NVDA Strengthens AI Dominance with Hugging Face Acquisition

**SPEAKER_1** (0:00)
Hello, it's time to turn our attention to Nvidia after announcing it's acquiring Hugging Face yesterday. Let's welcome in our next guest to discuss that move, Yael Ossowski, the Deputy Director of the Consumer Choice Center joining us. Yael, great to have you on the show. Now, Nvidia is paying close to $13 billion for Hugging Face. It's its second largest acquisition ever.
What does Nvidia see here that makes Hugging Face worth that kind of premium?

**Yael Ossowski** (0:26)
Well, essentially Hugging Face is the playground for anyone who's playing with local AI models. And local AI models are just another way that Nvidia is providing those chips, those wonderful chips that we've been talking about that are propping up the US economy. The more people that are moving to local AI, that means that they're running inference on their own computers, they're doing the compute on premises in their businesses or at home.
They're going to be using Hugging Face as the actual repository where they can get all of those models, plug into them and play around with them. So Nvidia sees that the entire future of AI, yes, right now it's with the large frontier companies, but it'll continue to evolve and much more of it is going to come local. As consumers come online more, businesses want to have a lot more sovereignty over their data. There's going to be a huge move there and I think that's what Nvidia is thinking and overall a pretty good move.

**SPEAKER_1** (1:19)
Does this deal also signal to us that open source AI is becoming just as strategically important as some of proprietary models that are coming out of the bigger labs?

**Yael Ossowski** (1:31)
I do think so and I think every company that has played around with artificial intelligence, they've gotten some token spend in there. They've noticed how much it actually cost. Engineers are sometimes putting together hundreds of thousands of dollars of token spend. But if you're able to use that using open weight models or open source models, you can actually get rid of those costs altogether. I think the more exciting thing, especially for consumers and technology nerds like me, is that the models are getting better every single day. There's an entire competitive category. Hugging Face is the central place where all that is. It's the source of truth for open source AI. So you can find all of those models, you can plug into them, and you can change them basically at will. So I think we see that there is going to be a move. We've seen Palantir and many other companies that have really tried to put their foot down and say, look, we want sovereign AI. We want the ability to control our compute, not necessarily give it up just to open AI and Anthropic. I think you're going to see other companies do exactly the same thing.

**SPEAKER_1** (2:30)
Let's dive into some of those numbers that they're buying in this community, a developer community. They've got a repository of more than three million models, 500,000 data sets. When you think about what additive value that has to Nvidia's long-term strategy, I mean, just looking back over the last couple of years, they moved from GPUs into networking, into software, into full AI systems. I mean, where does this fit in their continued domination of owning the whole AI stack?

**Yael Ossowski** (3:02)
Well, I think they're not owning the stack necessarily, but they are providing the power, the thing that runs it. I think that's really what Hugging Face is about, because it's about connecting the people who are tinkering, who are playing around with these models, and who are more likely to host at home or on their premises. So they're really buying into that.
I think we put so much focus, particularly on the markets, on the big models and what is being released now. But there's so much value up and down the stack just for consumers and small businesses. There are the harnesses, there's all the different apps that are happening. There's all these tools that are being developed that do require more power, more chips, and we see that in the general marketplace, computers are getting more expensive, a lot of desktops are getting a lot more expensive, but it's also making AI more available. So we don't have to necessarily use a subscription from one of the larger Frontier Labs if we want to, we can actually do it on our own. I think that's what Nvidia is putting its money towards, is the idea that people are going to want custom models. They're going to want things that they can build, they can keep their data sovereign, protected. I think that's going to be a big advantage because just imagine what we'll have for schools and hospitals and small businesses. Hugging Face will be the central source of truth. They'll be able to get to that and all of that will be running on Nvidia chips and they need to do that now because all the frontier labs are cutting their own cost and they're starting to make their own chips. This is definitely a nice route for Nvidia to take because so many of those labs are changing course.