Nvidia Pushes Open Models Into Robotics Amid Job and Software Fears

Nvidia chief executive Jensen Huang told attendees at the Computex trade show in Taiwan that recent progress in agentic AI systems is creating fresh demand for software rather than displacing existing companies. Huang argued that autonomous agents will require more tools than prior generations of software, provided those tools are designed for machine use. He directly addressed concerns that agentic systems would render traditional software firms obsolete, calling the current period an "incredible time" to build software.

Huang's remarks come as shares of several established software providers have fallen amid investor worries over AI substitution. Companies such as Salesforce and Atlassian have faced questions about whether large language models could automate core functions. Huang countered that agents increase the number of interactions with digital tools, potentially widening the market for well-structured software interfaces.

Venture capitalist Bill Gurley offered a similar historical perspective during a recent podcast discussion. Gurley noted that fears of widespread job loss from AI mirror warnings issued during the Industrial Revolution, when Pope Leo XIII cautioned that mechanization would harm workers and increase inequality. Gurley observed that those earlier predictions did not hold, as new industries and higher productivity ultimately raised living standards. He suggested current policy debates risk repeating the same oversight by focusing on displacement rather than adaptation.

Nvidia also unveiled hardware and models aimed at physical applications of AI. The Isaac Gr00t platform combines a Unitree humanoid chassis, tactile hands, and Jetson Thor processors with open software intended to accelerate robotics research. Separately, the company released Cosmos 3, an open world model trained on large volumes of multimodal data including video, audio, and action sequences. The model is designed to help developers simulate rare or hazardous scenarios for robots and autonomous vehicles.

These releases reflect Nvidia's broader push beyond semiconductors into foundational models and reference platforms. Partners including Agile Robots and Runway have joined an effort to customize the models for specific hardware needs. The open nature of the releases allows independent developers and manufacturers to adapt the tools without licensing restrictions typical of closed systems.

Market responses to these developments have varied. While some software equities declined on substitution concerns, companies positioned to supply the underlying compute and models have seen continued interest from investors. Historical patterns in prior technology shifts show that capital and labor tend to reallocate toward higher-value activities once initial disruptions subside.

Observers tracking employment data note that automation episodes have often coincided with net job growth in new categories, from computer programming to logistics coordination. Gurley emphasized that policy focused on preserving existing roles can slow the very adjustments that produce broader gains. Huang's comments align with this view by stressing the need for software that agents can readily access, implying continued demand for human-designed interfaces even as machine autonomy expands.

The Computex announcements underscore how advances in both digital agents and physical robotics depend on layered infrastructure of chips, models, and application software. Rather than a zero-sum replacement of existing firms, the emerging stack appears to multiply the points at which specialized tools can be deployed.