Nvidia Floats Per-Chip Software Subscription as Memory Data Extends the AI Demand Horizon
Published on 10/29/2026 at 03:31 | Editorial boerse-global.de
Nvidia is testing a second revenue stream on top of its hardware business. At GTC Berlin on October 22, chief executive Jensen Huang sketched an annual licensing fee of roughly $2,000 per graphics processor for software used to run AI agents, according to media reports. The figure is a talking point, not booked business — no concrete contribution to earnings can be derived from it yet — but the logic is plain enough: a yearly charge per chip would give Nvidia recurring income tied to how its processors are actually used, not merely to how many are shipped.
The proposal lands alongside the OpenShell Reference Architecture, unveiled on October 21. Nvidia says the framework supports building, operating, monitoring and governing AI agents across enterprise infrastructure and decentralized systems, and it named a broad group of technology partners working with the design. The two announcements should be kept apart: Huang's price remark does not establish that OpenShell itself will carry that fee. What matters for investors is the pairing — Nvidia is laying technical groundwork for agentic AI while airing ways to monetize it.
Demand Signal From the Memory Market
Fresh results from SK Hynix gave the sector a lift. The memory maker posted a record quarterly operating profit of 83.3 trillion won, beating expectations and buoying sentiment toward AI chip names including Nvidia. More telling was its outlook: SK Hynix expects demand to outstrip supply through 2030. That pushes back against the idea that the investment wave is about to break, though the claim is confined to the market SK Hynix describes and does not translate directly into Nvidia's growth or margins. Reports also pointed to continued financing of AI infrastructure by the large cloud providers, meaning the immediate impulse came from the industry backdrop rather than any new Nvidia announcement.
Should investors sell immediately? Or is it worth buying Nvidia?
Existing partnership plans give that backdrop some substance. Meta intends to broaden its use of Nvidia infrastructure, including Vera Rubin systems, Vera CPUs, Spectrum-X and BlueField-4, while the two companies deepen work on the Muse AI models and related agents. Nvidia's announced participation in Meta's share offering amounts to $3 billion — technical collaboration backed by its own capital. The arrangement could support future demand for Nvidia's AI chips, but it carries no guaranteed revenue.
Oracle's planned multi-gigawatt expansion of its cloud infrastructure fits the same pattern. Superclusters built on Vera Rubin systems are planned, with those systems and Vera CPUs already generally available there while further build-out is scheduled. Such projects describe where Nvidia hardware is meant to go; they are no proof that the mooted annual software fee forms part of any agreement. Hardware expansion and software revenue remain separate line items in any valuation.
A Market Move Without a Clear Cause
Nvidia shares slipped 2.1% in the prior session. No company-specific trigger has been identified for the decline, so the licensing discussion does not explain it. The distinction that matters for fundamentals is unchanged: the infrastructure partnerships map out intended deployments, Huang's pricing remark points to a possible additional source of income, and neither yet supports firm earnings estimates.
Ad
Nvidia Stock: New Analysis - 29 October
Fresh Nvidia information released. What's the impact for investors? Our latest independent report examines recent figures and market trends.
