NVIDIA plans to acquire Hugging Face for 14 billion U.S. dollars, which will become the largest acquisition in the AI sector this year.

On September 2, according to The Financial Association, Nvidia is reportedly about to reach an agreement to acquire Hugging Face for a value of nearly 14 billion U.S. dollars this week.

As of now, the final terms of the transaction have not been finalized, and there is still room for adjustment in the transaction price and schedule. Neither party has released an official announcement and both have declined to comment.

It is reported that Nvidia is in-depth negotiations to acquire artificial intelligence startup Hugging Face, with the total transaction amount likely to be around 14 billion U.S. dollars. People familiar with the matter revealed that the cash consideration for this transaction is about 12.9 billion U.S. dollars, which may be reached as early as this week; in addition, a 1 billion U.S. dollar employee retention incentive plan is set up to retain Hugging Face's core R&D personnel.

If this acquisition is successfully completed, this tens-of-billion-dollar transaction will become one of Nvidia's largest M&A projects so far, and it is also the highest-value acquisition case in the global artificial intelligence track since the beginning of this year, which will rewrite the competitive landscape of the global open-source AI industry.

The negotiation process has drawn high attention from the Silicon Valley tech circle. Some market analysts believe that after the subsequent transaction is completed, it will face strict review by U.S. anti-monopoly regulators, and the market competition risks brought by the integration of software and hardware ecosystems will become the key assessment direction of the regulatory authorities.

01

Hugging Face was founded in 2016 by three French entrepreneurs in New York, the United States. The company's original dream was to build a fun chatbot for teenagers. After open-sourcing the Transformers tool library in 2018, the company's development route ushered in a major turning point, and it gradually transformed into an AI model sharing and collaboration platform for global developers, known in the industry as "GitHub in the AI field".

Public operation data shows that the Hugging Face Hub platform currently hosts more than 3 million public AI models, the number of datasets launched has exceeded 1 million, bringing together more than 13 million developers from all over the world, with more than 18 million monthly active visitors, about 5 million registered users, and more than 2,000 enterprises paying to use its Enterprise Hub. A large number of open-source model R&D teams at home and abroad such as Meta, Google, DeepSeek, Zhipu AI and others all take this platform as the primary channel for external model distribution and version iteration.

In April last year, they acquired French open-source robotics company Pollen Robotics and started selling robots. In addition, the company has also launched paid enterprise hosting, model inference deployment, and customized development services, gradually building a commercial revenue system, with its business extending to almost every corner of the AI industry chain. According to foreign media estimates, Hugging Face's current annualized operating income is about 150 million U.S. dollars, and its revenue growth rate is in a rapid upward channel.

From the perspective of capital background, Nvidia was already an investor as early as the D-round financing stage of Hugging Face in 2023. Other tech giants that participated in the investment in the same period included Google, Amazon, Intel, Salesforce and others. After the completion of this round of financing, Hugging Face's post-investment valuation was 4.5 billion U.S. dollars. Compared with the current acquisition offer of nearly 14 billion U.S. dollars, the valuation has increased by nearly three times.

02

Earlier this year, Hugging Face once rejected Nvidia's 500 million U.S. dollar investment offer. The two sides moved to full acquisition negotiations after a few months, which also reflects Nvidia's determination to lay out the open-source track from the side.

Looking at Nvidia's strategic layout in recent years, the high-priced acquisition of Hugging Face is not a short-term financial investment, but a key layout focusing on the right to speak in the mid-to-long term AI ecosystem.

The most direct strategic benefit is to take the core entrance of global open-source model distribution, and complete the closed loop of the industry chain from the underlying computing power hardware, CUDA development framework to the upper-level model circulation channel. For a long time, Nvidia's commercial advantage has been rooted in the GPU computing power hardware market, while Hugging Face has massive developer resources. After the acquisition is completed, Nvidia's computing power products can directly reach the open-source model R&D group, opening up the path between computing power supply and model development.

At the same time, this acquisition is also regarded by the market as a defensive measure for Nvidia to hedge potential industry risks. At present, a number of closed-source large model vendors such as OpenAI and Anthropic are accelerating the self-development of AI server chips, aiming to reduce their business's dependence on Nvidia's GPU products, and the competitive landscape of the computing power market is gradually moving towards diversification. Nvidia management's core judgment is that the more prosperous the open-source AI ecosystem is, the more scattered the global computing power demand will be, and it will be difficult for a single closed-source vendor that develops chips independently to cut off Nvidia's basic market.

In addition, by taking over Hugging Face, Nvidia will take the initiative in the direction of open-source model adaptation and community technology trends. Combined with Nvidia's own Nemotron open-source large model project, it is expected to realize the integrated ecological collaborative layout of hardware computing power, self-developed models and open-source community, consolidating its position as the infrastructure in the global artificial intelligence industry.

Since the beginning of this year, Nvidia has also successively completed a number of acquisition projects in the software ecosystem field. In the first half of the year, the company completed the acquisition of Israeli computing power scheduling vendor Run:ai. This transaction has been approved by EU regulators, and the related products will strengthen Nvidia's supporting capabilities in GPU cluster resource scheduling and hybrid cloud computing power management. In June, Nvidia completed the full acquisition of U.S. predictive AI startup Kumo-AI for consideration of no less than 400 million U.S. dollars.

At present, in the Silicon Valley tech M&A market, the AI middle-layer infrastructure has become the target track that all major giants are scrambling for. Not long ago, U.S. online payment service provider Stripe just acquired AI model routing platform OpenRouter for more than 7 billion U.S. dollars. Nvidia's this tens of billions of dollars negotiation will further set off an acquisition boom in the open-source platform field.

This article is from "Jiemian News", written by Shen Xiaoge, and published with authorization by 36Kr.