Post a Comment Print Share on Facebook

Nvidia, Google, Amazon... The heavyweights of AI enter the capital of the French champion Hugging Face

Nvidia, Google, Amazon, Intel, AMD, IBM and Qualcomm.

- 4 reads.

Nvidia, Google, Amazon... The heavyweights of AI enter the capital of the French champion Hugging Face

Nvidia, Google, Amazon, Intel, AMD, IBM and Qualcomm. This prestigious list of champions of artificial intelligence, whether through their components or their cloud capabilities, is that of the new investors of Hugging Face, one of the main players in open source AI. All of these groups participated in a funding round, announced this Thursday and led by Salesforce Ventures, of 235 million dollars. This operation more than doubles the valuation of this company, created by French people and operating between Paris and New York. She is now worth $4.5 billion.

Hugging Face, named after this emoji that reaches out for a hug, was previously financed by American funds. Its last fundraising, of 100 million dollars, dated from May 2022. “For this new round, we wanted to bring together a large number of key players in artificial intelligence. These are groups with which we already have partnerships,” explains Julien Chaumond, co-founder and chief technology officer.

In early August, Nvidia announced that Hugging Face services would be hosted on its DGX cloud, which houses supercomputers specializing in AI. "This collaboration will allow companies to take their destiny in AI into their own hands thanks to open source", commented then the general manager Clément Delangue. Hugging Facing similar deals with Amazon.

But unlike some fundraisers that result in the obligation for start-ups to donate part of the money raised in cloud hosting, Hugging Face has no spending constraints on its new investors. “And, with the exception of Salesforce Ventures, their tickets are identical,” underlines Julien Chaumond. Sound Ventures, the private equity firm of American actor Ashton Kutcher, is also participating in the funding round.

Unknown to the general public, Hugging Face is a sharing platform specializing in AI and machine learning. Developers from all over the world can deposit pre-trained models or datasets there, which will be reused, adapted and improved by others. For example, Meta's AI lab chose Hugging Face to host Llama 2, its latest language model. And groups like Renault, Pzifer or Roche draw from this open source library to design their AI-powered services without having to start from scratch.

This approach is the opposite of that of OpenAI, which sells access to the power of its GPT-4 language model but prevents companies from lifting the hood. Clément Delangue was able to express his views on these differences in orientation during a hearing in June in the American Congress, which was conducting a fact-finding mission on AI developments.

Hugging Face engineers participate in this collaborative research. The new fundraiser will allow the teams to grow. The company currently has 170 employees, 80 of whom work in France. "We have opened several small offices in the country, as well as in Europe and the United States," says Julien Chaumond. Hugging Face succeeds in attracting profiles attracted by the open source spirit of the company. “But there are currently a lot of investments in AI start-ups”, such as the 105 million dollars raised by the French Mistral AI (see page 26), “and it is on these new projects that the competition is played out to attract talent", continues the technical director.

The money raised will also strengthen the company's cash flow and allow it to consider acquisitions. “But there are no specific plans at the moment”, specifies Julien Chaumond. Hugging Face, which also sells business services, does not disclose its revenues. But according to The Information, they should be 30 million dollars for the year 2023.

Avatar
Your Name
Post a Comment
Characters Left:
Your comment has been forwarded to the administrator for approval.×
Warning! Will constitute a criminal offense, illegal, threatening, offensive, insulting and swearing, derogatory, defamatory, vulgar, pornographic, indecent, personality rights, damaging or similar nature in the nature of all kinds of financial content, legal, criminal and administrative responsibility for the content of the sender member / members are belong.