So Nvidia is buying Hugging Face. The kings of closed source in control of the GitHub of AI. Holy moly. The fee is $12.9 billion, and for the role Hugging Face plays, that’s truly a steal.

On paper, such a deal arguably carries more weight than Nvidia’s attempted Arm takeover. There, it tried to buy the processor architecture the world increasingly relies on and failed. Now, it’s reaching for the software and community layer everyone depends on.

Nvidia already dominates the hardware used to train and run models hosted there. It’s like the captain of the school football team is also a chess champion, popular with the ladies, and can actually fix computers.

Now where this potential deal differs from Arm is that what the British chip designer offered was intellectual property. Something tangible. With Hugging Face, Nvidia is just buying a platform and something more altogether. It’s not buying the model weights themselves. More a distribution hub, if you will, something that could quite easily be mirrored elsewhere, thereby limiting that lock-in fear we all initially got when we first glimpsed the headlines.

The argument could be made then that such a deal comes as hyperscalers and AI labs, once so reliant on its chips, are now going it alone and developing some truly powerful silicon. So you could read this as Nvidia almost buying insurance against losing hardware dominance and not simply extending its market position that little bit further.

During the company’s latest mammoth earnings, CEO Jensen Huang faced questions on open source AI directly, telling investors: “The world will need both closed models and open models." Fitting for a company that frames itself as the most open and yet the most closed in the world.

“I would say nearly all open models run on Nvidia and the reason for that is because Nvidia’s footprint around the world is the highest and our architecture is the most fungible. It's everywhere," Huang said. "Closed models, we know, are doing incredibly well. The frontier labs, their scale – their sales are skyrocketing, their margins are fantastic. They're generating profitable tokens. They are only limited by the amount of compute. That is equally true for open models. And our position in open models is very good because the CUDA ecosystem is literally everywhere.”

During that call, Huang pretty much summed up the reasoning behind the now confirmed deal: “Our market footprint of all AI models, I think we're the only platform. I'm fairly certain we're the only platform that runs every frontier model, whether it's closed or open. … So long as models succeed, I'm very happy. And both closed and open models are going to succeed, and they're both simultaneously driving our sales.”

We’ve played these games before

So, will Nvidia do with Hugging Face what it couldn’t do with Arm?

Yes. And no.

The argument for rests on the current state of the U.S. antitrust system. Consider Hewlett Packard Enterprise’s (HPE's) Juniper Networks takeover and the alleged behind-the-scenes shenanigans from President Donald Trump administration officials. Attempts to fight that settlement in court failed when a judge ruled that despite acknowledging they’d raised legitimate questions about the Department of Justice's (DOJ) process, challengers hadn’t shown the settlement wasn’t in the public interest.

That was for a takeover concerning two firms that control a little over two-thirds of the wireless local area network market. But when it comes to who controls the AI stack? Nvidia has enough money and highly paid lawyers to swing such a vaguer debate in its favor.

The case against? Hugging Face is the only true bright spark of Europe’s AI malaise. You’ve got to be cruel to be kind and admit the continent lost the AI race years ago. Next to no real AI accelerator players and stuck between a Trump and a Chinese place, Hugging Face was Europe’s last bit of change, and Nvidia is hungry.

For semantics' sake, Hugging Face is legally incorporated as an American company, but holds deep roots in Europe, owing to its French founding. But if you think Europe will willingly let the most prominent AI company built by Europeans go off into the sunset under the ownership of Americans at a time when the term sovereignty is constantly being bandied about, think again.

To be fair to Nvidia, it has made a fair few strides when it comes to open source AI. It routinely releases ever more powerful versions of its Nemotron line of AI models through permissive licenses, and Huang himself hyped up "agentic scaling" when lauding OpenClaw at GTC this year.

Training recipes, pretraining corpora, and domain-specific data are also shared regularly, with Hugging Face’s own CEO crowning Nvidia “the largest organization” on the platform.

“I'm officially calling it: Nvidia is the new king of American open-source AI,” CEO Clem Delangue said in a LinkedIn post just five months ago. How the turntables turn.

Hugging Face is believed to have turned down Nvidia’s advances before. The FT reported in January that Delangue and co. flat-out rejected a $500 million investment offer from Nvidia, valuing itself at the $7 billion mark. Just shy of $13 billion is a heck of a lot larger than that initial offer, and there is a possibility that this doesn’t come to anything.

But it’s fair to say, though, that yes, Nvidia will get its way because the system will let it. HPE-Juniper’s precedent is the tell. And vague harms lose to expensive law firms every single time.

The answer is no when you consider the asset won’t sit still. Nvidia would be buying a community, something you can’t write yourself. Sure, they’ve turned what was an annual gathering of hardware nerds at GTC into a literal festival of sorts, so they have the means, plus the cash. But that lock-in fear that this is Arm 2.0 may be overstated by the simple fact that open-source infrastructure is, by definition, harder to enclose.

Unless you’re European, regulators won’t be the problem here. So Nvidia will likely come out on top in this fight. But is it a fight it doesn’t need to win? The prize may shrink the moment it’s claimed.

UPDATE: This op-ed has been revised after Nvidia announced its intent to acquire Hugging Face on September 9.