Nvidia reportedly agrees to buy Hugging Face for $12.9B, but deal is unsigned
Hugging Face co-founders Clement Delangue, Julien Chaumond and Thomas Wolf previously rejected a $500M Nvidia investment over concerns about investor influence. The reported acquisition would value Hugging Face at $12.9B, against roughly $150M in annualized revenue.
By RuntimeWire Staff · Published
Primary source: Forbes
Why it matters
The reported $12.9B price is roughly 86 times Hugging Face's reported annualized revenue. Nvidia ownership would also put Hugging Face's cross-hardware libraries and model platform under the leading AI accelerator supplier, making continued support for AMD, AWS, Google, Intel and Furiosa hardware a material concern for customers.
Hugging Face co-founders Clement Delangue, Julien Chaumond and Thomas Wolf rejected a proposed $500M Nvidia investment in late 2025 after Hugging Face said it did not want a dominant investor able to sway its decisions. Less than a year later, The Information reported that Nvidia had agreed to acquire Hugging Face for $12.9B.
The reversal has yet to produce an undisputed deal. The Information published its report on August 26, while Business Insider reported that negotiations had not produced a signed contract and could still collapse. Reuters documented the conflicting accounts. Neither Nvidia nor Hugging Face had publicly confirmed a definitive agreement as of August 31.
The founders drew a line at investor control
Delangue, Chaumond and Wolf founded Hugging Face in 2016 as an AI chatbot venture. They pivoted into machine-learning infrastructure after deciding that the natural-language-processing models beneath the chatbot held greater value, according to an interview with Chaumond. Hugging Face is based in New York and has a substantial operation in Paris.
Delangue, Hugging Face's CEO, previously co-founded VideoNot.es and UniShared and worked at eBay, Mention and computer-vision startup Moodstocks, according to his published biography. Chaumond serves as CTO. Wolf, a physicist and former patent attorney, became chief science officer and describes his work as an effort to make machine-learning research broadly accessible on his personal site.
Nvidia was already an investor. Hugging Face raised $235M in August 2023 at a reported $4.5B valuation, with Nvidia joining a group that included Salesforce Ventures, Google, Amazon, Intel, AMD, Qualcomm, IBM and Sound Ventures, according to TechCrunch's account of the financing. Salesforce Ventures led the round.
The commercial relationship extended beyond equity. In August 2023, Nvidia and Hugging Face announced a DGX Cloud collaboration intended to give Hugging Face users access to Nvidia computing infrastructure for training and tuning models.
The rejected $500M proposal would have valued Hugging Face at $7B, TechCrunch reported, citing the Financial Times. The reported acquisition price is about 2.9 times Hugging Face's 2023 financing valuation and nearly twice the valuation attached to the rejected proposal. Those comparisons involve different types of transactions, and Nvidia and Hugging Face have not confirmed final acquisition terms.
The price runs well ahead of reported revenue
The Information put Hugging Face's annualized revenue at approximately $150M, up from roughly $100M two months earlier. Nvidia and Hugging Face have not independently confirmed those figures.
At $150M in annualized revenue, a $12.9B purchase price equals roughly 86 times revenue. The calculation uses a reported annualized run rate rather than audited trailing sales, but it shows how much expected value lies beyond Hugging Face's present income.
Hugging Face earns money from subscriptions and computing services attached to its software platform.
The Hugging Face Hub hosts models, datasets and machine-learning applications called Spaces. Hugging Face also maintains software libraries including Transformers and Optimum. Hugging Face's open-source FAQ says source-code licenses allow outside contributors to fork many libraries.
Nvidia ownership would reach across competing hardware
Hugging Face helps developers move models across chips and cloud services that compete with Nvidia. Optimum documentation includes hardware-specific installation paths for Nvidia systems, AWS Trainium and Inferentia, Google TPUs, Intel Gaudi, Intel OpenVINO and Furiosa accelerators. Hugging Face separately documents support for AMD GPUs.
A sale would place decisions about staffing, testing and release priorities for those integrations under Nvidia, the leading supplier of AI accelerators. Open-source licenses would preserve the ability to fork covered libraries, but maintaining hardware integrations still requires engineers and testing infrastructure. The current documentation establishes cross-vendor support; no confirmed acquisition agreement establishes how Nvidia would govern it.
For customers, that makes maintenance of non-Nvidia backends a concrete vendor-management issue if the transaction closes. For the founders, the reported agreement would resolve the control question in the opposite direction from their 2025 decision: Nvidia would move from a rejected dominant investor to Hugging Face's owner.
That outcome remains unconfirmed. Delangue, Chaumond and Wolf have not publicly said that they accepted Nvidia ownership, and the reporting still conflicts over whether a contract has been signed. The $12.9B transaction remains a reported proposal rather than a completed acquisition.