Hugging Face has turn into one of the vital essential corporations in artificial intelligence with out ever changing into considered one of its most well-known. Ask most AI customers what it does and you could get a clean look, or maybe a dim reminiscence that it was on the middle of the OpenAI runaway-agent controversy only a few weeks in the past. Ask an AI developer, although, and there’ll be way more recognition. They’ll have downloaded a mannequin from it, uploaded considered one of their very own, examined a system there, or used considered one of its instruments.
That makes reviews that Hugging Face could be acquired by Nvidia for $12.9 billion particularly vital. The deal has not been publicly confirmed by both firm, and neither responded instantly to Quick Firm’s request for remark. But when it goes by means of, it might put the AI business’s de facto repository for open fashions, datasets, and instruments beneath the management of the world’s dominant provider of AI chips.
Hugging Face has turn into a sort of GitHub for AI. Hundreds of thousands of builders use it to add and obtain fashions, share datasets, collaborate on initiatives, and deploy AI methods. It performs a central function within the AI ecosystem past the closed companies run by large tech corporations like OpenAI and Anthropic.
“It feels slightly bit like déjà vu when GitHub was rumoured to be in conversations to be acquired,” says Stefano Maffulli, an open supply advocate. (GitHub was finally bought by Microsoft for $7.5 billion in 2018.) “Similar to GitHub captured a lot of the open supply ecosystem, facilitating all of these exchanges, forking, copying, and collaboration options, I see the identical form of strategy with Hugging Face.”
Nvidia isn’t a stranger all of the sudden barging into the open supply ecosystem. It invested in Hugging Face’s 2023 $235 million funding round, when the startup was valued at $4.5 billion, and the 2 corporations have labored carefully collectively for years. In 2023, they announced a partnership connecting Hugging Face customers on to Nvidia’s DGX Cloud computing infrastructure.
That relationship has solely deepened. Hugging Face now provides Training Cluster as a Service, developed with Nvidia, to make massive GPU clusters simpler for researchers and firms to entry. But there has beforehand been a restrict to how shut Hugging Face was prepared to get. Earlier this yr, the corporate reportedly turned down a $500 million Nvidia investment that might have valued it at $7 billion, partially as a result of it didn’t need one dominant investor to achieve an excessive amount of affect. An outright acquisition would clearly go significantly additional.
“The important piece is that place that’s pretty impartial,” says Maffulli.
Duane O’Brien, govt director of the Open Supply Initiative, says Hugging Face’s significance comes not simply from the sheer variety of fashions and weights it hosts, however from the truth that builders can encounter the fashions and weights it hosts on comparatively impartial floor. An acquisition turns into dangerous if that neutrality begins to erode.
“Platform neutrality has worth,” O’Brien says, notably if it ensures that fashions and weights are offered on their deserves fairly than tilted towards an proprietor’s proprietary know-how. The hazard, he provides, is that an acquirer might “inhibit the power of open weights and open fashions to thrive.”
Sustaining that neutrality issues as a result of Hugging Face works throughout AMD GPUs, Intel hardware, AWS accelerators, and Google TPUs, all of which compete with Nvidia’s {hardware}, whereas its hosted companies interface with the likes of Amazon Net Providers, Microsoft Azure, and Google Cloud. In current months, it has additionally collaborated with AMD, an Nvidia competitor, on optimizing AI models for its chips.
“What might get misplaced if throughout the acquisition is that impartial path,” Maffulli says. Turning Hugging Face from what he describes as a sort of Switzerland for AI growth right into a vertically built-in route towards one firm’s infrastructure would, he says, be “a foolish transfer.”
Bradley M. Kühn, coverage fellow on the Software program Freedom Conservancy, provides a much less rosy view. He factors once more to Microsoft’s $7.5 billion buy of GitHub in 2018, and what adopted. “Microsoft shopping for GitHub was the leap of catastrophe that led us to all of this, proper?” he says, pointing to GitHub Copilot and the commercialization of software program hosted on the platform. “I’d count on one other period of darkish ages to be bootstrapped by an acquirer.”
Nvidia has good causes to need Hugging Face. As a few of its greatest clients, together with Google, Amazon, and Microsoft, develop their very own AI chips, Nvidia is increasing into different elements of the AI stack. Proudly owning Hugging Face would give it a direct relationship with tens of millions of builders deciding which fashions to make use of and what {hardware} to run them on.
The large unanswered query is what Nvidia would do with that energy.
Maffulli believes builders would finally reply if Hugging Face grew to become too carefully tied to its proprietor. “Ought to that occur, I’m fairly assured that builders will route across the choke level,” he says.
