Meta News

Nvidia is reportedly closing in on one of its most strategically important acquisitions yet:Hugging Face, the platform that has become a central hub for sharing, downloading and deploying open AI models. The Information reports that Nvidia has agreed to acquire the company for $12.9 billion, although other reporting indicates that a final agreement may not yet have been signed and negotiations could still collapse. Neither company has publicly confirmed the transaction.

If completed, the deal would give Nvidia control of one of the most influential platforms in the open-source AI ecosystem. Hugging Face hosts models, datasets and development tools used by researchers, startups and enterprises around the world, making it an important layer between AI developers and the computing infrastructure required to train and run their models.

Nvidia’s dominance of AI computing has been built primarily around GPUs and CUDA, but the largest AI companies are increasingly developing their own chips. Google, Amazon, OpenAI and Anthropic are all working to reduce their dependence on Nvidia hardware, creating a long-term strategic challenge even as demand for Nvidia’s accelerators remains enormous.

Owning Hugging Face would provide Nvidia with a different way to defend its position. Open models give developers alternatives to proprietary platforms operated by the major AI labs, and many of those models ultimately run on Nvidia hardware. Supporting a strong open-source ecosystem therefore helps preserve a large and diverse customer base for Nvidia’s computing infrastructure.

The company has already invested heavily in open AI models of its own. Hugging Face would give it something potentially more valuable: direct access to the community and infrastructure through which much of the open AI ecosystem is distributed.

Founded in 2016, Hugging Face evolved from a chatbot startup into what is effectively a GitHub-like platform for machine learning. Developers can publish models, download existing ones, share datasets and deploy workloads without building all of the underlying infrastructure themselves.

That position has become increasingly valuable as open-weight models improve. Chinese AI laboratories in particular have released models capable of competing with leading American systems on several benchmarks while often operating at substantially lower costs. Hugging Face CEO Clem Delangue has warned that China is becoming increasingly competitive in open-source AI and has publicly supported policies that keep open models available in the United States.

Nvidia and Hugging Face have already aligned themselves politically around that issue. Jensen Huang, Delangue and representatives from other technology companies recently signed a letter urging the US government to support rather than restrict open models.

The acquisition would not only strengthen Nvidia’s software position. It could provide another route into the cloud computing market.

Nvidia previously attempted to expand its DGX Cloud business but reportedly scaled back those ambitions. Hugging Face already allows developers to run AI workloads using rented computing infrastructure, providing Nvidia with an existing distribution channel rather than forcing it to build another cloud platform from scratch.

There is also a financial incentive. Nvidia has committed to supporting tens of billions of dollars in cloud computing agreements involving customers. If those customers fail to consume all the capacity they reserve, Nvidia could potentially be exposed to unused computing resources. Hugging Face’s large developer community could provide another market through which that capacity is sold.

This would move Nvidia further up the AI stack. Instead of merely supplying processors to companies running AI workloads, it could participate directly in distributing models and providing the infrastructure on which developers execute them.

A price approaching $13 billion would represent an extraordinary increase in Hugging Face’s valuation. The company raised $235 million in 2023 at a valuation of $4.5 billion, with investors including Salesforce Ventures, Google’s GV, IBM Ventures and Nvidia itself.

Hugging Face also reportedly rejected a $500 million Nvidia investment late last year that would have valued the company at approximately $7 billion. At the time, the company was concerned that allowing a dominant investor to gain too much influence could compromise its independence.

The economics now look considerably different. Hugging Face is reportedly generating around $150 million in annual revenue, up from approximately $100 million only two months earlier, and Delangue recently said the business was approaching profitability. Even with that rapid growth, a $12.9 billion acquisition would represent an enormous revenue multiple.

For Hugging Face, however, joining Nvidia would provide access to vastly greater financial and computing resources at a moment when building competitive AI infrastructure is becoming increasingly expensive.

The potential acquisition also fits a broader consolidation trend. Companies that provide neutral infrastructure between developers and AI models are becoming strategically valuable to much larger technology companies.

TechCrunch points to the recent acquisition of OpenRouter by Stripe as another example. OpenRouter, which helps developers route workloads between different AI models, had been valued at roughly $1.3 billion in May before Stripe reportedly agreed to acquire it for more than $7 billion.

Hugging Face occupies an even more influential position because it sits at the intersection of model distribution, open-source development and AI compute. Nvidia acquiring it would combine one of the industry’s dominant hardware suppliers with one of its most important model-distribution platforms.

That combination could be enormously powerful, but it would also raise questions about Hugging Face’s neutrality. Developers currently use the platform across different clouds, hardware architectures and AI frameworks. Under Nvidia ownership, competitors and developers would inevitably scrutinize whether that neutrality could be maintained.

The strategic significance of the acquisition is ultimately larger than Hugging Face’s current revenue. Nvidia’s most important competitive advantage has never been silicon alone; CUDA and the software ecosystem surrounding its GPUs have made switching away from Nvidia considerably more difficult.

Hugging Face could extend that strategy into the next layer of AI development. Nvidia would have hardware at the bottom, CUDA and development libraries above it, and one of the world’s largest communities for discovering and deploying AI models sitting closer to developers.

At the same time, open-source AI could become an increasingly important counterweight to vertically integrated platforms from OpenAI, Google, Anthropic and Amazon. Those companies want to control their models, infrastructure and increasingly their own chips. Nvidia has a different incentive: it benefits when thousands of independent companies and developers can build competitive AI systems without owning specialized hardware infrastructure themselves.

Acquiring Hugging Face would therefore be more than another AI acquisition. It would be a bet that the open AI ecosystem itself can become one of Nvidia’s strongest defenses against the growing number of technology giants attempting to reduce their dependence on Nvidia chips.

⁠Original report at TechCrunch