Nvidia CEO Jensen Huang has put a number on the future of the company's software business: about $2,000 per GPU per year to run all of Nvidia's software that powers AI agents in so-called "AI factories". He said it in a Q&A session at GTC Berlin on October 22, first reported by The Register and picked up by TechSpot and Tom's Hardware.
The status matters: this is a statement, not a published price list. Huang offered the figure as an example — "for example, it's about $2,000 per GPU per year" — and said software licensing "will become increasingly a major part of our company's business". Nvidia has not confirmed which package the figure would cover, when it would apply, or how it would relate to its current offering.
The arithmetic explains the market interest: at $2,000 per GPU a year, each million licensed GPUs would mean $2 billion in annual recurring revenue — an illustrative calculation, not company guidance. The appeal for Nvidia is revenue that keeps flowing after the chip has been sold, rather than depending only on new hardware sales.
The open question is the comparison with what already exists. The Nvidia AI Enterprise suite already lists at $4,500 per GPU for one year, including support, or $18,000 for five years — equivalent to $3,600 a year. Because Huang's figure is lower than both, it remains unexplained whether it refers to a new, narrower agent-focused bundle, a future tier, or simply a rounded estimate given in public.
One day before the Q&A, on October 21, Nvidia launched the OpenShell Reference Architecture, a reference design for building, operating and governing AI agents safely, with Vera CPUs and BlueField-4 networking chips running Sentry software. Nothing disclosed so far establishes that OpenShell itself would carry the $2,000 fee: the technical groundwork for agents and the pricing discussion are two parallel moves.
In short, Nvidia has signalled a direction — software as a growing share of revenue — but the $2,000 per GPU a year figure remains an example given by its CEO, with no confirmed product, scope or date.
Image credit: TechSpot, in its article on Jensen Huang's remarks at GTC Berlin.
Fontes: TechSpot; The Register (reportagem original, via TechSpot); AI Weekly / Tom's Hardware.
Transparency: this article was produced with the assistance of artificial intelligence, cross-checking public posts on X and sources on the internet. Check the original sources to confirm the details before sharing.
— Muse
