Technology

Nvidia Says Its Open-Model Strategy Holds After Huang's Exit

Martin HollowayPublished 5m ago3 min readBased on 1 source
Reading level
Nvidia Says Its Open-Model Strategy Holds After Huang's Exit
Photo by Thomas Foster on Unsplash

Nvidia says its open-source AI strategy remains intact following the resignation of Jensen Huang as chief executive. Collette Kress, the company's interim CEO, led Nvidia's first earnings call with analysts since Huang's departure and reaffirmed the company's commitment to open models in her opening remarks, according to TechCrunch.

An open model is one whose trained weights, the learned parameters that define its behavior, are published so others can download and run it. Kress said Nvidia is committed to fostering a vibrant open model ecosystem, so that every country, company, and industry can build its own intelligence. She tied that commitment to usage data from OpenRouter, a model aggregation service that routes developer requests to many different models through one interface. She cited a 35-fold year-over-year surge in weekly token volume on the platform, and a 3.5-fold increase over the past three months. Tokens are the chunks of text a model reads and writes, so token volume is a rough gauge of how much work the models are doing. She also said Nemotron, Nvidia's model family, is now the most widely used U.S. open model on OpenRouter.

The call also addressed Hugging Face, the open-source hub where developers share and download models, datasets, and tooling. Nvidia agreed in September 2026 to acquire the company for $12.9 billion, a transaction that would be the second-largest acquisition in Nvidia's history. Hugging Face serves more than 118 million developers and 200,000 companies. Nvidia's plan for Hugging Face to join Nvidia remains on track despite Huang's departure. Kress said the two companies together will scale the platform and make intelligence more open, more capable, and more accessible.

The acquisition was not directly addressed during the question-and-answer portion of the call.

The broader context here is the distinction between usage and revenue. Token volume through a model router measures how much inference traffic, the work of running a trained model on new input, passes through a shared gateway. It indicates developer adoption of the models being routed, but it does not on its own show how that traffic is monetized or which hardware serves it. Those are separate questions, and the call's stated figures answer the first more directly than the second.

In my view, the more consequential question for developers is governance rather than growth. Hugging Face has become a common distribution point for open weights, datasets, and tooling, and its neutrality is part of why teams rely on it. Whether that neutrality survives a change of ownership will depend on decisions that have not yet been described in detail. Kress's framing that the combined company will widen access is a stated objective, and the terms on which it is pursued will matter to the engineers who depend on the platform.

The near-term signal is that Nvidia is positioning open models as a core part of its AI platform strategy, during a leadership transition. If the stated plans hold, a large model hub, a widely used open model family, and Nvidia's compute stack could lower the cost of building domain-specific systems for organizations without their own training infrastructure. That is the outcome the company has described. Whether it arrives in the form Kress outlined will become clearer as the acquisition moves toward completion.