September 3, 2026
WHEN: Today, Thursday, September 3, 2026
WHERE: CNBC’s “Squawk Box”
Following is the unofficial transcript of a CNBC exclusive interview with Nvidia Founder & CEO Jensen Huang and Hugging Face CEO Clément Delangue who joined CNBC’s “Squawk Box” (M-F, 6AM-9AM ET) today, Thursday, September 3 to discuss Nvidia’s deal to acquire Hugging Face for $12.9 billion. Following is a link to video on CNBC.com: https://www.cnbc.com/video/2026/09/03/nvidia-ceo-jensen-huang-on-12-point-9b-hugging-face-deal-open-models-matter-greatly-to-our-company.html.
All references must be sourced to CNBC.
BECKY QUICK: Joining us to talk about this deal right now is Nvidia CEO Jensen Huang and Hugging Face CEO Clem Delangue. Gentlemen, welcome to both of you. It's good to see you. And, Jensen, I like your sign. It's the Hugging Face. All right, gentlemen, let's welcome you to the—
CLÉMENT DELANGUE: And I have one in the background too.
QUICK: Yes, I see the one right behind you. This deal has been rumored. There have been people speculating about this for well over a week at this point. What happened? How did you two get together?
DELANGUE: Well, we, oh, go ahead, Jensen.
JENSEN HUANG: No, Clem, I'm going to let you go first.
DELANGUE: Yes. So, I mean, during the summer I think we realized that Hugging Face and open source AI in general was at a turning point and that it needed more resources, more scale, more visibility. So, we went to see Jensen, and we told him, we want to make open source AI big. And he told us, let's do it. So, that's really how it started. And a few weeks later, here we are.
QUICK: Clem, what happened over the summer? Because when you say something happened over the summer, the first thing that pops to my mind was OpenAI's agents kind of breaking in between May and July, you catching them, finding that was there. Was that the catalyst, or was there, were there other things happening that made you think we really need deeper pockets?
DELANGUE: When that happened, what we realized is that we needed open models. Why? Because if you remember, we couldn't defend ourselves with proprietary closed source APIs, so we had to use open models to defend ourselves. So, it did show the importance of open source. And I think in, more generally in the field you're going to see the IPO of Anthropic in a few months. I think there are two paths, where there's a path where proprietary APIs are dominating the field and everyone is kind of like outsourcing their AI to them, and there's a path where open-source AI is available to everyone, and everyone can actually become like an owner, a builder of AI and not just renting it or using it from other people. And so I think it became clear that there was these two paths and that we needed to double down on open-source AI to really distribute the technology as much as we can all over the world.
QUICK: Jensen, you've been recently talking an awful lot about how open AI is a very important path, you and other leaders, but you have been pretty outspoken on this. But the closed AI, the large language models, those companies like OpenAI and Anthropic, they've been some of your biggest customers and continue to be some of your biggest customers. How do you walk this path between closed AI and open AI?
HUANG: In fact, if you talk to the closed model labs, they'll also tell you they are very supportive of open models, and they believe there's a place for open models. Nvidia is the largest AI computing platform in the world. We support clouds, we support neo clouds, we support enterprise on-prem. We're all over the world. And what you're seeing right now is that in the last year or so and really accelerating in the last six months, AI has become useful. And these tokens are, these AI models are generating productive tokens, and they're doing so for both closed models as well as open models. I recommend that people use closed models as much as they can, you know, because it's off the shelf. It's incredibly good. It's advancing very quickly. We use Cloud Code here. We use Codex. We use Perplexity. We use Cursor. Whenever we could use a rented service because it's done so well, we should use it as much as we can, and we do. However, there are so many companies, whether it's fundamental science research companies who have domain expertise and proprietary information that they want to protect. Maybe it's something to do with a regulatory reason or sovereignty reason. They need to have control, ownership and the continued advancement of their own intellectual property. And so AI models are now sufficiently good. The open frontier AI models are sufficiently good, and the agent harness systems that makes it easy for you to build your own AI are now sufficiently good that both sides are now growing incredibly fast. And so you know that cloud service providers represents only half of Nvidia's business, and yet the other half of our business are really largely driven by open models. And so Nvidia's growing in both directions, and we want, our fundamental goal is just to make sure that AI advances as quickly as possible. And it's really, really important right now as the open models are really accelerating, that we make sure that we provide Hugging Face the platform to continue to scale and for the resource for them to scale and extend the open model ecosystem and community.
QUICK: Yes. Jensen, when Clem came to you and said that they wanted to partner up, how did you react? What did you think? And how did you all get to the $12.93 billion price tag? I think $4.5 billion was the valuation back in 2023, when there was a big round of others like Google and Nvidia and Salesforce all kind of stepping in at that point too.
HUANG: Well, Hugging Face, as you know, they have two hundred thousand enterprise customers, 18 million developers around the world, 3 million models that are now, 3 million models that are available on Hugging Face. It goes across languages and world foundation models, and physics models, and chemical models, and biology models. It basically cuts across every single field, robotics models, every single field of AI and when he came to me, you know, and also we're the largest contributor of open models in the world by far. And when he came to me and said that they're thinking about their next chapter, my first thought was, of course, I want to make sure that this, this platform, this company is in great hands and to make sure that it can continue to thrive and grow, because, as it turns out, they've been working on this now for ten years. Their vision is spot on, that the large, there's a large community and industry of people who really need open models. So, their vision was spot on. And at a time when open models are accelerating this is really a very, very delicate time, and we want to make sure that it has all the support necessary. And so when he came to me, my first thought was, oh, no, you know, this, everything is going so well, and I really wish that they would continue on a standalone basis. But there are other bidders, and $12.9 billion is what it took to close the deal, and it's worth every single penny. And, you know, as you know, Nvidia as I mentioned, Nvidia is the largest AI computing platform in the world. Open models matters greatly to our company, which is the reason why we invest so much ourselves. There are so many industries beyond languages that are, that benefit from open models, and we are completely committed to it. We want, and we're so dedicated to what we're ready. It's really important to us that it lands in a good place. And Nvidia is a great home for them.
QUICK: Jensen, Andrew asked the question before the break when we came up here. Just to talk a little bit about why you did this, was it, and it sounds like you're confirming that you were concerned that if it wound up in different hands, it might break the system that was there. You paid $12.9 billion to keep it out of others' hands?
HUNAG: No. Well, we paid $12.9 billion because, one, that's what it's worth, and, two, open models means so much to us. And if it has, if Hugging Face was looking for the next chapter and a home for their company, Nvidia should be it. And Clem was so gracious to come to me, and he was, he, you know, told me about his home and about the company he's built, and it's time for him to consider a next chapter, and he's got other interests and other companies who are interested in the company. And so the fact that this is so important to us, such a large growth driver of our company, and together, we can scale open community even faster than they're able to do today, which will be great for Nvidia. And so this is great for Hugging Face, this is great for the open model community, it's great for Nvidia.
QUICK: Clem, what what are your other interests, and are you going to stay with Hugging Face?
DELANGUE: Yes, the three founders and the whole team is going to to join Nvidia. Jensen was in our whole company team meeting just a few minutes ago, and you know you couldn't imagine how everyone was excited about that. So the goal really is to to join Nvidia, to continue to run independently neutral platform within within the Nvidia team, and everyone is super excited to get to this next level and and reach even more impact. As as Jensen mentioned, we have 18 million AI builders using us. I think in the next few years we should aim at getting 100 million AI builders on Hugging Face, and really empower anyone in the world to not just be an AI user, but also be an AI builder themselves and own their intelligence. I think if we achieve that, we will have kind of like a massive impact in the world.
QUICK: There had been reports that there could be as much as a billion-dollar retention plan put in place to keep the talent at Hugging Face. Is that the case?
HUANG: Yes.
DELANGUE: So we not—
HUANG: Yes. Hugging Face is all about the talent.
DELANGUE: — Not commenting that.
HUANG: No. There's there's no no reason not to. It's, the talent is everything. The talent is absolutely everything, and so Hugging Face, the the people of Hugging Face and the employees of Hugging Face have this is their this is their life's work, and they've done such an incredible job and such an incredible service to the world and the industry to create this platform where all of the community can come together to advance AI together. And when people work in the open, it is safer. It is more secure. The technology advances more quickly. That's the reason why Nvidia contributes so much to Hugging Face, and and we're so committed to Hugging Face. And this opportunity allows us to bring the two companies together and move even faster and scale the open platform even more broadly and even more quickly, and so this is this is a huge win for for Hugging Face. This is a huge win for the community, huge win for us. I'm just I was just happy that that when he was you know of course I'm always happy when companies thrive and and they continue to be a great partner of ours. However, if he's gonna if he's thinking about the next chapter of Hugging Face, it absolutely should be at Nvidia.
QUICK: Clem, Jensen mentioned that there were other bidders involved. Did you go to Jensen first? Who were the other bidders?
DELANGUE: Yeah, I went to to Jensen first because, as as I said, I think Nvidia is the perfect home for Hugging Face. Right? The alignment is obvious. I think when Jensen led this letter in support of open source during the summer, it was it was another example of of the alignment, of the mission alignment, of the culture alignment. I think we're trying to achieve very similar missions in in the world, and and so Nvidia was definitely the best home for Hugging Face.
QUICK: The the letter that went out that that was before your discussion started taking place?
DELANGUE: Yes, yes, it was. I mean, Nvidia has always been a great partner of ours over the years, as Jensen mentioned, the the largest American contributor to to the platform with open models, but also open data sets. So obviously we've known each other for for quite a while, but the the discussions went quite quite fast because it was quite quite obvious, I think, to to to everyone that this was kind of like the best fit, and that was giving us a chance to to to really go further, bigger, get 200 million AI builders, as I was mentioning, and and really make sure that open source AI is is a great force in in the field and in the world.
QUICK: Who who were the other bidders? Were they some of your other investors?
DELANGUE: I don't think it matters matters too much. Again, I think Nvidia was was the best home for us. We, we’re super happy. As as Jensen mentioned, we started 10 years ago. We we hope that in the next 10 years with with Nvidia, we can we can achieve great great things.
HUANG: Oh Becky, it doesn't matter who the other bidders were. It only matters who wins.
QUICK: Yeah, we're interested. We want to know who is in the running for this stuff too. Jensen, you put out in this blog post this morning. You said Nvidia compute will not be required to build on or deploy through Hugging Face. You've mentioned already it's pretty important that this stays open. I guess maybe there's the question of why not say hey we're just going to make this Nvidia powered only and and and really kind of push our advantage on that. Why not do that?
HUANG: That's just not the open ethos. You know, when you think about what NVIDIA does, we contribute more to open models and open data sets than anybody in the world. Just take a step back and just think. You know, there are a lot of people contributing to open source and open models and open data and open scripts and things like that, and and we contribute more than anybody else. And we contribute the model, the data set that we use to train it, the scripts necessary so that you can reproduce our model as exactly as we have it, and and we contribute all of that to the world and runs on everybody else's platform. And so, this is the open ethos. If you really believe in advancing the technology quickly, safely, securely, and you want this to benefit every single industry, every single company, including companies that you might compete with, you know that's the open model, that's the open model ethos. And so, we're completely bought into it. So much of the CUDA stack is open. The original large model training training script called Megatron Core. That Megatron Core was open sourced by us, and it enabled everybody to scale up their training of AI models. The inference platform called Dynamo completely open enabled everybody to inference at very high scale, and so, so we we really believe in advancing this technology for the benefit of of society, and of course, accelerating the industry. Now, of course, it's great for our business because we're also the world's largest AI computing platform for training and for inference, and therefore, you know when when AI advances, whether it's in the cloud, in a in an on-prem data center like a drug discovery company, an incredible supercomputer center that Lilly built, or all the way out at the edge in a self-driving car or a robotic system, Nvidia benefits whenever AI advances, and so this is great for the industry. It's great for companies. It's great for countries, and it's great for us.
QUICK: Clem, you said you wanted to do this because, look, there are a lot of concerns about what's happening with with AI and how quickly things are moving. There was a report just yesterday that looked at the new Astra model that OpenAI is bringing up, and real concerns from security experts saying there's opaque recurrence that is happening with this model that really has experts rattled about what this means. Can you stop some of these agents? They're doing things that hadn't been expected before. The chain of thought will be more difficult to monitor. What's your concern on this level, and what do you think you can do about it?
DELANGUE: Well, I think what we learn is that locking everything behind closed doors doesn't work, right? Because that that's what's been happening. Like if if you look at the models that attacked us during the cyberattack, it was actually an unreleased model behind closed doors. So this approach doesn't work. What we feel like works is distributing this technology, especially with with open weights, so that you leave all the playing fields, and you give everyone kind of like the ability to defend themselves, right? The real risk is when you have massive asymmetry of of power and capabilities between two players, because that's when the attacks can be the most dangerous. But when you leave all the playing fields and you give kind of everyone the same tools, the system continues to work, right? And if the attacks become more sophisticated, the defense will get more sophisticated. And so, at the end of the day, this is a system that that works. And in in this system, obviously, open models, open data sets, distributing the technology as much as possible is the most important thing.
QUICK: Clem, you, oh go ahead Jensen.
HUANG: And Becky, this is such an important point. Becky, if I could jump in for a second, this is such an important point. If you recall, the transition from a computer being a personal computer to the computer being an infrastructure for the world is when the world needed open-source Linux. The fact that you can build open-source Linux in plain sight, Kubernetes in plain sight, it's the safest, it's the most secure way of doing something. In the world of cybersecurity, everybody knows asymmetry defense is ultimately your best defense, and when I say asymmetric capability, there are way more people who are protecting than there are people who are attacking. And so, the benefit of having the community come together with open models, so that they can collaborate all transparently with each other gives the defenders an asymmetric advantage over the the attackers.
QUICK: Jensen, are you are you concerned about what OpenAI is building then?
HUANG: Hang on a second. Ask. Hey, can somebody take? Yeah.
QUICK: Clem, let me ask you. We'll get back to Jensen with that in a second. I think he must have the president calling, Clem, or maybe it dropped off because we hit 8:30. Clem, let me ask you. You have said that OpenAI, or that that you think China is winning on the OpenAI front. Do you think that's still the case?
DELANGUE: Well, what we're seeing on the platform, and that's kind of like also a beauty of what we're announcing today is that with open-source AI, it can come from anywhere around the world, right? I'm obviously I'm French, as you can hear from my accent. We have a big part of the team in in France, and and we have big and important contributors to open-source AI there, right? With with Mistral, in in Canada we've we have Cohere, in in China there there are a lot of important contributors to to the to the open-source AI field. So, and and one of the things that that we want to focus on with with Nvidia is to to continue to do that, push more sovereign AI, so give any any country the ability also to own their own AI and and build their own AI, and we hope it's going to accelerate in the next few years.
QUICK: Great. Hey, I think Jensen can hear us again. Jensen, my question was based on what you just said about why things need to be open, are you concerned about the models from the closed, the closed models that are here, OpenAI and Anthropic? Are you worried about what they're doing at this point?
HUANG: Well, I'm not exactly sure what they're doing, but but there's a place for closed models and there's place for for open models. The fact of the matter is, is defenders need access to frontier capability open models, and that that access is really enables the defenders to be able to innovate and build their their domain-specific cybersecurity expert harnesses and turn them into great AIs. And you saw the other day the partnership that we announced with CrowdStrike using Nvidia's open models, Nemotron, they were able to create a frontier AI cybersecurity AI defender, and that capability, I think, you're going to see a lot more of. I think that the bottom line is that that the way to think about that is you want to put in the hands of defenders all of the best technology you can as quickly as you can, and there's just no better way of putting great technology in the hands of defenders and the world to defend themselves than open technology, open model, open source, and which is the reason why we're doing all of this. The collaboration with CrowdStrike is the first step. You're going to see a lot more collaborations around open models in the world of cybersecurity because that's what it takes to defend the world. You can't rely on just one or two companies to do so. You have to enable the entire crowd, if you will, the asymmetry of defenders who want to defend themselves to be able to enable enable that to happen.
QUICK: Okay, Jensen, Clem, I want to thank you both. Congratulations on the deal, and we appreciate your time this morning.
HUANG: Thank you, Becky. Congratulations, Clem.
DELANGUE: Thank you very much. Thank you, thank you, everyone. Thank you, Jensen.
HUANG: Thank you.
For more information contact:
Stephanie Hirlemann
CNBC
