Nvidia has agreed to buy Hugging Face for $12.9 billion, according to The Information, giving the world’s dominant AI chipmaker control over the platform where most open-source AI models are hosted. The deal, reported Wednesday night US time, has not yet produced a signed agreement and could still fall through, Business Insider noted.
The price is a staggering jump from Hugging Face’s $4.5 billion valuation in 2023. It also represents a reversal: Hugging Face turned down a $500 million Nvidia investment late last year that would have valued it at $7 billion, reportedly because it did not want a dominant investor swaying its decisions. At $12.9 billion, that hesitation appears to have been temporary.
Why Nvidia wants it
The strategic logic is straightforward. Hugging Face, founded in 2016, is the central hub where developers share and download open-source AI models — think of it as GitHub for machine learning. Buying it gives Nvidia three things at once.
First, a foothold in open-source AI at the exact moment open models from China and elsewhere are closing the gap with closed systems from OpenAI and Anthropic. Hugging Face CEO Clem Delangue has spent this year publicly aligned with Nvidia’s open-source push, even appearing on CBS’s Face the Nation to argue against US restrictions on open-weight models. A letter signed by Jensen Huang and 24 other companies urged the US government to support open models rather than restrict them.
Second, a route back into cloud computing. Nvidia scaled back its DGX Cloud business about a year ago. Hugging Face already helps developers run models on rented compute — owning it gives Nvidia a cloud platform without building one from scratch.
Third, a hedge against unused capacity. Nvidia has promised to help cover the cost of tens of billions in cloud computing deals for its customers. If those customers do not use all the compute they signed up for, Nvidia could get stuck with the bill. Hugging Face’s user base gives it a place to sell that spare capacity.
The open-source question
What stands out here is the concentration risk. The single largest open-source AI repository — used by millions of developers, hosting models from Meta, Google, DeepSeek, and hundreds of others — would be owned by the company that also sells the chips those models run on. Nvidia is already an investor in Hugging Face, Anthropic, OpenAI, xAI, and others. Adding ownership of the distribution layer is a different category.
Delangue warned in a July CNBC interview that China is “clearly dominating” open-source AI. The implicit argument is that Nvidia ownership strengthens the Western open-source ecosystem. Maybe. It also means a single American company controls the platform that the global open-source community relies on — which is precisely the kind of concentration that makes policymakers nervous.
Neither Nvidia nor Hugging Face has publicly commented. TechCrunch reported that Nvidia’s silence is notable, given the company has historically moved quickly to address reports it considers inaccurate.
How we got here
Hugging Face raised $235 million in 2023 at a $4.5 billion valuation, led by Salesforce Ventures with participation from Alphabet’s GV, IBM Ventures, and Nvidia itself. The company has since grown its revenue through its hosted inference business and enterprise subscriptions, reaching what Delangue told TechCrunch was “close to profitability.” A $13 billion exit for a company that was worth $4.5 billion three years ago is a steep multiple, but Nvidia can afford it — the company’s market cap has ballooned past $4 trillion on the AI spending boom.
The deal also arrives against the backdrop of our earlier reporting on Nvidia’s $250 billion OpenAI data centre guarantee and the broader consolidation of AI infrastructure. Stripe’s recent acquisition of OpenRouter, a model-routing startup, points in the same direction: the infrastructure layer is consolidating, and the companies with the deepest pockets are buying their way in rather than building.
What it means for New Zealand
For NZ developers and researchers using Hugging Face — and many do, particularly in university ML courses and startups — the practical question is whether access and pricing stay the same. Nvidia has no track record of restricting open-source model access, but owning the platform changes the incentives. NZ’s small AI ecosystem relies heavily on open-source models precisely because it cannot afford the compute bills that closed frontier models generate. Anything that narrows that access would matter more here than in larger markets.
This connects to the open-weights convergence we covered in December, where the gap between open and closed models was narrowing fast enough to change the calculus for small countries. Nvidia buying Hugging Face does not close that gap — but it puts a gatekeeper in front of it.
❓ FAQ
Will Hugging Face stay free for developers? Nvidia has not commented on the deal, so there is no official statement on pricing. Hugging Face’s free tier is central to its developer community, and Nvidia would have little reason to alienate that user base. But the enterprise and inference tiers — where money actually changes hands — are where changes are more likely.
Does this affect open-source model availability? In the short term, no. Hugging Face hosts models from hundreds of organisations, and restricting access would trigger an immediate exodus to alternatives. The longer-term question is whether Nvidia uses its ownership to steer the platform toward its own hardware and cloud services.
What happens to Hugging Face’s existing investors? Salesforce Ventures, GV, IBM Ventures, and Nvidia itself all hold stakes from the 2023 round. A $12.9 billion acquisition would represent roughly a 3x return on that last valuation — a strong outcome for early backers, though the deal has not yet closed.
🔍 THE BOTTOM LINE
A $12.9 billion acquisition that puts the largest open-source AI repository under the control of the largest AI chipmaker is not just a business story. It is a structural shift in who controls the infrastructure layer of the AI economy. Nvidia is buying distribution, cloud revenue, and a hedge against its own customers building chips — all in one deal. The open-source community gets a wealthy patron with a clear commercial interest in keeping models accessible. Whether that patronage stays benign is the question nobody can answer yet.
