Nvidia Is Building a Bigger Open-Source AI Bet

This article first appeared on GuruFocus.

Nvidia Corp. (NVDA, Financials), the dominant maker of artificial intelligence chips, is expanding its push into open AI models as it looks for another way to increase demand for its hardware.

The company launched Nemotron 3.5 Lightning, a 30-billion-parameter model designed for long-running agentic AI workloads.

Nvidia also introduced NeMo Switchyard, an open-source routing tool that can direct AI requests to the most suitable model without forcing developers to rebuild their applications. The bigger story may be what comes next.

Nvidia is reportedly developing Nemotron 4, with its largest version potentially reaching at least 1 trillion parameters. The company wants the model family to compete more closely with leading open-source systems.

That creates an unusual dynamic because Nvidia is increasingly competing with some of the AI developers that also rank among its largest chip customers.

But the strategy has a clear economic logic. More capable open models could encourage companies to deploy more AI locally and in data centers, creating additional demand for Nvidia GPUs.

For investors, Nemotron is increasingly looking less like a side project and more like another way Nvidia can expand the market for its chips.