Nvidia keeps its open-source course and $12.9 billion Hugging Face deal after Huang's exit
Interim Nvidia chief Colette Kress told analysts on the earnings call, after Jensen Huang's resignation, that the company's bet on open AI models, which can be downloaded and fine-tuned, is not changing. The $12.9 billion acquisition of Hugging Face, a model hub for 118 million developers, remains in force. If your product runs on Nemotron or models from Hugging Face, you don't need to change your stack urgently. But it is wise to keep your own copies of the weights.
AI-processed from TechCrunch; edited by Hamidun News
When a founder who actively championed open AI leaves, developers ask a simple question: what happens to the tools they depend on? Picture a team whose config specifies Nvidia's Nemotron model, with weights pulled from Hugging Face. If Nvidia decided after Jensen Huang's resignation to close its models or change the platform's terms, that product would need an urgent rebuild. Nvidia tried to answer exactly this fear on its first earnings call without Huang. Interim CEO Colette Kress said in her opening words that the company still intends to grow a "living ecosystem of open models, so that every country, company and industry can build its own intelligence."
What Kress actually said
Open models are those whose weights (trained parameters) are published, so anyone can download them, run them locally and fine-tune them. Kress cited figures that Nvidia uses to justify its course: weekly token volume on the OpenRouter aggregator grew 35x over the year, including 3.5x in the last three months alone. A token here is the unit of text a model reads or generates, so the growth shows how much more actively developers are turning to open models. According to Kress, this proves that such models speed up AI adoption.
A caveat: OpenRouter is just one service where developers can easily switch between models. It reflects the behavior of some developers, not the whole market. Still, as an indicator of interest in open weights, the figure is telling.
Kress also singled out the Nemotron family, open models that Nvidia builds for long-running agentic systems and reasoning tasks. Huang promoted this line of work actively inside the company. According to Kress, Nemotron is now "the most used American open model on OpenRouter" and the place "where developers start building their own intelligence." This is the company's own claim, with no independent verification in this piece, and the word "American" narrows the comparison.
Hugging Face stays in the deal
The main question for the community is the fate of Hugging Face. It is the platform where open models are published and downloaded; in September Nvidia agreed to buy it for $12.9 billion. That is the second-largest acquisition in Nvidia's history. The platform serves more than 118 million developers and 200 thousand companies, and a significant share of open models, datasets and tools lives there.
Kress said the two companies together will "scale the platform and make intelligence more open, more powerful and more accessible." For now, the plans look unchanged. But analysts did not discuss the deal directly in the Q&A session, so no one has received any guarantee of the platform's independence. After Huang's resignation, the head of Hugging Face called what happened a "sad day" and promised to "fight for open AI even harder."
The background you can't ignore
Huang stepped down last week after reports about his extramarital affairs, including a relationship with an executive at the Singapore company Megaspeed, which is suspected of working for Chinese intelligence. He is also alleged to have had relationships with two direct reports. He is due to testify in Congress before Thanksgiving, and the SEC, as WSJ reports, is examining whether Nvidia's disclosures misled investors. All of this is a risk for the company, and the situation around it is far from calm.
What to do in practice
If your product depends on Nemotron, Hugging Face Hub or open weights in general, there is no urgent reason to change your stack: the company's public position has not changed, and the deal is formally in force. But words on an analyst call are not a contract. Sensible insurance is simple:
- don't tie yourself to a single model or a single storage location; keep a fallback;
- pin weight versions in your config and copy critical files to your own storage;
- watch the terms of the deal's closing and any changes to Hugging Face's licenses and rules.
For Nvidia, open models are not charity: the more developers build on its models, the more demand there is for its GPUs. That calculation did not change with the CEO, which is why the course was confirmed so quickly. The real test will come when the deal closes and it becomes clear how independent the platform that hosts open AI today will remain.
- Nov 20, 2026Anthropic, Maker of Claude, Raises $81.3 Billion in IPO at a $1.5 Trillion Valuation
- Nov 20, 2026Anthropic raises $81.3 billion in IPO, beating SpaceX's $75 billion record
- Nov 20, 2026Anthropic goes public and raises $81.3 billion — shares jump 17% on day one
- Nov 20, 2026AMD Launches EPYC Venice: 512 Threads in a Single Chip and 2.7x Faster Memory
Key Questions Answered
What Kress actually said?
Open models are those whose weights (trained parameters) are published, so anyone can download them, run them locally and fine-tune them. Kress cited figures that Nvidia uses to justify its course: weekly token volume on the OpenRouter aggregator grew 35x over the year, including 3.5x in the last three months alone. A token here is the unit of text a model reads or generates, so the growth shows how much more actively developers are turning to open models.
What are the key details on Hugging Face stays in the deal?
The main question for the community is the fate of Hugging Face. It is the platform where open models are published and downloaded; in September Nvidia agreed to buy it for $12.9 billion. That is the second-largest acquisition in Nvidia's history. The platform serves more than 118 million developers and 200 thousand companies, and a significant share of open models, datasets and tools lives there.
What are the key details on The background you can't ignore?
Huang stepped down last week after reports about his extramarital affairs, including a relationship with an executive at the Singapore company Megaspeed, which is suspected of working for Chinese intelligence. He is also alleged to have had relationships with two direct reports. He is due to testify in Congress before Thanksgiving, and the SEC, as WSJ reports, is examining whether Nvidia's disclosures misled investors. All of this is a risk for the company, and the situation around it is far from calm.
Test Frontier LLMs & Autonomous Agents in One Workspace
Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.