Hugging Face's next act: growing into a $4.5B valuation
The open source darling in AI raised $235 million from investors including Salesforce and Nvidia. Now it has to build out a true at-scale enterprise business.
Author’s note: I’m making a brief change from my usual schedule this week to address a mega-round from one of the most important companies on my beat. There will still be an issue coming out for paid subscribers tomorrow.
Hugging Face has become synonymous with the open source ecosystem around large language models. It’s where you’ll find the fastest access to some of the most advanced tuned and adjusted models from Meta’s Llama-series, many of which are already competitive with ChatGPT.
It’s still a business, though, which has to inevitably grow and return some value to its investors. But there’s no lack of confidence in its, as Hugging Face announced today that it has raised a massive $235 million funding round at a $4.5 billion valuation. The investors include Google, Amazon, Nvidia, Intel, AMD, Qualcomm, IBM, Salesforce, and Sound Ventures.
You see an enormous number of large enterprises and hardware providers in there. That’s part of the point—Hugging Face is expanding into enterprises to grow into becoming a one-stop shop for open source model usage. There’s the whole process of hosting those models (which it has over 1 million repositories and 500,000 models), but it also has experiences around that with its inference endpoints and spaces.
“We believe more in really integrating infrastructure with the collaboration layer, with the features, and for that we don’t really compete on raw compute or cost,” Hugging Face Clement Delangue told me in an interview. “But more we focus on providing value by integrating the features and the compute. The Spaces GPU is a good example, because you aren’t buying the infrastructure, you’re buying the spaces and feature. You can build an app faster. “
Also important given those investors: there are no spend commitments from Hugging Face as part of this round, the company said.
Hugging Face became a darling in the early machine learning ecosystem, looking to be the GitHub of AI. It was valued at $2 billion last year, a number that seemed eye-popping to many at the time. But with the Cambrian explosion in development around language models following the launch of ChatGPT, most investors I speak with are kicking themselves for not taking it more seriously (much like they are for Dbt Labs).
Now as it continues to grow, it has to juggle the challenge of cultivating one of the largest open source communities in machine learning—which is very much a thorny proposition—while building an enterprise business in tandem. It’s clearly on Delangue’s mind, as we spoke in depth around how it manages its community and works closely with its own kind of “influencers.”
“An interesting part of our culture is we make sure all team members are interacting with the community members themselves all the time,” Delangue told me. “Some companies have a small set of developer relations people, we try to do it differently with having all people, including me interacting with the community members regularly to have them, priority number one, to build the right product.”
I spoke with Delangue about a number of topics, ranging from managing its growing costs for storing those huge models to whether they’d eventually consider providing a tipping system for these creators.
Here’s my full interview with Delangue:
Supervised is an independent journalism publication. Your support means I can continue to grow and deliver stories and interviews!