Nvidia agrees to buy Hugging Face for $12.93 billion

Nvidia, the AI chip designer, agreed Thursday to buy Hugging Face, the platform where developers publish, find and run open models, for $12.93 billion. The purchase would put the company that already sells the chips under most large AI systems in charge of the public square sitting on top of those chips.

The company entered a definitive agreement on Wednesday, September 2. About $11.9 billion is payable to Hugging Face stockholders, subject to adjustments, plus an equity-based retention program of up to about $1.0 billion for employees who join Nvidia. The filing does not say whether stockholders will be paid in cash, Nvidia stock, or a mix. Required regulatory approvals are a closing condition. Nvidia did not name the agencies.

Nvidia shares closed Thursday at $228.45, up 1.8%, valuing the company at $5.53 trillion. That session is when Nvidia posted the agreement; the filing does not attribute the move to the deal, and Microsoft, the software company, rose 2.7%. Hugging Face will not appear in a fund. Nvidia already accounts for a large share of ASPY, AQQQ, AVGT and ASMH:

Hugging Face's 18 million builders

Huang's post is the source for Hugging Face's scale. More than 18 million developers, researchers and creators use the platform to share more than 3 million models, 500,000 datasets and 1 million applications. More than 200,000 companies use it to discover, evaluate, customize and deploy AI.

Hugging Face is a private company. It sells subscriptions and metered compute around that library: a Pro plan at $9 a month, Enterprise at $50 per user per month, and on-demand GPUs starting at $0.40 an hour on its price list. Nvidia did not disclose its revenue. The Information in late August estimated annualized revenue at $150 million, up 50% in two months. That is an estimate, not a company figure.

Nvidia was already a backer. Hugging Face raised $235 million in a 2023 Series D that Salesforce Ventures led, with Google, Amazon, IBM, Nvidia, AMD, Intel, Qualcomm and Sound Ventures participating, at a $4.5 billion valuation. Some of the proceeds will go to those prior backers; stake sizes and individual amounts were not disclosed.

Huang wrote that Nvidia is Hugging Face's largest contributor of open models and data, with more than 500 models and 250 open datasets on the platform. Hugging Face chief executive Clément Delangue approached him weeks before the agreement, according to CNBC; Huang wrote that Delangue came to him as he considered the company's next chapter.

The openness pledge

The strategic claim is not a near-term earnings number. Huang wrote that Nvidia will "scale Hugging Face's platform, strengthen its infrastructure and expand access to AI for developers and institutions worldwide." He also wrote the sentence regulators will read first: "Hugging Face will remain an open platform for the entire AI ecosystem." Developers, he said, will choose their models, frameworks, clouds, inference providers and computing platforms. "NVIDIA compute will not be required to build on or deploy through Hugging Face."

The filing matches that pledge. Nvidia committed to keep the platform open, consistent with Hugging Face's existing practices, including support for other silicon vendors and the right for model makers to upload and download what they choose.

That is a promise about requirement, not about ranking. The filing does not say how Hugging Face will order models on the hub, which hardware will be the default for inference, or what Nvidia may do with the usage data the platform already collects. Huang's own count, more than 500 of Nvidia's models among more than 3 million on the hub, is a reminder that most of what the platform distributes was not built by Nvidia.

Forrester principal analyst Naveen Chhabra offered a different mechanism. Nvidia, he said, "gains visibility into customer's preferences and the AI models they use" and can see which models, datasets and architectures gain traction before they show up in mainstream coverage. In his reading, Nvidia is "securing the software layer because hardware-only companies get commoditised over time." That is an analyst's interpretation, not Nvidia guidance. No customer or competing chipmaker went on the record Thursday to endorse or oppose the transaction.

A first-half 2027 close

Nvidia says the transaction is expected to close in the first half of 2027, subject to customary conditions including required regulatory approvals. The filing names no reviewing agency. Nvidia's purchase of Run:ai, which orchestrates GPU clusters, was announced in April 2024 and closed on December 30, 2024, after the European Commission cleared it. Until a notice is filed and an agency puts a clock on Hugging Face, the dated fact is Nvidia's own window, not a clearance.

The $12.93 billion package is small against Nvidia's $5.53 trillion value. For a holder of ASPY, AQQQ, AVGT or ASMH, Thursday did not add Hugging Face to the portfolio. It added a claim on a private platform that, if the deal closes, becomes part of Nvidia sometime next year. The written test is dated and narrow. After a first-half 2027 close, the filing's commitment is that a developer can still build and deploy on Hugging Face without Nvidia compute, and that the platform still supports other silicon. Ranking, default hardware, and whether the preference data Chhabra described becomes an advantage are not in that sentence. Those distinctions become checkable only once the platform has a single owner.