⏱️ Lectura: 9 min
The Nvidia Hugging Face negotiation already has a concrete figure on the table: over $13 billion dollars for the platform that hosts millions of open models and datasets used daily by developers around the world.
📑 En este artículo
- TL;DR
- Introduction: why the Nvidia Hugging Face deal matters
- What happened
- Context and history of the Nvidia Hugging Face relationship
- Technical details: what Nvidia would control
- How to try it today without waiting for the deal to close
- Impact and analysis
- What’s next
- Frequently Asked Questions
- Has Nvidia confirmed the Hugging Face acquisition?
- How much is Hugging Face worth according to this negotiation?
- Why had Hugging Face previously rejected Nvidia’s money?
- What happens to Hugging Face’s neutrality if Nvidia buys it?
- Does Microsoft also want to buy Hugging Face?
- How can I use Hugging Face today without depending on who owns it?
- References
If the deal goes through, it would be one of the largest acquisitions in Nvidia’s history and would give it direct control over the central repository of the open AI ecosystem: the same one that today also serves models optimized for competitors’ hardware like AMD and Intel.
TL;DR
- Nvidia is negotiating to buy Hugging Face for over $13 billion dollars, according to Business Insider (August 27, 2026).
- The talks are recent and might not close, warns the source cited by the outlet.
- Microsoft also met with Hugging Face, but those talks are not currently active.
- Nvidia already invested in Hugging Face in 2023: it took part in a $235 million round that valued the company at $4.5 billion.
- In late 2025, Hugging Face rejected a $500 million investment offer from Nvidia that valued it at $7 billion.
- Nvidia has $18 billion committed in capital investments for the rest of its fiscal year, on top of the $47.9 billion it already holds in private companies.
- Hugging Face was founded in 2016 by Clément Delangue, Julien Chaumond, and Thomas Wolf.
- The platform hosts millions of models and datasets used by developers worldwide, including hardware from Nvidia rivals like AMD and Intel.
Introduction: why the Nvidia Hugging Face deal matters
For any developer working with AI in Latin America, Hugging Face isn’t just another name: it’s probably the site where you downloaded your last language model, uploaded your own dataset, or tested a Gradio space before taking it to production. The platform works as a kind of central repository for machine learning models, and that position is exactly what makes it valuable to Nvidia.
The Nvidia Hugging Face negotiation isn’t an isolated rumor. Nvidia has already been an investor in the company since 2023 and had tried to deepen that relationship in 2025 with an offer that Hugging Face rejected. That a figure nearly double that rejected valuation is now being discussed shows how much weight the hub carries in Nvidia’s strategy for 2026.
What happened
According to Business Insider, Nvidia and Hugging Face held acquisition talks in recent weeks over a deal that would value the platform at over $13 billion dollars. The figure comes from a source with direct knowledge of the process, and the outlet clarifies two important things: the companies haven’t reached an agreement yet, and the talks could fall through before anything is signed.
Nvidia and Hugging Face did not respond to the outlet’s requests for comment. Nor is it the only company that has shown interest: Microsoft also met with Hugging Face to explore a possible deal, although according to two sources cited by Business Insider those talks are not active at this time.
The report comes at a time when Nvidia is actively spending its cash to secure positions across the entire AI chain. The company reported the Wednesday before the story broke that it has $18 billion committed in capital investments for the rest of its fiscal year, on top of the $47.9 billion it already holds in private companies.
Context and history of the Nvidia Hugging Face relationship
The relationship between the two companies didn’t start this week. Nvidia took part in a $235 million funding round in 2023 that valued Hugging Face at $4.5 billion. It was a minority investment, without control, consistent with Nvidia’s role as a financial backer of much of the AI ecosystem.
In late 2025, as the Financial Times had previously reported, Nvidia offered to invest an additional $500 million at a $7 billion valuation. Hugging Face rejected the offer: the company explained at the time that it didn’t want a dominant investor capable of influencing its decisions. That refusal is what makes it especially notable that a full sale is now being discussed for nearly double that rejected valuation.
| Moment | Amount | Implied valuation | Outcome |
|---|---|---|---|
| 2023: funding round | $235 million (Nvidia’s contribution) | $4.5 billion | Closed; Nvidia enters as a minority investor |
| Late 2025: investment offer | $500 million | $7 billion | Rejected by Hugging Face |
| 2026: acquisition talks | Over $13 billion | Over $13 billion | In negotiation, no deal closed |
Hugging Face was founded in 2016 by French entrepreneurs Clément Delangue, Julien Chaumond, and Thomas Wolf. What started as a chatbot app ended up becoming the go-to repository for open models: today it hosts millions of models and datasets, and its transformers library is one of the most widely used pieces of software for loading language models in Python.
Technical details: what Nvidia would control
Hugging Face isn’t just a website, it’s infrastructure. Its Hub distributes model weights, model cards with metrics and licenses, versioned datasets, and Spaces where anyone can deploy a demo with Gradio without managing a server. All of that runs on a storage layer and an API that thousands of production pipelines query every day.
For Nvidia, controlling that layer means something concrete: greater influence over which runtime and weight format becomes the standard. Nvidia already pushes TensorRT-LLM and its NIM microservices as the preferred way to serve models on its own GPUs. If Hugging Face were to integrate more closely with those tools, every download from the Hub could tip the balance toward an inference path optimized for Nvidia hardware, instead of staying neutral among manufacturers.
flowchart TD
A["Developer"] --> B["Hugging Face Hub"]
B --> C["Model or dataset"]
C --> D["huggingface-cli download"]
D --> E["Nvidia GPU (CUDA / TensorRT)"]
subgraph Inference flow
D
E
end
How to try it today without waiting for the deal to close
While the negotiation moves forward (or falls apart), the Hub keeps working just like always and any developer can use it today. The first step is installing the client libraries:
# Windows (PowerShell)
pip install huggingface_hub transformers
huggingface-cli login
# macOS / Linux (bash)
pip install huggingface_hub transformers
huggingface-cli loginThat command installs the Hub client and the transformers library, and starts the authentication flow to link your personal token.
With the CLI authenticated, you can now download a full model to disk and load it locally:
from huggingface_hub import snapshot_download
from transformers import AutoModelForCausalLM, AutoTokenizer
model_id = "meta-llama/Llama-3.2-1B"
snapshot_download(repo_id=model_id, local_dir="./modelo-local")
tokenizer = AutoTokenizer.from_pretrained("./modelo-local")
modelo = AutoModelForCausalLM.from_pretrained("./modelo-local", device_map="cuda")
entrada = tokenizer("How does the Hugging Face Hub work?", return_tensors="pt").to("cuda")
salida = modelo.generate(**entrada, max_new_tokens=50)
print(tokenizer.decode(salida[0]))That block downloads the model’s weights to a local folder, loads the tokenizer and model, and generates a response from a prompt. With device_map="cuda", inference runs directly on your Nvidia GPU if one is available.
💡 Tip: if you’re worried about depending on a single provider, snapshot_download saves a full copy of the model to your disk. Once downloaded, you don’t need a connection to the Hub to keep using it.
Impact and analysis
The biggest risk the reports themselves point to is the loss of neutrality. Hugging Face hosts models trained and optimized for AMD and Intel chips too, direct Nvidia competitors. Having the hub’s owner also be the market’s dominant GPU maker changes the incentives: a neutral platform choosing what to show first is not the same as a platform whose owner makes money when the user picks a specific piece of hardware.
⚠️ Heads up: Hugging Face had already rejected a Nvidia offer before, precisely to avoid ceding that kind of influence. That it’s now negotiating a full sale suggests the calculus changed, or that competitive pressure increased.
For developers in Latin America, the practical effect will depend on whether the Hub stays open on the same terms. Today, uploading or downloading a model costs nothing beyond bandwidth, and the formats are compatible with any hardware that supports PyTorch. A change of ownership won’t alter that overnight, but it is a move worth monitoring if your stack depends on the Hub in production.
What’s next
The deal still hasn’t closed, and the very source cited by Business Insider warns it might not happen. If it moves forward, it’s reasonable to expect antitrust regulators to ask questions given the weight Nvidia already carries across the AI chain: GPUs, inference software, and now potentially the model distribution hub. It also remains to be seen whether Microsoft resumes its own talks with Hugging Face if the deal with Nvidia falls through.
📖 Summary on Telegram: View summary
Try it yourself: run pip install huggingface_hub transformers and download a model from the Hub today to see firsthand the infrastructure Nvidia wants to buy.
Frequently Asked Questions
Has Nvidia confirmed the Hugging Face acquisition?
No. Both companies declined to comment, and according to the source cited by Business Insider, the talks might not lead to a deal.
How much is Hugging Face worth according to this negotiation?
Over $13 billion dollars, according to the August 27, 2026 report.
Why had Hugging Face previously rejected Nvidia’s money?
Because it didn’t want a dominant investor that could influence its decisions, as the company explained when it rejected the $500 million offer in late 2025.
What happens to Hugging Face’s neutrality if Nvidia buys it?
It’s the main concern: the platform hosts models optimized for competitors’ hardware like AMD and Intel, and an acquisition could compromise that neutrality.
Does Microsoft also want to buy Hugging Face?
It met with the company, but according to the sources cited by Business Insider those talks are not currently active.
How can I use Hugging Face today without depending on who owns it?
By installing the huggingface_hub and transformers libraries, and downloading models locally with snapshot_download, as shown in the “How to try it today” section.
References
- Business Insider: exclusive report on the negotiations between Nvidia and Hugging Face.
- Hugging Face: official site of the platform and its open model hub.
- Nvidia: official site of the GPU maker and its AI investment strategy.
- huggingface/transformers: official repository of the library used to load and run models from the Hub.
- Wikipedia: history and founding of Hugging Face in 2016.
📱 Like this content? Follow @programacion on Telegram for daily tech content in Spanish: quick summaries, fresh content every day.
Imagen destacada: Foto de Brecht Corbeel en Unsplash
0 Comments