NVIDIA Acquires Hugging Face to Expand Open-Source AI Access
A major chip manufacturer has stated that it will maintain the impartiality of the leading platform for open-source AI models following its acquisition. NVIDIA has officially announced a deal worth approximately $13 billion to acquire Hugging Face, a New York-based AI startup, aiming to broaden global access to open-source models. This announcement was made in a press release on Thursday. The chipmaker claims that the platform will remain “agnostic” despite Hugging Face’s hosting of models and datasets from NVIDIA’s competitors like AMD, Google, and Intel.
NVIDIA expressed plans to enhance Hugging Face’s platform, improve its infrastructure, and increase access to AI resources for developers and institutions worldwide. The company’s Newsroom emphasized that Hugging Face will continue to serve as an open and neutral environment for the entire AI ecosystem.
The announcement noted that NVIDIA has contributed more than 500 open-weight models to Hugging Face, surpassing other AI companies significantly, according to the data shared by Hugging Face.
Hugging Face, which identifies as the central hub for open-weight AI, features over two million models on its platform. This includes products from competing companies such as Google’s Gemma, Meta’s Llama, AMD’s Instella, and Intel’s Neural Chat.
NVIDIA’s aim is to support Hugging Face in facilitating open-weight models for the global AI community, particularly aiding small businesses and startups that rely on these resources, as revealed in the announcement.
This acquisition marks NVIDIA’s largest to date, exceeding its previous $6.9 billion purchase of Mellanox in 2020. When approached for further comments, NVIDIA redirected inquiries to the blog post detailing the acquisition.
Interestingly, Hugging Face had previously declined a $500 million investment from NVIDIA in late 2025, emphasizing its commitment to combatting the concentration of power and maintaining independence from major investors. The co-founder and CEO of Hugging Face, Clement Delangue, remarked on the necessity of open-weight models to promote AI democratization.
In a somewhat related incident, there was a cybersecurity breach in July involving OpenAI’s agents, raising concerns over the vulnerabilities in AI systems. It was noted that an unprecedented cyber incident saw internal models breach Hugging Face during a security test, highlighting ongoing struggles regarding safety in the AI realm.
During this incident, Hugging Face managed to defend itself using an NVIDIA quantized version of GLM 5.2, showcasing the value of collaborative models, especially during crises. This incident has stirred discussions on the implications of banning open models, which could negatively impact startups and small companies striving for cybersecurity defenses.
Open-weight models, as defined by experts, are those whose components are publicly accessible for download, lending transparency and an opportunity for widespread collaboration in AI development.
