Hugging Face

Hugging Face is the leading open-source AI platform and community, often called the "GitHub of machine learning." Founded in 2016 by Clément Delangue, Julien Chaumond, and Thomas Wolf, the company has become the central hub where the AI community shares models, datasets, and applications.

The Hugging Face Hub hosts millions of models, more than a million datasets, and hundreds of thousands of demo applications (Spaces), making it the largest repository of open-source AI assets in the world. The company's Transformers library is the most widely used Python library for working with pre-trained models, and its ecosystem includes libraries for training (Accelerate), evaluation (Evaluate), and deployment (Inference Endpoints). Virtually every major AI research paper releases its models on Hugging Face.

Hugging Face has raised over $400 million, with investments from Google, Amazon, NVIDIA, and Salesforce. The company generates revenue through enterprise subscriptions (private model hosting, dedicated inference), while keeping its core community platform free. Hugging Face's role as the neutral repository for the entire AI ecosystem — hosting models from OpenAI, Meta, Google, and thousands of independent researchers — gives it an unmatched position in the AI developer community — a position that rests on staying independent. The company has retained a bank to gauge acquisition interest, with no deal reached and no buyer named.

Hugging Face is also now the corporate home of the local-inference stack. On February 20, 2026 it acquired ggml.ai, the company Georgi Gerganov founded in 2023 to sustain ggml and llama.cpp — the MIT-licensed C/C++ engine that most on-device AI runs on, including Ollama and LM Studio, and whose GGUF file format is the de facto standard for distributing quantized model weights. Gerganov and his team joined Hugging Face, which committed to the projects staying "100 percent open-source and community driven" with the team keeping full autonomy over technical direction. It is a significant strategic position: alongside hosting the models, Hugging Face now employs the maintainers of the software a large share of the ecosystem uses to run them. It also sharpens the ownership question above, since the same acquisition interest that would change Hugging Face's owner would change llama.cpp's corporate parent for the second time in a year.

The company's ownership was settled in September 2026. After Business Insider reported in August that it had retained a bank to gauge takeover interest, NVIDIA confirmed on September 3, 2026 that it is acquiring Hugging Face for $12.9 billion — NVIDIA's second-largest deal ever, after the 20-billion-dollar Groq asset purchase in December 2025. The deal is agreed rather than closed, with completion expected in the first half of 2027. Chief executive Clément Delangue told CNBC that he approached Jensen Huang over the summer, having concluded that open-source AI was at a turning point and needed more scale; Huang wrote that the platform will remain open to the entire ecosystem and that NVIDIA compute will not be required to build or deploy on it. The tension is neutrality rather than price: the platform's value rests on serving every vendor, including NVIDIA's competitors AMD and Intel, and Hugging Face turned down a $500 million investment from NVIDIA in 2025 precisely because it did not want a dominant investor able to sway its decisions.