The AI community building the future — the de facto 'GitHub of machine learning' hosting 3M+ models
HQ: New York, NY (remote-first; Paris office) | Founded: 2016 | Employees: ~900-990 | Stage: Acquired by NVIDIA (Sept 2026, $12.93B; closes H1 2027) | Website: https://huggingface.co
Hugging Face is an American (French-founded) AI community and platform company that operates the largest open-source catalog of machine learning models, datasets, and AI applications in the world — hosting 3M+ models, 500K+ datasets, and 1M+ demo apps used by 18M+ developers and 200,000+ organizations. Founded in New York in 2016 as a chatbot startup named after the 🤗 emoji, it pivoted after open-sourcing a PyTorch port of Google's BERT in 2018, which became the enormously popular Transformers library. Today its full-stack offering spans open-source libraries (Transformers, Diffusers, Datasets, Tokenizers, TRL, Gradio, smolagents, LeRobot, safetensors), the Hugging Face Hub, hosted inference (serverless API, Inference Endpoints, Inference Providers marketplace), AutoTrain fine-tuning, and Team/Enterprise Hub governance tiers. In September 2026, it agreed to be acquired by NVIDIA for ~$12.93B, with commitments to keep the platform open, multi-cloud, and multi-accelerator.
Hugging Face executed one of the cleanest open-source community moats in software history — becoming the de facto 'GitHub of ML' by giving models away and monetizing only compute, storage, and enterprise governance. Its ~$150M ARR is small relative to its strategic gravity (3M+ models, 18M+ builders), which is exactly why NVIDIA agreed to pay ~86x revenue to acquire it for $12.93B rather than try to displace it. The deal crystallizes the company's founding fragility: the neutrality that made it trusted is being sold to the largest silicon vendor in the stack, and both regulators and the open-source community will watch whether 'open platform' holds below the surface.
Browse: Home | Blog | About | All Startups A-Z | View full profile on StartupWiki