With over twelve million users and millions of open AI models, Hugging Face may well be a bigger player than OpenAI or Anthropic. What role does the platform play in the AI ecosystem?
Hugging Face has been around for almost ten years and has since become a central player in the global AI ecosystem. Names like OpenAI, Google and Anthropic are now widely known to the general public, but Hugging Face positions itself more as an open platform for AI builders.
At AWS Re:Invent 2025, we speak with Jeff Boudier, product lead at the company. How is the role of the platform evolving for its 12 million users? “Hugging Face is the leading open platform for AI builders,” says Boudier. “Every day, more than twelve million data scientists, researchers, machine learning engineers, and increasingly software developers are using our platform.”
Hugging Face is all about openness. Users can find, share, and use AI models, datasets, and applications there. This is different from commercial providers who mainly focus on a small number of proprietary models. “If you want to use an open model, you use Hugging Face,” says Boudier. “That’s where all these models are hosted and shared.” Compare Hugging Face a bit with Github, but for AI.
read also
Hugging Face: the open alternative in the AI landscape
Millions of models
One of the main differences with the major AI brands lies in scale and diversity. “OpenAI might offer five models,” says Boudier. “We offer millions.” There are more than two million public models available on the Hugging Face platform. In addition, there are millions of private models developed within organizations. “There are more than 300,000 organizations active on Hugging Face that collaborate in teams on private models.”
That enormous amount of models also translates into a wide range of applications. According to Boudier, it is a misconception to equate AI with chatbots and large language models (LLM’s). “A false image has emerged that AI is always LLM’s, AI is much broader than that. LLM is just one use case, and perhaps the most boring one.”
AI models for text, speech, image, video, and even 3D data can be found on Hugging Face. They are even often specialized per sector. “If you are building an application for lawyers, you obviously want a model that is exceptionally good at legal content,” he explains.
From research platform to development environment
When Boudier started at Hugging Face five years ago, the focus was mainly on AI researchers. “It was then mainly the reference for people who built the large models,” he says. Since then, the platform has evolved strongly. First came data scientists who adapted models to company data, then machine learning engineers who had to scale up, and now Hugging Face is increasingly seeing software developers on the platform.
“AI has now become the standard way to build technology,” says Boudier. “You no longer start with a thousand lines of code, but with an AI model.” That shift explains why Hugging Face is also attractive to classic software teams today.
read also
Hugging Face: the open alternative in the AI landscape
Strategic collaboration with AWS
An important factor in that growth is the collaboration with Amazon Web Services. AWS was the first strategic partner of Hugging Face. “When I started, my first task was to set up that collaboration,” says Boudier. Hugging Face runs largely on AWS infrastructure, but the collaboration goes further than that.
Together, the companies are working on integrations with, among others, Amazon SageMaker. “We make it easy for AWS customers to use Hugging Face models,” says Boudier. This is done via ready-made containers, SDKs and direct links from the Hugging Face platform to AWS services.
The collaboration is strong in terms of hardware. “We work very intensively with the teams behind Inferentia and Trainium,” says Boudier. “So that our models run optimally on AWS chips and customers benefit from the efficiency of this.”
Security and responsibility
Openness also entails risks. Hugging Face therefore invests heavily in the security of public models. “We take responsibility for securing public models,” says Boudier. Every model that is online is automatically scanned for malware, leaked API keys and personal data. Hugging Face works together with companies such as Protect AI, JFrog and VirusTotal for this.
AI creates value in millions of small ways today.
Jeff Boudier, AWS Product Lead
For companies, however, part of the responsibility lies with themselves. Boudier warns against shadow AI. “We see companies with hundreds of users using models without IT having any visibility into this.” With Hugging Face Enterprise, companies can therefore set up a policy, access rights and audit logs.
Anyone can publish
Just like on GitHub, anyone can actually publish a model on Hugging Face. “It’s your content and you remain the owner,” says Boudier. Checks are always carried out after publication, not before. There are stricter selections for enterprise environments. For example, partners such as Dell and Microsoft show subsets of models to their customers that meet additional requirements regarding licenses and security.
“On Azure AI Foundry you only see models that have passed all scans,” says Boudier. This makes the platform suitable for regulated environments without completely abandoning the open character.
AI in all sectors
According to Boudier, there is no sector that does not benefit from AI. “All industries benefit,” he says. From quality control in factories with image recognition to legal analysis and document processing. “The use cases are often different from what you hear in the media.”
read also
From Workflow to Techno Beat: how Siemens Makes AI Work
That broad use is reflected in the community. “You see national libraries sharing models, but also car manufacturers and maritime transport companies,” he says. With twelve million users, Hugging Face is global and for various sectors.
Hype, bubble or sustainable reality?
Finally, the question of a possible AI bubble is discussed. Boudier acknowledges that there is overinvestment, especially driven by the dream of general AI. “There is certainly overinvestment in infrastructure,” he says. At the same time, he sees that apart from the daily reality. “AI creates value in millions of small ways today.”
New open models appear weekly on Hugging Face. “A ‘DeepSeek moment’ happens here every week,” he says. That puts the shockwaves on the stock exchange into perspective. According to Boudier, the infrastructure will eventually be used. “Not for one model, but for the millions of models that companies need to support all their technology.”
read also