Artificial Intelligence · Profile from AI Research Companion
Hugging Face
The New York- and Paris-based company that began in 2016 as a chatbot app for teenagers, open-sourced a small library for running Google's BERT model in 2018, and turned it into the default home of open machine learning: the transformers library, the Hugging Face Hub (more than 3 million public models, 500,000 datasets and 1 million demo applications by September 2026, per Nvidia's announcement) and a community of more than 18 million developers. It raised ~$395m in total, the last round a $235m Series D at $4.5bn led by Salesforce Ventures with Google, Amazon, Nvidia, Intel, AMD, Qualcomm and IBM all on the cap table (August 2023). It became the reference case for the economics of neutral open infrastructure: enormous reach, a free product most people never pay for, and revenue (an estimated ~$70m annualised at end-2023, Sacra; reported at ~$150m annualised in mid-2026, The Information) that was tiny next to its influence. On September 2, 2026 it signed a definitive agreement to be acquired by Nvidia for ~$12.93bn in total, ~$11.9bn payable to stockholders plus up to ~$1.0bn of retention equity for employees (Nvidia 8-K), with closing expected in the first half of 2027 subject to regulatory approval; the US Justice Department was reported in late October 2026 to be asking questions about the deal (Reuters, reported).
At a glance
- Founded
- 2016 in New York by Clément Delangue (CEO), Julien Chaumond (CTO) and Thomas Wolf (chief science officer), three French entrepreneurs
- Headquarters
- New York (Brooklyn) and Paris; staff distributed worldwide, largely remote
- Ownership / structure
- Private, venture-backed (Lux Capital, Sequoia, Coatue, Addition, Salesforce Ventures, Google, Amazon, Nvidia, Intel, AMD, Qualcomm, IBM and others). Agreed to be acquired by Nvidia (definitive agreement September 2, 2026; pending)
- Leadership (2026)
- Clément Delangue (co-founder, CEO); Julien Chaumond (co-founder, CTO); Thomas Wolf (co-founder, chief science officer)
- Products
- The Hub (model, dataset and "Spaces" app hosting with Git-based versioning); open libraries (transformers, datasets, diffusers, accelerate, peft, trl, tokenizers, safetensors, text-generation-inference, smolagents, LeRobot); Gradio (demo apps); paid tiers (PRO accounts, Team and Enterprise Hub); Inference Endpoints and Inference Providers (routing to partner clouds); HUGS (Hugging Face generative AI services for cloud marketplaces); own models (BLOOM via BigScience, StarCoder via BigCode, Zephyr, SmolLM, Idefics, SmolVLA); robots (Reachy Mini, HopeJR) after buying Pollen Robotics
- Scale
- >3m public models, >500,000 datasets, >1m applications, >18m users, >200,000 companies using the platform (Nvidia, September 2026); ~250 staff in 2024 (company posts, verify current headcount)
- Market value
- $2bn (Series C, May 2022); $4.5bn (Series D, August 2023); ~$12.93bn total consideration in the agreed Nvidia acquisition (September 2026)
- Signature moves (incl. failures)
- Chatbot app to transformers library (2016–19); Gradio, Spaces and the Hub as "GitHub for AI" (2021–22); BigScience and BLOOM (2021–22); Series D and the rivals' cap table (2023); cloud partnerships and monetising the Hub (2023–25); security breaches and malicious models (2024–26); Pollen Robotics and open hardware (2025); the Nvidia acquisition (2026)
8 strategic moves analysed
Each one gets a full strategy file in the companion: context, the decision, the financing, what happened, and the lesson.
2016
Chatbot App to Transformers Library
Founding / pivot
2021
BigScience BLOOM and Open Science
Model release / research collaboration
2021
Gradio Spaces and the Model Hub
Acquisition / launch
2023
Cloud Partnerships and Monetising the Hub
Partnership / pricing
2023
Series D and the Rivals Cap Table
Funding round
2024
Security Breaches and the Trust Problem
Incident / failure and response
2025
Pollen Robotics and Open Hardware
Acquisition / product launch
2026
Nvidia Acquisition
Acquisition (as target)