At Nvidia’s GTC conference, Jensen Huang leaned hard into a new narrative: AI is moving from chatbot demos to agentic software that can actually control systems. Reuters reports that Huang used the event to showcase OpenClaw, an open-source agent-control technology he framed as a kind of “new computer,” comparing its potential role to the old Windows-era software layer. The pitch is simple and ambitious at the same time: if models are becoming commodities, the real power may shift to the stack that lets them take action.

That framing matters because it expands Nvidia’s role beyond chips. The company already dominates the compute layer, but tools like OpenClaw and the enterprise-friendly NemoClaw push it deeper into orchestration, agent workflows, and enterprise AI operations. Reuters notes that Nvidia also used the event to reinforce its position as the infrastructure backbone for hyperscalers and major AI players, even as competition from AMD and in-house chips keeps growing.

For the market, this looks like Nvidia trying to own not only the hardware that runs AI, but also part of the software logic that makes AI usable at scale. If that works, Nvidia stops being just the company selling picks and shovels and starts looking more like a platform company for the agent era.