Clem Delangue is the co-founder and CEO of Hugging Face, the open-source AI company powering innovation across industries. He's championing transparency and collaboration in a field defined by competition and speed. Delangue embodies the movement to make AI open, responsible, and accessible. He's redefining how value is created in the age of intelligence.
Clem, thank you. Thank you for coming. The clapping is good, and I'm so sorry that I'm up here so much. You're going to get very tired of me very quickly. Clem, you are the only official AI CEO on this stage today. Everybody else is touching it in various ways. So I should just ask you the question directly that I think I maybe to some extent asked Roloff: Are we in an AI bubble?
That's a good question. That's like the not-even-billion-dollar question — the trillion-dollar question these days, right?
I think we're in an LLM bubble, and I think the LLM bubble might be bursting next year. But LLM is just a subset of AI when it comes to applying AI to biology, chemistry, image, audio, video. I think we're at the beginning of it, and we'll see much more in the next few years.
Why do you think we might see? I mean, are you suggesting you think some of these big LLM companies are going to fail, or at least have major problems next year?
I think all the attention, all the focus, or the money is concentrated into this idea that you can build one model through a bunch of compute and that is going to solve all problems for all companies and all people. I think the reality is that you'll see in the next few months, next few years, kind of like a multiplicity of models that are more customized, specialized, that are going to solve different problems. In a way, if you think about a banking customer chatbot, you don't need it to tell you about the meaning of life, right? You can use a smaller, more specialized model that is going to be cheaper, that is going to be faster, that maybe you're going to be able to run on your infrastructure as an enterprise. And I think that is the future of AI.
I'm wondering — it's interesting. Do you feel that? And I'm thinking of — we had Emily Choi from Coinbase up here earlier, and she was talking about when FTX blew up, kind of what that meant for everybody else in the crypto industry. They weren't involved with FTX, but there was residual damage that was kind of done to everybody. If there were one or two major LLM collapses next year, it's not what you guys do, but would it have an impact on Hugging Face? Some sort of negative second-order effects?
Yes, but AI is so big. I think it's so diversified, right? It applies to so many different industries, different companies that I think if part of it is a bit overvalued or a bit exaggerated, it's not going to have a massive impact on the field.
So Hugging Face is best known, or is known, because you're an open-source company, an open-source model. Why is open source important in an AI world where so many of these LLMs are closed?
Yeah, because it distributes power, it distributes capabilities, it distributes understanding. I think in AI you have this major risk of concentrating power in the hands of a few companies. There are natural tendencies for that, and with open source you can actually distribute that power, these capabilities, the wealth too that is created, because open source is a foundation, right? It levels the playing field to give everyone the ability to build themselves these kinds of AI capabilities.
So are you arguing — is open source the best business model or the best moral, philosophical stance?
Hopefully it's both. That would make Hugging Face a good company, right? If we can kind of have a positive impact in the world and at the same time be sustainable. That's kind of what we're seeing right now. We're probably one of the few AI companies that is close to profitability. We raised—
What does close to profitability mean?
It means we're not burning—
Because it means unprofitable, right?
Yes, but we raised $400 million so far. Half of it is still in the bank. We haven't touched yet the previous round that we raised two years ago. So in AI standards, that's called profitability, because the other guys — it's not hundreds of millions that they're spending, it's obviously billions of dollars.
I'm going to get back to open source in a minute, but I want to ask, since you brought up financing. As you said, you guys haven't raised money in a couple of years. There seem to be certain large AI unicorns that seem to literally be raising every couple of days. Anthropic again raised $15 billion today.
I assume your phone is ringing regularly with venture capitalists and private equity folks who want to give you money. Why haven't you raised in two years?
I think we're taking our time. I think we're taking a different approach, as I said, a more capital-efficient approach. We're building a platform for the community that we hope to be there for everyone in 10 years, in 20 years, in 50 years. So we're kind of taking a different approach than most AI companies. I think a lot of people right now are rushing or maybe even panicking and taking really a short-term approach to things. I've been in AI for 15 years now, so I've seen some of the cycles, and so we're learning from that and trying to build a long-term, sustainable, impactful company for the world.
I'm wondering, have you done or do you do secondaries or tenders for employees at— Is that something you do ongoing, or is that when you raise a new round, which has been a couple of years?
Both. We do it regularly. I think it's important for the field today. Talent obviously is a challenge for any AI company, so that's something that we need to do to have retention with your—
Last financing question: Will the next time Hugging Face raises money be in the private markets or the public markets?
It's a good question. I don't know.
I mean, you don't internally, I assume. I mean, do you expect the company to go public in the next 24 months? Is that — I mean, there would be a bit of a game plan. I appreciate things can change both internally and macro, but is that the goal?
Probably not. As I said, I think we're taking the time. The goal is really to optimize for the long run. We don't have any urgency. As I said, we have all the money in the bank from the previous round. So the idea is more to take a long-term approach. However, I'm really excited about the public markets because we have such a big community. We have 10 million AI builders using Hugging Face today, and I know because they tell me every day they would love to be shareholders of Hugging Face too. So that's why I'm excited in the very long run about the public markets.
Let me go back to open source for a minute. One of the co-founders of Databricks, I think yesterday, said something to the effect of we either are or risk the US losing the AI race to China because we are more closed and open — China is more open and closed. Is he right?
So for the past few months, for the first time in our history, the number of downloads from models coming from China is higher than the number of downloads from models coming from the US. Right? You've probably all heard of DeepSeek, of Qwen. All these companies are really sharing their models in open source, and a lot of the startups, academia, companies in the US are now using them, and they're becoming the standard. This is, in my opinion, a big risk for the US, and one of the reasons why we need here in the US to push more for open source. We're seeing OpenAI starting to do that this summer. They released their first open-source model on Hugging Face. Elon Musk released the previous generation of Grok on Hugging Face. So we're starting to see more and more open source in the US, but we need much more because this is really the foundational technology and foundational layer, and I think we want this foundation to be from the US for American companies, from China for a Chinese company, or at least have the choice between the two.
Does it matter though? As you said, like the last version of Grok, you know, the open AI-specific model — does it need to be the newest, newest? I think of this a little bit like when you hear about the US and chip exports, right? Well, we're not letting China have the good chips, we're letting them have last year's chip, last year's used car. Seems like a similar thing, though, right?
Yeah, it's not directly competitive in the way that you use open source for different reasons, and then closed source — it's not always the same use case, the same companies that are using it. But American models need to be competitive with Chinese models, which is not the case now. So that's more the thing that you want to compare it to.
When you talk about those Chinese models, I mean, there's a hardware component to this too, right? In terms of what's being used or what Chinese researchers are using.
Just talk to me a little bit about that, what you're seeing under the hood with that.
The interesting thing is seeing how open source is the foundation, right? And it has an impact on all the other layers of the stack. So what we're seeing right now is, because most of the best open-source models are coming from China, it gives them an ability to influence the decision for hardware, right? You can optimize these models for specific hardware to train and to optimize the inference. So I wouldn't be surprised if in the next few quarters we see this play out with more market share for Chinese hardware based on the success of their open source.
So, long Huawei, short Nvidia.
I mean, I think both will do well. But definitely I think Huawei is going to gain some ground, not only in China but in the other markets, in the US, in Europe obviously, in Africa, in Latin America, thanks to their position in open-source AI.
You've kind of got these insights because of the AI developer community you have. What else is under the hood right now that you're seeing that those of us on the outside looking in wouldn't necessarily see from the top?
LLMs and text AI are not growing as fast, but other modalities are. So AI for biology, for chemistry, audio, video, time series, images are going much, much faster than text. So I really see much more applications of AI in other domains than text in the next few years.
Hugging Face has always, it seems, been referred to as the GitHub of AI. That seems to be this tagline that you guys have always had. And then GitHub started putting some models out there and encroaching a little bit on your turf. Is GitHub still, if not an inspiration, at least kind of a model, or is it a rival?
Both. But I was reading the other day an interesting stat that GitHub released themselves, which was the number of AI-related repositories on GitHub, which was I think three million or a little bit more than that. We have doubled that on Hugging Face. We have over six million AI repositories on Hugging Face. So I think we're doing well, with GitHub being mostly the platform for software and Hugging Face being the platform for AI.
Do you have any interest in possibly encroaching on their turf and becoming also at least partially a platform for software?
Not in the short term. I think we're doing well together and focusing on our use case that obviously is growing, and just staying relevant, staying important, and staying the central platform for AI. It's already a full-time job, so I think we don't need another job.
