NVIDIA has agreed to acquire Hugging Face, the world’s largest open source AI platform, for $12.93 billion in cash and equity, in what is its largest acquisition to date. Announced on 3 September 2026 through a blog post by Chief Executive Jensen Huang, the deal brings a platform used by more than 18 million developers and 200,000 companies under the world’s most valuable chipmaker. The move signals a decisive push by NVIDIA beyond hardware into the software and distribution layer where open models are built and deployed.
What Is Hugging Face and What Is NVIDIA?
Hugging Face, Inc. is an American artificial intelligence company based in Manhattan, New York City. It was founded in 2016 by French entrepreneurs Clément Delangue (Chief Executive Officer), Julien Chaumond (Chief Technology Officer) and Thomas Wolf (Chief Scientific Officer). The name comes from the hugging face emoji, reflecting its origin as a chatbot app for teenagers. After open sourcing the model behind that chatbot, the founders pivoted to building tools for machine learning.
Today Hugging Face is best known as the home of open source AI. Its platform, often called the Hugging Face Hub, hosts more than 3 million models, 500,000 datasets and 1 million applications. More than 18 million developers, researchers and creators use it to share and discover models, and over 200,000 companies use it to evaluate, customise and deploy AI. Its most influential product is the Transformers library, an open source software library for natural language processing and other tasks that has more than 164,000 stars on GitHub. Other key products include Datasets, Spaces for hosting interactive AI demos, Inference API and AutoTrain. The company was valued at $4.5 billion in August 2023 after a $235 million funding round backed by Google, Amazon, NVIDIA, Salesforce, AMD, Intel, IBM and Qualcomm. It remains a private company.
NVIDIA Corporation is an American multinational technology company headquartered in Santa Clara, California. It was founded on 5 April 1993 in Sunnyvale, California by Jensen Huang, Chris Malachowsky and Curtis Priem. The founders first sketched the idea at a Denny’s diner in East San Jose. NVIDIA invented the Graphics Processing Unit (GPU) in 1999, which transformed gaming and later enabled accelerated computing. In 2006, it launched CUDA (Compute Unified Device Architecture), a parallel computing platform that lets developers use GPUs for general computing. CUDA became the software foundation for modern artificial intelligence.
NVIDIA is listed on NASDAQ under the ticker NVDA and is a component of the Nasdaq-100, Dow Jones and S&P 500. It reported record revenue of $215.9 billion for fiscal year 2026, with data centre revenue accounting for the largest share, and employs more than 42,000 people in 38 countries. Jensen Huang has served as President and Chief Executive Officer since inception.
Details of the $12.93 Billion Deal
The definitive agreement was entered into on 2 September 2026 and announced publicly on 3 September 2026. According to the filing with the United States Securities and Exchange Commission (SEC), the transaction is valued at $12.93 billion in total. This includes about $11.9 billion payable to Hugging Face stockholders, subject to adjustments, and an equity based retention programme of up to $1 billion for Hugging Face employees who join NVIDIA.
The deal is structured as a cash and equity transaction. It will be funded from NVIDIA’s balance sheet, which held more than $60 billion in cash and short term investments in late 2025. The transaction is expected to close in the first half of 2027, subject to customary closing conditions. These include approval by regulators in relevant jurisdictions and other standard requirements. NVIDIA has stated that regulatory review is expected to be substantive given its dominant position in AI chips and Hugging Face’s central role in open model distribution.
Transaction at a Glance
| Item | Detail |
|---|---|
| Buyer | NVIDIA Corporation |
| Target | Hugging Face, Inc. |
| Total value | $12.93 billion |
| Cash to stockholders | About $11.9 billion |
| Employee retention equity | Up to $1 billion |
| Announcement date | 3 September 2026 |
| Expected closing | First half of 2027 |
| Mode | Cash and equity |
| Regulatory condition | Required approvals must be received |
Why Is NVIDIA Buying Hugging Face?
The acquisition is a bet on open source AI at a time when open models are closing the gap with closed systems from OpenAI and Anthropic but at much lower cost. By owning the largest distribution hub for open models, NVIDIA extends its influence from chips into the software layer where developers choose, fine tune and deploy models.
Three strategic drivers stand out. First, platform control. Hugging Face is where models are discovered and shared, similar to what GitHub is for code. Owning this layer gives NVIDIA visibility into which models are popular, how they are used, and where inference demand is growing. Second, AI factory vision. Jensen Huang has described NVIDIA as a full stack AI infrastructure company that powers AI factories, large scale computing clusters that train and run models. Integrating Hugging Face helps connect those factories directly to developers. Third, defence of the ecosystem. Open models run on many chips, including those from AMD and Intel. By committing to keep Hugging Face neutral, NVIDIA can position itself as the steward of openness while still ensuring its hardware is optimised for the most downloaded models.
The timing also matters. Competition for AI talent and intellectual property is intense, with large licensing deals becoming common. Securing a trusted developer community of 18 million users gives NVIDIA a long term distribution advantage that is difficult to replicate by building a new platform.
What Is Open Source AI and How Does Hugging Face Work?
Open source AI refers to artificial intelligence models whose weights, code and often training data are made freely available for anyone to use, modify and share. This is different from closed source AI, where the model is kept private and access is controlled through an application programming interface. Open source models give developers more control, lower costs and the ability to adapt the model for specific needs.
Hugging Face makes this easier through several interconnected tools. The Transformers library is a standard framework that defines how models are built for text, vision, audio and multimodal tasks. It allows a single model definition to work across training frameworks like PyTorch and inference engines like vLLM, which has made it the default way researchers share new architectures. The Hugging Face Hub is a cloud platform that hosts models, datasets and applications. Developers can upload a model with one command and others can download and run it, and the Hub now holds more than three million public models and around half a million datasets. Spaces lets users host interactive demos of models in the browser, while Inference API and Inference Endpoints let companies deploy models without managing their own servers. The Datasets library simplifies sharing and processing of large datasets, and Gradio, which Hugging Face acquired in 2022, helps developers build simple web interfaces for machine learning apps in Python.
In simple terms, Hugging Face acts as both a library and a marketplace. It standardises how models are written and provides the warehouse where they are stored, found and launched.
Is Hugging Face Free and Is It Open Source?
Access to the Hub is free for public models and datasets, and the Transformers library is open source under the Apache 2.0 licence. The company also offers paid tiers for private hosting, enterprise security, dedicated inference and support. It earns revenue from these paid services and from partnerships with cloud providers. Users can run models locally or in the cloud, and they can use many models commercially depending on the model’s own licence.
How Does This Compare With NVIDIA’s Past Acquisitions?
NVIDIA has historically grown through small, technology focused purchases. Its largest completed acquisition before this deal was Mellanox Technologies, an Israeli networking company that makes InfiniBand and high speed Ethernet products for data centres. NVIDIA announced that deal in March 2019 for $6.9 billion, funded entirely in cash, and closed it in April 2020. The combination united NVIDIA’s computing platform with Mellanox’s networking technology, enabling higher performance for supercomputers and cloud data centres.
In December 2025, NVIDIA entered a non exclusive licensing agreement with AI chip startup Groq for about $20 billion, accompanied by the hiring of Groq founder Jonathan Ross and other senior leaders. Groq will continue to operate independently under its cloud business, GroqCloud. Although widely reported as NVIDIA’s largest transaction, it was structured as an asset licence and talent addition rather than a full company acquisition, and NVIDIA itself described it as not acquiring Groq as a company.
Viewed strictly as a company acquisition, the Hugging Face deal at $12.93 billion is therefore the largest full acquisition NVIDIA has undertaken, surpassing Mellanox. If the Groq licence is counted as a transaction, it remains larger in headline value but different in structure.
| Transaction | Year Announced | Value | Type | Purpose |
|---|---|---|---|---|
| Mellanox Technologies | 2019 | $6.9 billion | Full acquisition, all cash | Data centre networking, InfiniBand and Ethernet |
| Groq assets and licence | 2025 | About $20 billion | Non exclusive licence plus hiring of team | Low latency inference processors |
| Hugging Face | 2026 | $12.93 billion | Full acquisition, cash and equity | Open source AI platform and developer community |
Failed bids also provide context. NVIDIA’s proposed $40 billion acquisition of Arm in 2020 was abandoned in 2022 after regulatory opposition.
Why Will Hugging Face Remain an Open Platform?
The central concern after the announcement was whether a chipmaker that competes with AMD and Intel could remain neutral while owning the main venue where developers compare hardware. Jensen Huang addressed this directly in the blog post and in the SEC filing.
He stated that Hugging Face will remain an open platform for the entire AI ecosystem. Developers will continue to choose the models they want, the frameworks they want, the clouds and inference providers they want, and the computing platforms they want. NVIDIA compute will not be required to build on or deploy through Hugging Face.
Specific commitments included continuing to support open source and open weight models from every model builder, and continuing to support multi cloud and multi accelerator development and deployment. The filing added that Hugging Face would continue to permit uploading and downloading of models and datasets of the user’s choosing and support other silicon vendors.
Huang pointed to a recent open letter on open weights that he co signed with researchers and executives across the industry, arguing that open models broaden access, strengthen competition and give developers control over deployment. The founders and existing team of Hugging Face are expected to stay on, and NVIDIA said it will use its infrastructure and engineering resources to improve platform reliability, safety, model evaluation and deployment without closing the ecosystem.
Significance for the Global AI Ecosystem and India
For the global AI ecosystem, the deal consolidates two critical layers, chips and model distribution, under one company. On the positive side, Hugging Face can gain stronger infrastructure, better security, faster inference and global reach. NVIDIA’s engineering investment could improve uptime and evaluation tools for the millions of models hosted. For developers, this could mean more reliable access and smoother deployment.
At the same time, it raises questions about concentration. Hugging Face is viewed as a neutral warehouse for open weights, and trust in that neutrality is a core asset. If developers fear that search rankings, recommendations or optimisation will favour NVIDIA hardware, they could shift activity to alternatives. Regulators in the United States and Europe are therefore expected to examine whether the deal affects competition in AI infrastructure and model distribution.
For India, the implications are direct. India is one of the largest user bases for Hugging Face, with students, startups and researchers using open models for natural language processing in Indian languages, agriculture analytics and education tools. NVIDIA already has a significant presence in India with offices in Bengaluru, Pune and Hyderabad, and runs programmes for startups and developers such as NVIDIA Inception, which supports more than 15,000 startups globally, and the NVIDIA Developer Program with over 7.5 million members. Lower friction access to open models, along with cloud agnostic deployment, could help Indian startups build cost effective applications without depending on expensive closed APIs.
However, cost and access remain watch points. Even though models are open, running them at scale requires compute. The promise that NVIDIA hardware will not be mandatory will be tested in how inference pricing and optimisation evolve on the platform.
The Way Forward
The acquisition is not yet complete. Closing is expected in the first half of 2027, pending regulatory approvals and other customary conditions. During this period, both companies will continue to operate independently. Hugging Face will keep publishing models and datasets, and NVIDIA will continue to expand its AI factory architecture that integrates computing, networking and software.
Observers will watch three milestones. First, the outcome of antitrust reviews and whether any conditions are imposed to preserve openness. Second, how NVIDIA invests in Hugging Face’s roadmap for reliability, safety and evaluation while honouring multi accelerator support. Third, whether the founders’ continued leadership maintains developer trust, which was a key reason Hugging Face became the default home for open source AI.
If those tests are met, the combined entity could accelerate the adoption of open models worldwide. If neutrality erodes, the deal could spur the growth of alternative hubs, which would fragment the community it seeks to unite.
Key Takeaways
- NVIDIA agreed to acquire Hugging Face for $12.93 billion in cash and equity, announced on 3 September 2026 by CEO Jensen Huang.
- The deal includes about $11.9 billion to stockholders and up to $1 billion in equity retention for employees, with closing expected in first half of 2027.
- Hugging Face, founded in 2016 in New York by Clément Delangue, Julien Chaumond and Thomas Wolf, hosts over 3 million models and 500,000 datasets for 18 million developers.
- Hugging Face was valued at $4.5 billion in August 2023 after a $235 million round backed by Google, Amazon, NVIDIA and others.
- As a full company acquisition, it is NVIDIA’s largest to date, surpassing Mellanox ($6.9 billion in 2019), while the Groq licence (about $20 billion in 2025) was a larger transaction structured as a licence, not a full acquisition.
- NVIDIA, founded on 5 April 1993 and headquartered in Santa Clara, California, said Hugging Face will remain an open, multi cloud and multi accelerator platform with no requirement to use NVIDIA compute.
- NVIDIA trades on NASDAQ as NVDA and invented the GPU (1999) and CUDA (2006), the software platform central to modern AI training and inference.
