Hugging Face is an AI company that runs the Hugging Face Hub, an open platform where people share machine learning models, datasets, and demo apps. It also maintains the Transformers library and many other open-source tools. On September 3, 2026, NVIDIA announced an agreement to acquire it for $12.93 billion, while promising the platform will stay open and not require NVIDIA hardware.
Hugging Face at a glance
- Founded: 2016 in New York
- Founders: Clément Delangue, Julien Chaumond, Thomas Wolf
- Known for: the Hub, Transformers, Datasets, Spaces
- Scale: more than 18 million users sharing over 3 million models, according to NVIDIA
Why does Hugging Face matter?
Hugging Face sits where open AI research, developers, and businesses meet. Its Hub hosts public models, datasets, and Spaces, and also supports private collaboration for companies. According to NVIDIA, more than 18 million developers, researchers, and creators use it to share over 3 million models, 500,000 datasets, and 1 million apps, and more than 200,000 companies use it to find and deploy AI.
How did Hugging Face start?
Its first product was a chatbot app for teenagers, named after the hugging-face emoji. The turning point came after Google released BERT in 2018. Hugging Face quickly published an open-source PyTorch version, then shifted fully toward open machine learning tools in 2019.
The Transformers library became the symbol of that shift. A 2019 paper described it as a unified way for researchers and engineers to use state-of-the-art transformer models.
What are the key milestones?
- 2016: founded in New York around a chatbot idea.
- 2018 to 2019: pivot to open-source transformer tools.
- 2021: Spaces lets developers host interactive AI demos in the browser.
- 2022: the BigScience project, launched with partners, releases BLOOM, a 176-billion-parameter multilingual model built by hundreds of researchers.
- 2023: raises $235 million from investors including Salesforce, Google, Amazon, NVIDIA, and IBM.
- 2026: NVIDIA announces it will acquire Hugging Face.
Which Hugging Face tools should you know?
- Transformers: load, fine-tune, and run transformer models in PyTorch, TensorFlow, or JAX.
- Datasets: access and share datasets.
- Diffusers: image, video, and audio generation models.
- PEFT and TRL: efficient fine-tuning and reinforcement learning for language models.
- Safetensors: a safer format for storing model weights.
What does the NVIDIA deal change?
NVIDIA said developers will keep choosing their own models, frameworks, clouds, and hardware, and that NVIDIA compute will not be required. That promise matters, because the platform’s value comes from being neutral. The open-source community will be watching whether that neutrality holds.
How should teams use Hugging Face well?
- Read the model card first. It explains intended use, limits, and evaluation.
- Check the license. Open access does not always mean free commercial use.
- Pin versions. Record the exact model, dataset, and library versions you use.
- Keep sensitive work private. Use private repositories for company data.
- Test with your own data. Public benchmarks are a starting point, not a guarantee.
Frequently asked questions about Hugging Face
Is Hugging Face free?
Public models, datasets, and Spaces are free to browse and use under their licenses. Paid plans add private collaboration and enterprise features.
Who owns Hugging Face?
NVIDIA announced on September 3, 2026, that it agreed to acquire Hugging Face.
What is the Transformers library?
An open-source Python library that makes it easy to use and fine-tune pretrained transformer models.
The takeaway
Hugging Face grew by spotting bottlenecks in AI work and turning them into shared tools. Its biggest contribution may be cultural: it made sharing models and data feel normal. Whatever the NVIDIA deal brings, that idea of open building blocks is what made it important.
