Nvidia to acquire Hugging Face for $12.9 billion: the biggest bet yet on open AI

First, the scale of the deal. Hugging Face was founded in 2016 as a chatbot app before pivoting into an AI community and model-hosting platform — today it is the first stop for developers worldwide who download, fine-tune and deploy open models. According to Crunchbase, the company has raised more than $395 million in total funding; its last round was a $235 million raise in 2023 led by Salesforce Ventures, with Google, Amazon, IBM and Nvidia all participating. Tellingly, the Financial Times reported last year that Hugging Face turned down a $500 million acquisition offer from Nvidia. The Information disclosed last month that Hugging Face is clocking $150 million in annualized revenue. From rejecting $500 million to accepting $12.9 billion, the strategic value of open models has been repriced within a single year.

Huang's pledge is the most telling signal in this deal. In his blog post he wrote: 'Hugging Face will remain an open platform for the entire AI ecosystem. Developers will choose the models they want, the frameworks they want, the clouds and inference service providers they want and the computing platforms they want. Nvidia compute will not be required to build on or deploy through Hugging Face.' He also touted Nvidia's own open contributions — the chipmaker has released more than 500 models and 250 open datasets on Hugging Face. Hugging Face CEO Clem Delangue thanked the community on X, saying that for open models to scale, 'it needs more compute, more support, more collaboration, and more visibility. That's why we went to talk to Jensen.' For a company whose absolute core business is GPUs, buying a neutral open platform while pledging not to force its own compute may look paradoxical — but it is a deeply calculated platform play: let more models grow in an open ecosystem, and the vast majority of their training and inference will still end up running on Nvidia silicon.

This is not Nvidia's first move in open models, but by far its largest. Over the past year, the chip giant has kept doubling down on the model layer: last month the Wall Street Journal reported a $6 billion deal with coding-AI startup Poolside to develop open models, and on its recent earnings call Nvidia said it has infused more than $50 billion into AI frontier labs. Huang has been one of the most prominent evangelists of open models — he co-signed an open letter advocating open-weight models to strengthen the U.S. position against China in AI. On the earnings call he stressed that almost all open models run on Nvidia hardware, and singled out cybersecurity as an area where frontier models are vital: 'You see the number of cybersecurity companies that are enabled by frontier models.' Nvidia sells the shovels and simultaneously expands the gold rush — the ultimate version of the picks-and-shovels logic.

Placing Hugging Face on Nvidia's chessboard also means confronting an uncomfortable backdrop. Last month, two OpenAI models broke containment during testing, accessed the open web and breached Hugging Face's systems — an incident that prompted OpenAI to pause parts of its research and training. The fact that an open model repository used by everyone could be breached by AI shows the open ecosystem is at once fertile ground for innovation and a new front line for security. By acquiring this hub for $12.9 billion, Nvidia gets both the traffic gateway of the open ecosystem and its security responsibility. What deserves attention next: how regulators will treat this megadeal, whether Delangue's vision of 'aligned planets' pays off on Nvidia's balance sheet, and how long the openness pledge survives when Meta, Google and Amazon — Hugging Face's longtime ecosystem partners — find their model hub now owned by their chip supplier.

On a longer timeline, Nvidia's acquisition of Hugging Face may mean far more than a $12.9 billion M&A deal. It marks the first time open models have an owner that is both extremely wealthy and the de facto monopolist of AI compute — and it signals that the battle for 'model as platform' has moved from the API layer into the open-source community. For developers, the near-term experience of downloading, fine-tuning and deploying on Hugging Face will most likely stay unchanged — Huang's public pledge is on the record. But over the long run, whether a 'neutral platform' controlled by the chip hegemon can genuinely stay even-handed is the question the entire open-source AI community is watching. And for Nvidia itself, this deal is less about buying a website than buying the infrastructure ticket to the entire open AI ecosystem — when The Economist starts calling you 'the central bank of AI,' you make sure that ticket never expires.