Hugging Face

NLP

Hugging Face is the central open-source AI platform for machine learning, founded in 2016 by Clément Delangue (CEO), Julien Chaumond, and Thomas Wolf in New York.

Reviewed by 7wData
Share Email

On this page

Profile

Hugging Face builds and operates the central platform where machine learning engineers discover, share, and deploy open-source AI models and datasets.

Hugging Face is the central open-source AI platform for machine learning, founded in 2016 by Clément Delangue (CEO), Julien Chaumond, and Thomas Wolf in New York. Originally conceived as a chatbot for teenagers, the company pivoted to building infrastructure for the AI community. Today it operates as "the GitHub of machine learning," hosting 2+ million open-source models, 500,000+ datasets, and 1+ million user applications (Spaces) on its Hub.

The Transformers library, Hugging Face's flagship open-source project, has grown to over 3 million daily installations and 1.2 billion total installs, with 400+ model architectures and 750,000+ compatible model checkpoints. In December 2025, Hugging Face released Transformers v5—a structural redesign emphasizing modularity and PyTorch-first simplification. The company reached $130M revenue in 2024, nearly double its 2023 ARR of $70M, primarily from cloud services, enterprise subscriptions ($20/month+ per user), and consulting contracts with major customers including Meta, Google, Amazon, Microsoft, Intel, and Nvidia.

Recent acquisitions underscore Hugging Face's expanding scope: Pollen Robotics (April 2025, adding open-source humanoid robots like Reachy 2) and GGML/llama.cpp (February 2026, bringing local inference infrastructure under one roof). In November 2025, Hugging Face deepened partnerships with Google Cloud (TPU support, cached model downloads) and Meta (OpenEnv, a standardized framework for AI agent environments). CEO Delangue maintains that Hugging Face has retained roughly half its $400M in total funding, positioning the company toward profitability rather than pursuing aggressive spending. The company maintains roughly 250 employees after a 4% layoff of its consulting-services team in early 2025, a reduction Delangue attributed to shifting focus from one-time contracts to recurring API usage.

Track Hugging Face and 240+ vendors.

335k+ subscribers read the daily AI & data note. One email, both newsletters. Unsubscribe anytime.

Who buys this

  • AI researchers and engineers building and fine-tuning models
  • Enterprise companies embedding open-source models into production applications
  • Robotics developers using open-source hardware and AI integration
  • Cloud providers (Google Cloud, AWS, Azure) integrating Hugging Face models into their platforms
  • Developers using local inference (llama.cpp) for on-device AI applications

Publicly disclosed clients

  • Meta
  • Google
  • Amazon
  • Microsoft
  • Intel
  • Nvidia
  • Salesforce
  • Bloomberg
  • Grammarly

Strengths and what to watch

Strengths

  • De facto standard open-source model repository with 2M+ models and unmatched ecosystem adoption
  • Transformers library dominates model definition across training, inference, and deployment frameworks
  • Strategic control over llama.cpp and local inference stack; partnerships with Google Cloud and Meta position Hugging Face as AI infrastructure commodity

Watch for

  • Revenue concentration: $130M ARR built largely on Nvidia/AWS/Meta contracts and one-time consulting; recurring platform revenue less clear
  • Robotics expansion (Pollen, Reachy) is small-scale pilot; market demand and unit economics for consumer/industrial robots unproven
  • 4% layoff of consulting team signals pivot away from high-touch services, but recurring model-hosting revenue per user remains undisclosed

Recent moves

Key Information

Industry
NLP
Founded
2016
Headquarters
New York

Frequently Asked Questions

What is Hugging Face?

Hugging Face is the central open-source AI platform founded in 2016, hosting 2+ million models, 500,000+ datasets, and 1+ million applications. It serves as the GitHub of machine learning, enabling engineers to discover, share, and deploy AI models and datasets at scale.

How many models does Hugging Face have?

Hugging Face hosts 2+ million open-source machine learning models and 500,000+ datasets on its Hub. The platform also runs 1+ million user applications called Spaces, making it the largest ecosystem for discovering, sharing, and deploying AI models and datasets at scale.

What is the Transformers library?

The Transformers library is Hugging Face's flagship open-source project, used daily by 3 million developers with 1.2 billion total installations. It features 400+ model architectures and 750,000+ compatible checkpoints, enabling engineers to train, fine-tune, and deploy advanced AI models efficiently.

What is Transformers v5?

Transformers v5 was released in December 2025 as a major structural redesign of Hugging Face's flagship library. It emphasizes modularity and PyTorch-first simplification, maintaining 3 million daily installations while enabling developers to build, train, and deploy AI models more efficiently.

What are Hugging Face's recent acquisitions?

Hugging Face acquired GGML and llama.cpp in February 2026 to consolidate local inference infrastructure. This strategic move brings efficient, on-device AI capabilities under one umbrella, enabling developers to deploy models locally while maintaining the projects' open-source nature and full community autonomy.

Who are Hugging Face's main customers?

Hugging Face serves AI researchers, enterprise companies, robotics developers, and cloud providers. Major customers include Meta, Google, Amazon, Microsoft, Intel, Nvidia, Salesforce, Bloomberg, and Grammarly. The platform supports model training, fine-tuning, deployment, and local inference across research, production, and edge applications.

How Hugging Face compares

Direct head-to-head against 3 competitors. Picked by 7wData.

This company

Hugging Face

Positioning
Hugging Face builds and operates the central platform where machine learning engineers discover, share, and deploy open-source AI models and datasets.
Customer segments
AI researchers and engineers building and fine-tuning models
Strengths
De facto standard open-source model repository with 2M+ models and unmatched ecosystem adoption
Watch for
Revenue concentration: $130M ARR built largely on Nvidia/AWS/Meta contracts and one-time consulting; recurring platform revenue less clear
Recent moves
Transformers v5 released: major redesign toward modular, PyTorch-first library with 3M daily installs

Replicate

Positioning
Serverless API for running open-source ML models, absorbed into Cloudflare's developer platform after November 2025 acquisition.
Customer segments
App-layer engineers and product developers deploying pre-trained models via API, favoring pay-per-prediction pricing.
Strengths
50,000-plus open-source models accessible via a single API call with no GPU infrastructure to configure.
Watch for
Cloudflare acquisition subordinates ML roadmap to edge-platform priorities; community-focused model-hosting features risk deprioritization post-integration.
Recent moves
Acquired by Cloudflare, announced November 17, 2025, folding model inference into Cloudflare's global edge network.

Together AI

Positioning
AI-native cloud for open-source model inference and fine-tuning, targeting AI-native startups and global enterprises.
Customer segments
AI-native companies and enterprise engineering teams (Salesforce, Zoom, SK Telecom), plus 450,000-plus independent developers globally.
Strengths
Dedicated Blackwell GPU clusters with enterprise SLAs for open-source model inference and fine-tuning at scale.
Watch for
200 MW GPU capacity commitments against $305M raised; burn rate and profitability timeline remain undisclosed.
Recent moves
$305M Series B at $3.3B valuation closed February 2025; Series C talks at $7.5B valuation reported March 2026.

Weights and Biases

Positioning
ML experiment tracking and model registry, now folded into CoreWeave's GPU cloud after May 2025 acquisition.
Customer segments
ML engineers at model-building organizations and AI research teams, including OpenAI, Meta, NVIDIA, and Toyota.
Strengths
Experiment tracking that links model versions to training run lineage, adopted across major AI labs worldwide.
Watch for
CoreWeave acquisition creates compute-stack lock-in; neutral multi-cloud positioning is no longer credible for enterprise buyers.
Recent moves
Acquired by CoreWeave for $1.7B, completed May 5, 2025, vertically integrating model tracking with GPU compute.

Sources

  1. huggingface.co — Company mission, products (Hub, Transformers, Diffusers, Inference Endpoints), customer roster (Meta, Google, Amazon, Microsoft, Intel), community scale (5M users, 50k organizations)
  2. sacra.com — Revenue trajectory ($70M ARR 2023, $130M 2024), Series D $235M funding round August 2023, valuation $4.5B, customer count (50k organizations, 13M users)
  3. techcrunch.com — CEO Clem Delangue's recent market commentary, financial strategy (retained ~50% of $400M funding), strategic direction toward specialized models
  4. techcrunch.com — Pollen Robotics acquisition April 2025, Reachy 2 robot specifications ($70k), founding of Pollen team (Matthieu Lapeyre, Pierre Rouanet)
  5. huggingface.co — GGML/llama.cpp acquisition February 2026, leadership (Georgi Gerganov), commitment to maintain llama.cpp open-source and community autonomy
  6. huggingface.co — Google Cloud partnership November 2025, TPU support, CDN gateway for model downloads, security enhancements, usage scale (petabytes downloads monthly)
  7. huggingface.co — Transformers v5 December 2025 release, 3M daily installs, 1.2B total installs, 400+ model architectures, 750k checkpoints, modular redesign, PyTorch focus
  8. www.theinformation.com — 4% layoff February 2025 (~10 employees from 250-person company), Expert Support Program consulting service reduction