Nvidia already sells the chips. Now Jensen Huang has put a yearly price on the software that runs AI agents on them. Meet the Nvidia AI software GPU licensing fee.
The Nvidia AI software GPU licensing fee started as one line in a Q&A. At GTC Berlin on October 22, 2026, CEO Jensen Huang said running Nvidia’s agent software costs about $2,000 per GPU per year. However, Nvidia has not launched a product, a price list or a start date.
Nvidia AI Software GPU Licensing Fee: What Huang Said in Berlin
Huang made the comment during a question session at GTC Berlin, Nvidia’s European developer event. In particular, his point was about where the company’s money will come from next.
“Software licensing will become increasingly a major part of our company’s business,” he said, as reported by Notebookcheck.
Then he gave the number. In short, running “all the software that Nvidia provides that runs agents on these AI factories” costs about $2,000 per GPU per year, Huang said. In Nvidia’s language, an AI factory is a data center built to produce AI output at scale.
About $2,000 per GPU per Year, for Agent Software
So what would the money buy? Nvidia has not said which products the figure covers. Notebookcheck and Geo News both stress that the company is still weighing the idea.
Still, the timing gives a strong hint. One day earlier, Nvidia had unveiled a full software and hardware stack for running AI agents safely inside companies. Geo News links the fee to that stack.
AI agents are programs that use tools, read data and finish tasks with little human help. As a result, businesses that run thousands of them need software to schedule the work and keep each agent inside its permissions. That management layer is what Nvidia wants paid for.
NVIDIA AI Enterprise Pricing vs the $2,000 Figure
Nvidia already charges for software, though. Its NVIDIA AI Enterprise suite has a public price list, and it is higher than the new number. Here is how the official list prices compare with what Huang floated.
| Licence | Price per GPU | Term |
|---|---|---|
| Agent software fee (floated by Huang) | about $2,000 | per year |
| NVIDIA AI Enterprise subscription | $4,500 | 1 year |
| NVIDIA AI Enterprise subscription | $13,500 | 3 years |
| NVIDIA AI Enterprise subscription | $18,000 | 5 years |
| NVIDIA AI Enterprise perpetual | $22,500 | 5 years of support |
| Education and Inception price | $1,125 | 1 year |
| Cloud marketplace, production | $1 per hour | pay as you go |
In other words, the $2,000 figure is less than half the one-year AI Enterprise price. Tom’s Hardware made the same point in its headline. That said, the two may cover different software, so nobody should read this as a price cut yet.
OpenShell and Sentry: The Agent Stack in the Background
On October 21, Nvidia launched the OpenShell Reference Architecture at the same event. It is a blueprint for building, monitoring and controlling AI agents across data centers and edge machines.
The design runs models on Vera Rubin systems and agent tools on Vera CPUs. Meanwhile, BlueField-4 data processing units run NVIDIA Sentry, which watches agent traffic and can stop an agent that breaks its rules. The OpenShell runtime itself is open source.
Smaller machines are also part of the plan. Nvidia lists RTX Spark, DGX Station and the NVIDIA DGX Spark desktop as edge systems for the same agent setup.
“AI capability and safety must advance together,” Huang said in the launch release. Partners range from Anthropic, OpenAI and Microsoft to SAP, Dell and CoreWeave.
Why the Nvidia AI Software GPU Licensing Fee Matters to Investors
Today, Nvidia earns most of its money from hardware. A yearly software charge would add income that repeats for as long as a GPU stays in service.
For a sense of scale, Notebookcheck ran a simple illustration. At $2,000 a year, five million paid licences would bring in $10 billion annually, or $50 billion over five years with renewals.
However, that is arithmetic, not a forecast. Nvidia has not said how many GPUs would qualify. It also has not confirmed whether $2,000 is a final price or a standard one.
What an Nvidia GPU License Means for Buyers
For companies, therefore, the fee would turn part of the GPU bill into a subscription. As a result, budget planners would need to count software years, not only the upfront hardware cost.
Some buyers already know this model. NVIDIA AI Enterprise includes a five-year term with H100, H200 NVL and A800 PCIe GPUs, according to Nvidia’s licensing guide. In the cloud, the suite also costs $1 per GPU per hour on top of the instance price.
Smaller teams can wait and watch. For instance, developers on a single workstation card are not the audience for an AI factory licence, at least not yet.
What We Still Don’t Know About the Nvidia AI Software GPU Licensing Fee
Even so, most of the details are still missing. Nvidia has not named the products, the start date or how the fee fits with AI Enterprise.
Moreover, it is unclear whether customers who already pay for AI Enterprise would pay twice. Geo News notes that Nvidia has not confirmed any overlap between the two offerings.
For now, the takeaway is the direction. Nvidia wants software to become a major part of its business, and AI agents are the reason it gives. We will update this story when Nvidia publishes real terms.
Want More on Nvidia AI Software GPU Licensing Fee?
If you build agents yourself, start with our guide to the top agentic AI frameworks for building AI agents. Then compare hardware costs in our pick of the best GPU for AI and local LLMs.
Frequently Asked Questions
What is the Nvidia AI software GPU licensing fee?
It is a yearly charge of about $2,000 per GPU that Jensen Huang floated for Nvidia’s agent software. Nvidia has not launched it or named the products it covers.
How much does NVIDIA AI Enterprise cost per GPU?
The official list price is $4,500 per GPU for one year with support. Longer terms cost $13,500 for three years and $18,000 for five years.
When did Jensen Huang mention the $2,000 fee?
He mentioned it in a Q&A at GTC Berlin on October 22, 2026. A day earlier, Nvidia had launched the OpenShell Reference Architecture for AI agents.
Does the fee replace NVIDIA AI Enterprise?
Nobody knows yet. Nvidia has not said whether the two overlap, so reports warn against reading the $2,000 figure as a price cut.
What is NVIDIA OpenShell?
OpenShell is open source runtime software for running AI agents with clear limits. In addition, the reference architecture pairs it with Vera Rubin, Vera CPUs and BlueField-4 DPUs running Sentry.
Is NVIDIA AI Enterprise included with any GPUs?
Yes. According to Nvidia’s licensing guide, a five-year term comes with H100, H200 NVL and A800 PCIe GPUs. Other cards need a paid licence.
Sources
Notebookcheck (October 25, 2026), Geo News (October 26, 2026), Tom’s Hardware, NVIDIA press release, OpenShell Reference Architecture (October 21, 2026), NVIDIA AI Enterprise licensing guide, pricing. *Photos: Jensen Huang, Nvidia Keynote, CES 2025 Las Vegas by Pronoia, CC0, cropped and resized; Nvidia DGX GB200 by Pokiiri, CC BY-SA 4.0, resized and placed on a blurred background.*
