Nvidia CEO Jensen Huang says market misreads software, outlines why

Jensen Huang’s comments come at a moment when Wall Street analysts are wrestling with divergent earnings forecasts for the enterprise‑software sector. In a CNBC interview aired this week, Huang pushed back against the prevailing narrative that generative‑AI agents will cannibalize traditional SaaS licences. He argued that “AI agents will actually drive more usage of software,” suggesting that the next wave of automation will sit on top of, rather than replace, existing applications (Capital Aidaily). The nuance in his stance is that the underlying codebases, data pipelines, and integration points remain essential, and that the “agentic layer” will act as a productivity multiplier for end‑users.

The CEO’s perspective also hints at a broader strategic implication for Nvidia’s own ecosystem. By positioning AI agents as demand generators for software, Huang reinforces the value proposition of Nvidia’s GPU and AI‑accelerator portfolio, which powers the inference workloads behind those agents. If enterprises indeed embed Nvidia‑optimized models into their existing ERP, CRM, and analytics stacks, the resulting increase in compute consumption could translate into higher sales of Nvidia’s data‑center chips—a trend that investors have already begun to price into the company’s forward‑looking guidance.

Huang’s remarks, however, run counter to a growing chorus of investor commentary that warns of “software disruption” as AI models become capable of autonomous decision‑making. The market’s concern is that firms like Salesforce or ServiceNow could see a contraction in licence renewals if AI agents automate routine tasks traditionally handled by those platforms. By asserting that usage will rise, Huang is essentially betting that the integration cost and learning curve of deploying agents will be outweighed by the efficiency gains they unlock, prompting firms to expand—not shrink—their software footprints.

The interview also underscores the competitive dynamics shaping the AI‑software interface. While Nvidia touts its hardware advantage, rivals such as Microsoft and Google are rolling out their own agent‑centric services built on proprietary clouds. Huang’s confidence that the “software market’s got it wrong” implicitly acknowledges that Nvidia must continue to deepen partnerships with ISVs to embed its chips into the emerging agent stack. Without such collaborations, the company risks being sidelined as cloud providers deliver end‑to‑end solutions that bypass Nvidia’s hardware altogether.

In short, Huang’s thesis reframes the AI‑software debate from a zero‑sum game to a symbiotic growth story. By insisting that agents will amplify software consumption, he signals to investors that Nvidia’s revenue trajectory remains tied to expanding compute demand across the enterprise stack, rather than being threatened by a speculative wave of software displacement. The market will now watch the next earnings season for evidence that software usage metrics are indeed climbing in tandem with Nvidia’s hardware sales, a correlation that could validate—or refute—Huang’s contrarian outlook.