Hugging Face explores a sale at a $13bn valuation, nearly triple its last one

Hugging Face has retained bankers to test the market for a sale at a valuation of $13bn or higher, according to Business Insider, which reported that talks are early and that no bidder has been identified.

It is a striking development for a company that has spent the past decade positioning itself as the neutral ground of machine learning, and that only recently went looking for help with compute costs by asking OpenAI for $100mn of it.

The number represents close to a tripling of the $4.5bn valuation Hugging Face carried after its last outside round, a $235mn raise in August 2023 led by Salesforce Ventures with Nvidia, Google, Amazon, Intel, Qualcomm, IBM, Sequoia Capital, and Lux Capital alongside. That the company has not raised externally since is unusual in a sector where three years is several funding cycles.

Founded in New York in 2016 by Clement Delangue, Julien Chaumond, and Thomas Wolf, Hugging Face started as a chatbot company and pivoted into the infrastructure layer beneath everyone else’s models. Its Hub now hosts more than three million public models and around a million datasets, which is what makes it strategically interesting to a buyer and awkward to price.

Revenue comes from paid subscription tiers, enterprise hosting, and compute services sold on top of the free repository, and the company has never published figures. As of November 2025, roughly half of the $400mn or so it had raised over its life remained unspent, which is a reasonable position from which to negotiate and a poor one from which to be forced into anything.

The question a buyer would have to answer is what exactly it is buying. Hugging Face’s value rests on being the place developers trust to publish and download open weights, and trust of that kind tends not to survive being owned by a company with its own models to sell.

That tension is not hypothetical. Mistral’s chief executive has argued that closed models hand providers leverage over their customers, and the same logic applies one layer down to the registry where the open alternatives live.

The Hub’s scale has also become a liability of its own. Researchers found hundreds of malicious models and agent skills planted on Hugging Face and ClawHub in a supply chain campaign aimed at AI infrastructure, and separate work has traced nudify tools the EU is moving to ban back to components hosted on the platform.

Moderating a repository of three million artefacts is a cost that grows with the asset, and any acquirer takes on the regulatory exposure attached to it. In Europe that exposure is not abstract, since the AI Act places obligations on providers of general-purpose models and the question of where a hosting platform sits in that chain has not been settled cleanly.

A sale at $13bn would also land in a market that has grown noticeably keener on AI infrastructure than on AI applications, where multiples are set by the assumption that whoever owns the pipes collects rent regardless of which model wins. Hugging Face owns something close to the distribution layer for open weights, and there is no obvious second version of it.

Hugging Face has been expanding rather than retrenching. It acquired the French robotics developer Pollen Robotics in April 2025, moving into hardware alongside the software business, and it has continued to fund open-source releases that generate no direct revenue.

Neither Hugging Face nor its advisers have commented on the report, and Business Insider’s account describes a process at the sounding-out stage rather than a deal in progress. Exploratory mandates of this kind frequently end in a large funding round instead, which for a company sitting on unspent capital and a valuation mark three years stale would be the less disruptive outcome.

The list of plausible buyers is short and mostly consists of Hugging Face’s own investors. Nvidia, Google, Amazon, IBM, and Salesforce all hold stakes and all have reasons to want the Hub inside their perimeter, which is precisely the arrangement much of the open-source community got on the platform to avoid.