After three weeks of speculation, Nvidia and Hugging Face ended the ambiguity on September 2 by signing a definitive agreement, with the announcement arriving September 3. The deal values the open-source AI hub at $12.93 billion, a price more than double Hugging Face's $4.5 billion valuation in its 2023 Series D round and roughly triple what Nvidia quietly offered in a $500 million equity stake that the company rejected in late 2025.
The shift from rumor to signed contract matters because it confirms the direction that started with Business Insider reports on August 23. Hugging Face CEO Clément Delangue now says he approached Nvidia's Jensen Huang over the summer, arguing that open-source AI "needed more resources, more infrastructure" to reach its next stage, a framing Nvidia echoed in its announcement blog.
How the deal is actually structured
The numbers lay out in two parts, and the split tells you where the real risk sits. According to an 8-K filed with the U.S. Securities and Exchange Commission, the transaction includes an approximately $11.9 billion purchase price payable to Hugging Face stockholders, subject to adjustments, plus an equity-based retention program worth up to $1.0 billion aimed at Hugging Face employees who join Nvidia.
The deal is expected to close in the first half of 2027, subject to customary closing conditions and regulatory approvals. That timeline leaves significant room for the deal to still fail, especially given the regulatory risk disclosures Nvidia added to its filing. Delangue discussed the closing timeline and terms in an interview with CNBC's Squawk Box.
Nvidia committed in the agreement to keep the platform open, consistent with existing practice. The language is specific. Developers keep choosing their own models, frameworks, clouds, inference providers, and computing platforms. Nvidia compute will not be required to build on or deploy through Hugging Face. The platform will continue supporting open weight models and multi-accelerator development.
The scale that justifies the price tag
The announcement cites a platform now hosting more than 18 million developers, 3 million models, 500,000 datasets, and 1 million applications, with more than 200,000 companies using it to discover, evaluate, and deploy AI. Nvidia describes itself as the single largest contributor of open models and data to the platform, with more than 500 models and 250 open datasets released there.
That last point is the strategic core of the acquisition. Nvidia is not buying code or revenue so much as the attention graph where AI models are discovered and standardized.
Our Read: the retention program and the open-weight bet
Look at the deal mechanics through an engineering and economics lens, and two things stand out.
First, the $1.0 billion equity retention program is a deliberate hedge against the one category of risk that dominates a capital-light software business: key person and brain-drain risk. Hugging Face's product is built by a relatively small engineering team, but the platform's value lives in the 18 million developers who treat it as default infrastructure. When you acquire a company whose main asset is a community plus a small group of engineers who maintain it, the retention structure is the deal. Bidding $11.9 billion in cash and equity for the company itself is almost secondary to locking in the people who keep the hub running.
Second, the revenue multiple looks insane until you price the right asset. Hugging Face was reporting around $150 million in annualized revenue as of late August, which puts this at roughly 86 times revenue. No traditional valuation supports that. But the buyer is paying for distribution, not profit. The same flywheel that makes Nvidia's CUDA software stack sticky works here: the more models ship through Hugging Face, the more the ecosystem standardizes around tools and hardware that Nvidia already profits from. Nvidia is buying the discovery layer for open-source AI, and the open-weight wave is exactly what is eroding the moat around closed models from OpenAI and Google.
This is where the regulatory risk in the 8-K becomes the central tension. The filing warns that governments may impose requirements on the development, release, or use of open-source models, and it explicitly flags that many of the most popular open models originated in China. For a data scientist who has fine-tuned a Qwen or DeepSeek checkpoint downloaded from the hub, this is the reason the platform exists at all. Any restriction on supporting region-derived models could constrain what lives on Hugging Face, which is partly what makes Hugging Face valuable. Nvidia is acquiring a company whose worth is entangled with open models that geopolitics wants to fragment.
The antitrust shadow
The commitment to keep Nvidia compute optional is a direct response to that concern. When a chip monopoly owns a gatekeeper to the model ecosystem, antitrust reviewers expect foreclosure promises. The multi-accelerator and multi-cloud language is Nvidia buying regulatory goodwill while also reassuring the community it will not be forced to run only on Hopper or Blackwell GPUs.
The Arm precedent looms. Nvidia's proposed $40 billion Arm acquisition collapsed in 2022 under global regulatory pressure. Reviewers may treat the Hugging Face deal through the same skeptical lens, especially since it combines hardware dominance with control over an open-source distribution monopoly.
What comes next
The first-half-2027 closing window gives both companies and regulators time. If it closes, Nvidia gains the primary routing point for open-source model discovery at exactly the moment open-weight models are closing the gap with closed systems. If regulators force carve-outs or block it entirely, Nvidia keeps its position as the platform's largest contributor without owning it.
For developers, the operating promise is stable: upload and download what you choose, run it where you want, and pick your own accelerators. Whether that promise holds under Nvidia ownership depends almost entirely on what regulators demand before the 2027 close.