Nvidia buys Hugging Face for $12.9 billion

Nvidia said Thursday it is buying Hugging Face, a library of open artificial intelligence models, for $12.9 billion as the chipmaker ramps up spending in an escalating global race to dominate the technology.

The deal combines Nvidia’s chip and data center infrastructure with millions of open-source Hugging Face artificial intelligence models that can be freely downloaded and modified. And that puts Nvidia deep in its pockets on one side of the debate about how AI systems should be built.

“Together, we will make AI more open, capable and accessible to people and institutions around the world,” wrote Jensen Huang, CEO of Nvidia. in a blog post.

The acquisition is also another sign of Nvidia’s growing role as Silicon Valley’s central banker, using its vast financial resources to support artificial intelligence start-ups and bring money to customers for its chips. With $197 billion in current assets and nearly $60 billion in revenue last quarter, the Silicon Valley giant is spending big on AI.

Just last month, Nvidia and six giant investment firms said they were raising $500 billion in financing to help the chipmaker’s customers pay for computing power. Then Nvidia agreed to pay up to $105 billion to support one of the largest data centers in the world. Nvidia also said it would underwrite more data centers for other start-ups.

Nvidia has invested nearly $50 billion in artificial intelligence labs producing advanced models, Colette Kress, Nvidia’s chief financial officer, said on a call with investors and analysts last week. It called the investment a “meaningful commitment” but “a small fraction of our expected free cash flow.”

In recent months, Nvidia has emerged as a champion of open-source technology. Open source offers some advantages: It can be freely shared and is much more affordable than so-called closed systems built by AI companies like Anthropic. But critics argue that open source creates security risks that cannot be quickly resolved.

That debate has found its way to Washington, where some Trump administration officials have considered restrictions on open source AI, particularly models created by Chinese companies.

The purchase of Hugging Face shows how important open-source technology has become for Nvidia. Hugging Face is essentially an online open-source library of artificial intelligence models that can be freely downloaded and modified. It is widely used by technology developers. Nvidia and Hugging Face argued that AI developers must be able to create open models so that humans can further develop technology and build new businesses.

“Open models allow start-ups, enterprises, universities and public institutions to build on advanced capabilities without having to train each model from scratch. They enable organizations to match the right model to the right job,” Mr. Huang he wrote in Thursday’s blog post. “In this way, AI can advance safely, strengthen cybersecurity and sovereignty, accelerate innovation, and reach factories, hospitals, farms, classrooms, and Main Street businesses around the world.”

Both companies last month backed an industry letter slamming open AI models after Anthropic and OpenAI lobbied against the models and suggested Chinese start-ups were stealing their technology to create competitive open alternatives.

Nvidia addressed the debate in a securities filing on Thursday, noting that government restrictions could negatively impact its acquisition of Hugging Face.

“We are committed to supporting the training, distribution and use of closed and open AI models and applications,” the company said. “Other parties are actively lobbying the US government and other stakeholders around the world to adopt legislative or regulatory measures that would limit or disadvantage open source models and their customers.”

By acquiring Hugging Face, Nvidia gained deeper insight into the ways developers use AI and the models that drive it, said Jeff Pollard, vice president of research firm Forrester. “Hugging Face is where a lot of the AI ​​ecosystem exchanges models, datasets and code,” he said, adding that Nvidia “now sits closer to the entire AI supply chain.”

Hugging Face, which has raised more than $400 million in funding, said it was valued privately at $4.5 billion in 2023. Mr. Huang said other bidders have also tried to buy Hugging Face during an interview with CNBC on Thursday. Clément Delangue, CEO of Hugging Face, said his company and Nvidia signed the agreement on Wednesday.

Hugging Face launched in 2016 and its main product at the time was a chatbot app for teenagers. The start-up later became a repository for open-source AI and a destination for developers looking to customize AI tools.

Hugging Face took off with the advent of AI. In 2021, the year before OpenAI released its first chatbot and accelerated the AI ​​race, Hugging Face hosted 13,590 open-source models, the company said. Today it has almost three million.

New York-based Hugging Face became widely known after AI agents created by OpenAI broke away from its developers and hacked into its network. Mr. Delangue began a publicity campaign soon after the incident.

There’s no doubt that Nvidia has the resources to spend. As recently as three years ago, Nvidia’s quarterly profit was $6.2 billion, roughly a tenth of what it reported last quarter. The company also said revenue more than doubled from a year earlier to $96.22 billion. Nvidia’s share price closed up 3.6 percent on Wednesday and was up about 1 percent on Thursday.

But that growing list of financial deals raises concerns that the AI ​​boom is being fueled by increasingly risky bets, with Nvidia at the center of many of them. Using unusual deals with chip makers, cloud computing providers and governments, AI startups are gaining access to computing power they couldn’t afford on their own and capital that lenders wouldn’t otherwise provide.

Many of these deals have drawn criticism for being circular. Some AI startups receive billions of dollars from tech giants before spending those billions with the same companies to pay for chips, cloud computing and other services.

In July, for example, Nvidia announced a partnership with Safe Superintelligence, a start-up created by the founder of OpenAI. The chip maker invested multibillion dollars in the start-up and provided computing power for its research. It concluded a similar agreement in March with the Thinking Machines Lab, which was founded by a former OpenAI manager.

Nvidia said in its filing that its acquisition of Hugging Face includes incentives worth approximately $1 billion for the startup’s employees who remain in the deal.