000 each per year, Nvidia would make $10 billion a year, or $50 billion over five years if customers kept renewing. These numbers are just examples, not Nvidia’s official forecast. The company has not said how many GPUs might qualify for the license or if $2,000 will be the final price.
Nvidia already offers its AI Enterprise software platform for $4,500 per GPU for a one-year subscription. This is more than double the proposed new fee, but the two prices may apply to different software or licensing terms. It is not yet clear how the new proposal would fit with Nvidia’s current products.
Nvidia wants to take advantage of the rising interest in AI agents, which can use tools, access data, and do tasks with little human help. Companies using these systems on a large scale need software to manage workloads and stop agents from reaching resources they should not access.
On October 21, Nvidia introduced its OpenShell Reference Architecture, which is meant to give AI agents a secure environment and limit their access to certain tools and resources. This setup uses Nvidia’s Vera Rubin computing system, Vera CPUs, and BlueField-4 data processing units (DPUs) with Nvidia Sentry. Nvidia Sentry watches agent activity on its own and can isolate agents that try to cross set security limits.