NVIDIA to Acquire Hugging Face in $12.93B Open-Source Bet

NVIDIA on September 3, 2026 agreed to acquire Hugging Face, the largest open-source AI model and dataset hub, in a transaction valued at $12,930,300,000 — roughly $12.93 billion. The chipmaker's CEO, Jensen Huang, announced the deal in a post on NVIDIA's official blog, calling Hugging Face "a vibrant home for the open model developer community."

The definitive agreement was signed on September 2, 2026, according to NVIDIA's Form 8-K filing with the U.S. Securities and Exchange Commission the next day. The structure is unusually cash-heavy: 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 Hugging Face employees who join NVIDIA. The retention pool pushes the headline figure to $12.93 billion.

Why now, and why Hugging Face

Hugging Face was founded in 2016 in Paris — the three co-founders, Clément "Clem" Delangue, Julien Chaumond and Thomas Wolf, started at Station F before pivoting from a consumer chatbot to an open machine-learning platform. Its Transformers library became foundational infrastructure for large-model development, and the Hub is often described as "the GitHub of AI." The platform now serves more than 18 million developers, researchers and creators who share over 3 million models, 500,000 datasets and 1 million applications; more than 200,000 companies use it to discover, evaluate, customize and deploy AI.

The strategic logic is about the developer funnel, not just software revenue. NVIDIA already sits at the bottom of the AI stack — GPUs, CUDA, and NVIDIA AI Enterprise. What it lacked was ownership of the layer where developers actually choose which models to build on, which datasets to pull, and which deployment tooling becomes standard. Hugging Face is precisely that layer. Owning it gives NVIDIA a direct, top-of-funnel channel to the people whose workloads will ultimately run on its hardware.

In a telling detail, Delangue — not Huang — initiated the transaction. "I am honored that Clem came to me as he considered the next chapter of Hugging Face," Huang wrote, adding that Delangue "believed NVIDIA would be a great home for the company, its community and the future of open models." In a CNBC interview, Delangue said he approached Huang over the summer of 2026 because open-source AI was "at a turning point" and needed "more resources, more scale, and more visibility."

The openness commitment — and its limits

The single biggest risk to the deal's value is the open-source community's fear of capture. NVIDIA has tried to pre-empt this with a commitment that is unusually specific and, per Chinese coverage of the SEC filing, legally binding. Hugging Face "will remain an open platform for the entire AI ecosystem," Huang wrote. "NVIDIA compute will not be required to build on or deploy through Hugging Face." The company pledged to keep supporting open- and open-weight models from every builder, and to maintain multi-cloud, multi-accelerator development and deployment.

That pledge is mirrored in the 8-K, which states NVIDIA committed to "keep Hugging Face's platform open" and that the platform "would continue to permit model makers, developers, and users to upload and download models and datasets of their choosing and to support other silicon vendors." The wording matters: it names AMD- and Intel-class competitors explicitly, not just in spirit.

NVIDIA also has a credible track record as an outside contributor — it is, by its own account, the largest contributor of open models and data to the Hub, with more than 500 models and 250 open datasets published there.

Price, valuation, and the deal's place in NVIDIA's history

The $12.93 billion price is steep on fundamentals. Hugging Face's annualized revenue was reported at roughly $150 million as of mid-2026 — meaning NVIDIA is paying about 86 times sales. That multiple reflects a strategic purchase of control over the model-distribution layer, not a cash-flow bet.

It is also a dramatic re-rating. Hugging Face raised at a $4.5 billion valuation in 2023, and — per reporting from Business Insider and Chinese outlets — turned down a much smaller NVIDIA investment offer, around $500 million, roughly a year ago. The final price more than doubles even the implied valuation of that rejected approach.

By NVIDIA's own deal history, this is its second-largest acquisition, behind only the roughly $20 billion purchase of Groq's assets in December 2025 (structured as an asset purchase rather than a full takeover). It dwarfs the $6.9 billion Mellanox acquisition completed in 2020, and stands in contrast to the abandoned $40 billion Arm bid that regulators killed in 2022.

Investors appeared to welcome it: Bloomberg reported NVIDIA shares rose about 2% on the announcement, adding roughly $132 billion to a market value then near $5.5 trillion.

Antitrust and the road to close

The transaction is expected to close in the first half of 2027, subject to "customary closing conditions, including receipt of required regulatory approvals." In practice that means a Hart-Scott-Rodino antitrust filing in the United States and a merger review by the European Commission, whose jurisdiction is engaged by Hugging Face's Paris headquarters and EU user base. No regulator had publicly commented as of the announcement.

The 8-K itself flags the risk: "Government restrictions may negatively impact our business and the Hugging Face platform," noting that "other parties are actively lobbying" for measures that would "restrict or disadvantage open-source models." For a company whose entire thesis here is openness, regulatory hostility to open weights is a genuine threat.

ORIGINAL INSIGHT: lock-in by gravity, not by contract

(The following is analysis based on verified facts, not a reported statement by either company.)

The openness pledge is likely sincere in intent and almost certainly insufficient in practice. NVIDIA does not need to violate the 8-K's letter to bend the funnel its way. The lever is integration: deeper wiring of TensorRT-LLM, NIM microservices and DGX Cloud into Hugging Face's deployment path makes NVIDIA hardware the path of least resistance even when AMD or Intel chips remain "supported." Developers rarely re-benchmark when the default "deploy" button is already optimized. This is lock-in by gravity rather than by contract — and it is the model NVIDIA has already perfected with CUDA.

What the deal signals more broadly is the closing of the open-AI stack. For three years the narrative was "open weights vs. closed labs"; the more durable consolidation is horizontal, across compute, models and distribution, with NVIDIA as the integrator. If the FTC or EU forces real neutral-access remedies, this becomes a template for how a dominant hardware vendor can own the developer gateway without technically excluding competitors. If they don't, the 86x-multiple bet only pays off if "open" and "NVIDIA-default" coexist — a balance that will be tested the moment a model optimized for a rival accelerator underperforms on Hugging Face's recommended stack.