Startup Hugging Face Inc., which operates a platform for hosting open-source artificial intelligence projects, today announced that it has raised $235 million in funding.

The Series D round included contributions from many of the tech industry’s most prominent players. Amazon.com Inc., Google LLC, IBM Corp. and Salesforce Inc. were among the participants. They were joined by chip giants Intel Corp., Nvidia Corp., Qualcomm Inc. and Advanced Micro Devices Inc. along with Sound Ventures.

Reports that Hugging Face could raise new capital first emerged earlier this week. On Tuesday, The Information reported that Salesforce was the lead investor in the startup’s latest raise. The new investment values Hugging Face at $4.5 billion, more than twice what it was worth after its previous funding round last May. 

Hugging Face operates a platform of the same name that developers use to share open-source AI models. It has an interface similar to GitHub. The Hugging Face page for an AI model is organized into three main sections: One contains a description of the model, another offers access to its code and the third includes a user forum.

Developers rely on Hugging Face to share not only AI models but also related technical assets. According to the company, its platform hosts thousands of training datasets that companies can use to build new neural networks. It also offers access to so-called Spaces, sample applications that demonstrate practical uses for the AI models on its platform. Overall, Hugging Face hosts more than 120,000 AI models and 20,000 datasets.

The company makes money by offering a paid version of its platform called Enterprise Hub. The offering allows companies to create an internal Hugging Face environment for their software teams. Developers can use that environment to share AI models and related assets with colleagues, as well as access files from the public version.

Alongside Enterprise Hub, the company offers two other commercial products that focus on easing AI model development.

The first offering is called AutoTrain. According to Hugging Face, users can upload a training dataset to AutoTrain and have it automatically find the most suitable open-source AI model for their project. After selecting a neural network, the service trains it on the user-provided dataset.

Developers can upload the AI models they create with AutoTrain to a Hugging Face service called Inference Endpoints. It provides a managed cloud environment for hosting neural networks. The service spares users the manual work usually involved in setting up AI inference infrastructure.

Co-founder and Chief Executive Officer Clement Delangue told The Verge that the startup will use its latest funding for “more open source AI and platform building.” The startup will expand its 170-person workforce to advance the effort. 

Image: Hugging Face