Nvidia's $2,000-per-GPU software discussion leaves the product scope unresolved

Reported remarks point to recurring agent-infrastructure software revenue, while overlap with existing enterprise licences remains unclear.

Nvidia chief executive Jensen Huang has discussed an approximate $2,000-per-GPU annual software licence for running agents on AI infrastructure, according to an October 25 Tom's Hardware report covering remarks at GTC Berlin. The report says the company has not clarified when that pricing would take effect or exactly which software it would cover.

That uncertainty matters because Nvidia already sells enterprise software subscriptions. Tom's Hardware cites a listed one-year AI Enterprise price of $4,500 per GPU and says the two figures have not been confirmed to cover the same products or terms. The new discussion therefore cannot establish either a universal additional charge or a price cut.

The commercial direction is broader than a GPU sale. Agents need runtime infrastructure, permissions, security monitoring and operating support alongside model inference. Software that supplies those capabilities can create a recurring relationship with the customer after the hardware has been purchased.

Revenue projections attached to the remarks also depend on assumptions about adoption and renewals. Multiplying a hypothetical five million licences by $2,000 produces $10 billion annually, but it does not establish that five million customers or devices have contracted for the service.

For infrastructure buyers, the unresolved item is the licence boundary: which components, support and deployment rights are included, and how those relate to existing subscriptions. Huang's remarks indicate the growing role Nvidia sees for software. A complete commercial comparison will require the actual product terms, rather than treating a discussed per-GPU figure as a settled bill.

Sources

- [Nvidia discusses $2,000-per-GPU annual fee for software that runs agents on AI factories](https://www.tomshardware.com/pc-components/gpus/nvidia-discusses-usd2-000-per-gpu-annual-software-fee-for-running-agents-on-ai-infrastructure-firm-already-charges-more-than-twice-that-amount-for-its-ai-enterprise-software-platform) — Tom's Hardware