Nvidia lobbed a $500mn (€420mn) investment offer at Hugging Face late last year, but the AI outfit turned it down.
The deal would have valued the group at $7bn (€5.9bn) and was more than it has raised since founding a decade ago.
Hugging Face said it does not want a single dominant investor that could sway decisions, and Nvidia declined to comment.
The snub lands while tech investors chuck billions at the latest AI darlings on frothy valuations, yet Hugging Face is playing it like it has options.
It helps that the outfit has become an influential arbiter for the global AI industry, not just another pitch deck with a GPU bill.
The company runs a platform hosting 2.5 million public AI models and more than 700,000 public datasets for users to download.
A model’s popularity on Hugging Face has become a key signal of its adoption by AI developers and deployment in real applications.
With 13 million users worldwide, Hugging Face has also pushed “open” models that developers can use freely, making it easier for groups like China’s DeepSeek and Alibaba to reach global audiences.
That approach cuts across the grain of dominant Silicon Valley players such as OpenAI, Google and Anthropic, which focus on “closed” models and sell access.
Hugging Face now sits in the middle of a geopolitical and commercial scrap over how artificial intelligence should be built, shared and owned.
Hugging Face, co-founder and chief executive Clem Delangue said “Open models “contribute to democratising AI, to fighting concentration of power, which is, in my opinion, the biggest risk in AI,””.
Hugging Face could afford to say no because its finances are relatively steady, running a “freemium” model where three per cent of customers pay for extras.
Delangue said Hugging Face was profitable in 2025, then slipped into a first-quarter loss this year after investing in datasets.
Hugging Face has raised $400mn (€335mn) in total and was valued at $4.5bn (€3.8bn) in 2023, with about half that capital still on the balance sheet.
Delangue said the company is not trying to maximise revenue growth and is more interested in nudging developers towards open alternatives for text, image and vision models.
Hugging Face, co-founder and chief scientific officer Thomas Wolf said “Open models don’t have to be second-class citizens, but can actually be really, really impressive.”
The White House has responded to China’s advances by pushing US AI giants to invest more in open-source models.
Betaworks, chief executive John Borthwick said the company is “contrarian” to Silicon Valley rivals racing to build frontier models that need billions and massive compute.
“Everybody’s running manically towards… this extreme version of AI superintelligence, that is almost a religious fervour. The Hugging Face guys have a much, much more pragmatic view of this technology,” Borthwick said.
