Nvidia discusses $2k-per-GPU annual fee for software

At GTC Berlin, Nvidia CEO Jensen Huang described an annual software license of about $2,000 per GPU for software that runs AI agents in data centers. At that rate, five million paid licenses would generate $10 billion a year, or $50 billion over five years if customers renewed; another estimate assumes 10 million eligible GPUs and 50% uptake. Nvidia has not specified when the fee might begin, which software it would include, how many GPUs could be covered, or whether $2,000 is a standard price. The company already lists its AI Enterprise platform at $4,500 per GPU for a one-year subscription, but the relationship between the offerings is unknown, so the lower figure does not establish a price cut.

The discussion reflects Nvidia’s effort to sell a broader AI infrastructure stack, not just chips. Its OpenShell architecture combines computing hardware with a secure runtime that limits agents’ access to tools and resources, while Nvidia Sentry monitors activity and can quarantine agents that breach those limits. The architecture extends the company’s earlier agent-safety work and is being incorporated into products from Anthropic and Salesforce. Nvidia presents security and management as essential for businesses deploying agents at scale, creating a potential reason to buy its software and support alongside its hardware. The commercial opportunity depends on adoption and on what the proposed license actually covers.