Nvidia CEO Advocates for Balance in Open-Closed AI Model Development
Nvidia CEO Jensen Huang is clear about one thing: if you're running an open AI model, it's likely on Nvidia hardware. In fact, he claims that there's only one platform capable of running every frontier model, open or closed.
Huang has been advocating for a balance between open and closed models in AI development. He co-signed an open letter with around 25 organizations last July, calling for the promotion of both open and closed model development to drive competition and innovation.
Nvidia has put its words into action by releasing several open-source AI models under permissive licenses. The Nemotron language model family, Cosmos for physical AI, and Nemotron 3.5 Lightning are just a few examples. These models provide developers with access to weights and training resources.
Huang sees open-source AI adoption as a key driver of enterprise GPU server deployments. Companies in SaaS, robotics, and industrial applications are building custom AI solutions rather than relying on closed model providers. This shift allows them to modify, deploy, and run models on their own infrastructure, not just in the cloud.