NVIDIA CEO Hints at Annual GPU Subscription Fees, Says Software Licensing Will Be a Major Part of Its Business

NVIDIA wants to make money long after its AI chips leave the factory, and Jensen Huang has put a number on that ambition. The NVIDIA CEO, floated a potential $2,000 annual subscription fee per GPU for running AI agents, opening another revenue opportunity beyond hardware sales.

NVIDIA’s AI software could cost $2,000 per GPU annually

Huang reportedly discussed the pricing example during a Q&A session at GTC Berlin on October 22. Huang said software licensing would become an increasingly important part of NVIDIA’s business. His $2,000 figure referred to the software NVIDIA provides to run agents inside what the company calls AI factories.

These AI factories are data center infrastructures built to power demanding AI workloads. The software layer helps businesses deploy and manage AI agents, which can use tools, access data, and perform tasks with limited human intervention.

However, NVIDIA has not announced a finalized $2,000 subscription price. The company has also not clarified which products would fall under this proposed licensing approach or when it could become available.

NVIDIA already charges for enterprise AI software

For those unaware, NVIDIA already sells its AI Enterprise platform through annual GPU-based subscriptions. According to the company’s official enterprise licensing guide, the standard one-year subscription for self-managed systems costs $4,500 per GPU, including support.

That makes Huang’s example notably cheaper on paper. However, the two figures may cover different software bundles and licensing terms, so this should not be treated as a confirmed price cut. The revenue potential could be substantial if Nvidia eventually licenses its software across millions of GPUs. For example, five million licenses at $2,000 each would generate $10 billion annually. That is an illustrative calculation, not Nvidia’s forecast.