- Lead. Nvidia on September 3 confirmed a $12.93 billion agreement to acquire Hugging Face, the open-source AI platform used by more than 18 million developers worldwide, giving the chipmaker control of the dominant repository for machine-learning models and datasets.
- Fact. The deal values Hugging Face at roughly 86 times its $150 million in annualized revenue; the company previously turned down a $500 million offer from Nvidia before ultimately accepting a figure 26 times larger.
- Stake. The acquisition creates a form of vertical integration previously absent in AI: a company that designs the chips on which models are trained also controlling the primary marketplace where 3 million models, 500,000 datasets and 1 million applications are distributed.
The deal, confirmed publicly on September 3, consists of approximately $11.9 billion payable to Hugging Face stockholders and a $1 billion equity-based retention program for Hugging Face employees who will join Nvidia. The transaction is expected to close in the first half of 2027, subject to regulatory clearance. Hugging Face will operate as a unit within Nvidia rather than being absorbed entirely, preserving the brand and its open-source governance structure.
CEO Jensen Huang stated that “Hugging Face will remain an open platform for the entire AI ecosystem,” a commitment interpreted as a response to pre-announcement concern within the developer community about Nvidia using the acquisition to favour its own hardware. The company had earlier suggested it would continue to support multi-cloud environments and would not make Nvidia accelerators a requirement for running models hosted on the platform.
How the deal came together
According to CNBC’s reporting, Hugging Face’s CEO approached Huang directly in the weeks before the announcement, a departure from the conventional pattern in which an acquirer pursues a target. Hugging Face had previously rejected a $500 million offer from Nvidia—a figure that at the time valued the company at roughly 3.3 times revenue. The final price of $12.93 billion represents a valuation expansion that reflects the platform’s growing centrality to the AI development stack: its hosted repository has become the de facto standard for distributing and discovering open-weight models, and its inference infrastructure is used by companies ranging from startups to sovereign AI programs.
Hugging Face had raised over $395 million across prior funding rounds, with backers including Salesforce, Google, Amazon and NVIDIA itself, which had participated as an investor before moving to acquire the company outright. Nvidia’s acquisition interest had been reported before the definitive agreement was signed, but the final price and the retention package terms were not known until the September 3 disclosure.
The vertical integration question
The deal raises a structural question that US and EU antitrust regulators will examine ahead of the H1 2027 close. Nvidia’s chips—particularly its H100 and B200 series—already account for the majority of AI training and inference workloads. If Nvidia also controls the principal platform through which models are discovered, fine-tuned and deployed, a new kind of data-flywheel advantage emerges: usage patterns across 18 million developers would give Nvidia earlier and deeper insight into where AI compute demand is heading than any competitor. The practical question for regulators is whether that informational advantage, combined with hardware dominance, constitutes a foreclosure risk for rival chip designers such as AMD, Intel’s Gaudi division and emerging domestic chip programs in China and the European Union.
Nvidia has moved to preempt that scrutiny by promising Hugging Face will maintain support for all major hardware backends. Whether that commitment survives regulatory review—or whether reviewers seek structural remedies such as a firewall between Hugging Face’s model usage data and Nvidia’s hardware business—will shape the terms under which the deal ultimately closes.