Nvidia floats $2,000-per-GPU software fee to run AI agents
What's the deal? Nvidia chief executive officer Jensen Huang floated a figure of roughly $2,000 per GPU per year for software to run AI agents on the company's infrastructure. He offered it as an example, not a formal price announcement — Nvidia has not said what software the figure covers or when it would apply.
Why now? The comment landed as Nvidia presented OpenShell and related infrastructure for securing and operating AI agents. The push comes as Microsoft, Google, and other platforms expand their own tools for building and governing enterprise agents.
What's the endgame? Nvidia wants to sell a broader stack beyond chips. The strategic prize is recurring software revenue tied to its vast installed GPU base, rather than one-off hardware sales.
The catch: The $2,000 figure cannot be read as a confirmed price, or a price cut. Nvidia already lists its AI Enterprise subscription at $4,500 per GPU annually, so the lower number is not directly comparable to any existing offer.
The signal: The remark signals Nvidia's intent to layer software revenue on top of its hardware dominance. But the commercial impact stays unclear until the company defines the bundle and its licensing terms.
Read more: Tom's Hardware
