AMD Acquires Fei-Fei Li’s World Labs for $8.2 Billion

AMD is acquiring World Labs. Congratulations to you both. Thank you. Lisa, I think a pretty obvious place to start is why? Why is AMD acquiring World Labs? Well, first Ed, thanks for having us. And today's a super exciting day. I can tell you that, um, you know, we are super excited to be acquiring World Labs. I have known Feifei for a long time. I've had tremendous respect, uh, you know, for her as a researcher and as a visionary. And if you think about where I is today, I mean, we are you know, I've said often, Ed, that we're still in the very early innings of AI, and I truly believe that. And what's becoming more and more clear is that, um, the interaction between, you know, the hardware and the software and the systems and the model layer are such that the more you understand end to end, the better system you're going to build. And so that's why we're acquiring World Labs, really is, uh, to have access to the world class talent that Feifei has brought together and really bring it together with AMD's complete capabilities around hardware, software and systems and, um. You know, build the future of AI. I think we should get to the back story. Uh, with both of you. But would you talk a little bit, Lisa, how it came about? You know, the catalyst for doing the actual transaction? Well, as I said, I've had a tremendous respect for faith. I mean, you know, everyone knows just how much she has contributed to I over the last couple of decades and from, you know, our personal relationship, we've always had, uh, that, you know, both mutual respect and trust. But what we've saw is, uh, you know, AMD was an early investor in world loves. So, you know, we saw FFA build when there were only, you know, just a very few researchers and really building to the team uh, with marble. And then with now the release of Atlas. And what we see is, uh, just a tremendous nucleus of capability. Uh, that would be a huge addition to, you know, already our very talented, you know, AMD workforce. So I was thrilled, Thrilled that, uh, you know, Feifei and her founders, uh, were excited about joining AMD and Feifei. Reflect on that. I mean, you and I have spoken very recently. Uh, this, of course, did not come up, but, um, I suppose you know, a good starting point, other than how this came together, is what opportunity did you see for your work in the team at World Labs? Work to be a part of AMD? Yes. So just to echo Lisa, this relationship has been going for many, many years. It started with mutual respect and really a genuine friendship at the personal and professional level. We're very grateful when AMD was an early believer and supporter of World Labs, but really was most fundamental and important and exciting about this merger is the shared vision for the future of AI. World labs has been a frontier, um, model building company. We have been running as fast as we can to build the next generation, uh, multi-model, um, omni model for our spatial and physical intelligence. And as our company grows, so is our ambition. So is our need for compute. And at this point, if we can join forces with AMD, we really are, um, giving ourselves and AMD in the ecosystem an opportunity to accelerate this flywheel between software and hardware development, which is critical critical for today's AI development. Lisa, I think let's get into how that would work in practice. This is an $8.2 billion all stock transaction. But I kind of see two distinct things that AMD is acquiring. There is the insight into the development of models and what the future of AI looks like that helps you inform future generations of hardware. Um, separate from just getting a great team of researchers who can also then offer models. Let's start with the hardware piece. What was your thinking there? Yeah, well, I think if you take a step back and just look at what we're trying to achieve, you know, across this AI ecosystem, I mean, it's a tremendous opportunity. A lot of the focus today has been on, you know, large language models. Uh, you know, really the large foundational model labs, you know, we have a a very deep and special relationship with, you know, folks like OpenAI and anthropic in this. Um, and then you have labs like, what? You know, Feifei has built with World Labs, which is actually, you know, pushing the envelope on sort of the next big thing around world models and physical intelligence. We think that's extremely important as well. So the foundational work that, um, you know, Feifei and her team have done in this area are things that we will continue. Uh, but I also see it as a nucleus for, you know, as you know, for the size and scope of our ambition at AMD, um, having a world class research team and, you know, based team will be the nucleus of our, um, you know, AI models team. Um, I think is absolutely critical when we think about all of the innovation that is to be done, um, over the next decade. Um, I think we want to do it, uh, really with innovation across the entire stack. And so I view this as, you know, both, um, you know, a big, uh, vote of confidence for, you know, just the work that's already been done. But even more importantly, what we will do together, uh, across the entire, you know, AI ecosystem. A lot of people will not have seen this coming. Feifei. I actually think there's a lot of value in saying, where did World Labs get to to date, you know, from founding to, to present day? Um, you've been through some of your own M&A activity to grow the team? Yes. So where do we sit today with World Labs? What is what is the specialism. Yeah. And the work you're focused on. So we're labs is a frontier model company specializing in, uh, spatial and physical intelligence. I was just here talking about our latest model called Atlas, which is a first of its kind, multimodal omni model that, uh, that can generate, uh, beautiful pixels, but also have the, uh, incredible 3D, 4D consistency and allows cameras and agents to interact within these worlds. So we, uh, we were founded in 2000 and, uh, 20th, 2024. Yeah. So we're a little more than two years old. Uh, what's really the founding principle that I is a civilizational technology. We really believe in the positive impact of. I want to push the envelope beyond just the language modeling and going into, uh, spatial and physical, because we're seeing the needs for human creators, design, uh, designers for, uh, education, health care, but also the imminent revolution of robotics. So all those is the foundation of world models. Over the past two years, we have built models and we have built team. We built we started with, you know, the kernel of computer vision, deep learning, uh, large models, as you said. Um, we also have rolled out, uh, a couple of different generations of models by now Marvel first and Atlas. And we also expanded our team with robotics talents. Uh, we recently acquired K'Nex, where, uh, the, the technologists there have built the, a first of its kind, uh, robotic simulation workflow and, uh, cutting edge robotic policy training, um, models. So, so we're lab is a very still a startup, small team, very ambitious. Like Lisa said, we are focusing on building the next generation, uh, technology. And we cannot be more thrilled to join force with AMD because, as I said, I really believe that the future of AI is this whole evolution of AI software. An AI hardware models will gain so much by using the most cutting edge hardware to accelerate the training, the inference, the servicing. But in the meantime, hardware will benefit so much from the insight and the the peek into the future of the workflows and needs of of hardware. Lisa, is it as simple as AMD taking that work and selling models? Um, offering, uh, software as a product? Can I first say like, please, ditto to everything that she just said, because I think she described perfectly, uh, both the, um, tremendous work and the opportunity. But look, um, you know, from our standpoint, AMD continues to be a, you know, platform and solutions company. It's hardware. We have software as an enabling, uh, capability. We have systems where if we're bringing that all together, we work very, very closely with our customers. Who are some of the leading, uh, model labs out there. So I wouldn't necessarily say that, you know, we're we're changing our model to be selling software. I really view this as, uh, you know, first of all, a tremendous foundation that, uh, World Labs has built. Um, I think we've always said we believe in the open ecosystem, and I continue to believe that that's, uh, one of the foundations of the AI innovation race going forward is you're going to have both open as well as proprietary models. Um, AMD has been a contributor there. I think we are going to expand that contribution, uh, with, uh, World Labs as part of the equation. And the ultimate goal is, you know, with World Labs, we build better. I mean, that is AMD's goal. And our our goal is, uh, our ambition is really to define the future of AI, compute and be, you know, the best partner to the largest, uh, consumers out there. And I think I, uh, World Labs will make us much, much better, um, at that, because of the insights and the opportunity to offer, uh, tremendous innovation across the stack. This deal is expected to close by the end of the year, which isn't too far away, subject to regulatory approvals. How is, uh, is Doctor League going to be integrated into the AMD team? What will her role be? Uh, what do you picture on? Well, I pitched her on, you know, joined the most exciting AI company in the world. That's what I pitched. Draw. But, uh, look, Feifei, uh, in addition to the World Labs team, I think Feifei has, uh, a tremendous portfolio and ability to add across AMD. So I'm very happy to say, uh, that she will be, um, AMD's chief scientist. So executive Vice President and chief scientist reporting to me. And, uh, we will, um, you know, really want and welcome, you know, the contributions of, uh, Feifei and the World's labs team across the AMD, uh, roadmap. And if you think about everything that we do, and one of the things I've always said is there's no one size fits all when it comes to AI, right? I mean, we talk a lot about the big cloud compute and what we're doing in terms of CPUs and GPUs. Um, but the physical AI world, The opportunity in robotics. The end to end AI capability across cloud, edge and client. You know, those continue to be places where I believe we're going to grow a lot over the next, you know, three, five, ten years. And those are all places that the world loves. Expertise is it's really in our sweet spot is you need to ask, you know, the World Labs had raised significant money earlier in the year. Um, you have had a lot of momentum. What options did you have? You know, was it always, uh, a path towards joining AMD. What did you consider? I think from the get go, we build a company to really, um, stick to our mission. We're very mission driven, and our mission is really to make a positive impact with the technology that we believe in. So it's not a matter of option as a matter of believe. And having taken this journey with AMD since the beginning, then being our investor as well as, uh, chip provider and then gradually talking to Lisa about our shared vision. It truly is the merging of, uh, of of the shared vision and mission. So it's not about option is about what we can do in the future. And I genuinely believe this is, uh, um, this is going to be, um, one plus one bigger than two. Well, to be fair, you're bigger than one compared to our labs. But it's really going to be, uh, a very, very exciting journey together in that shared vision. Uh, clearly, uh, emphasis on open source and, uh, is a non-negotiable in the direction of travel for I that speak a little bit more about that, Lisa. I mean, why is it important for AMD, from the company's perspective, to advocate for, for an open source market and environment? But also, you know, some of the models that the world labs have worked on have been open, some have been closed to proprietary. And how you balance that going forward. Yeah, I think, uh, the key is what they've said. It is about shared vision for what we're trying to accomplish. And we are trying to be a, um, a very big contributor to the expansion, the acceleration, the innovation in the AI ecosystem. Yes. And to do that, I think you're going to have, um, open models, you're going to have, uh, proprietary models, you're going to have large models, you're going to have small models, uh, they're going to run on different hardware. And, uh, that's been a promise for AMD in our entire high performance computing and AI roadmap going forward. So I think what you're going to see is more emphasis on having optionality. Um, enterprises want optionality. You know, when I talk to enterprise CEOs and CIOs, nobody wants to be locked into either, you know, one hardware ecosystem or one software ecosystem. People want to be able to choose what's best at a given time. And we also want to enable all of these entrepreneurs and, you know, new companies out there that are coming up with great ideas. So you're going to see us continue to lean into open ecosystems as a key way of enabling the entire AI ecosystem. And with that, um, you know, I do think, you know, part of what they've been, I've talked about is how to leverage some of the great work that we're has already done, uh, to actually, you know, accelerate some of the open ecosystem going forward. Please, can I add so the history of modern I cannot be possible without open ecosystem from the open sharing of, uh, neural network algorithms through scientific communities to my own work. Image that right? Well, yes. Image now was the first open source data set that was part of the the driving force for modern I revolution. So I think just to add on to what Lisa said, we need optionality. And it's so important that an open ecosystem help to cross-pollinate Discovery, innovation and entrepreneurship. Business opportunities. This is the meaning and impact of an open ecosystem. I think a lot of people will say FFA and World Labs has focused on what we've called spatial intelligence, uh, a focus on physical AI. But other, um. Areas of the market where a world model is more appropriate than the text and large language model. Do you now go broader in your work within AMD um, and have not? Such a narrow is a strange way to put it, but um, yes, a broader field of work across model types. Yeah. Look, um, we have a lot to look forward to in the future. Lisa and I barely got time to talk about all the plans. Obviously, this is going to accelerate and expand our roadmap, right? We are going to have a lot of conversations. We're definitely going to keep doing some of the most exciting work that we're alive is already doing, but I do envision that we will be expanding and accelerating. Lisa, this is the first opportunity I've had to speak to you, um, in the era of, uh, debate around AI safety. It's hard to make it succinct, but the argument that the capabilities of models are advancing at a pace greater than safety in alignment are advancing to keep up with that development. What is AMD in your position on how best to address all of the concerns that there are out in the world at the moment? Yes. Ed, um, this is an extremely important point, so I'm actually glad you brought it up because it is one that we have to be super clear about as an industry. I mean, there is no optionality in this. This is, uh, we must have great technology and it must be safe. And that's just unequivocal. And when I look at, you know, just history of, you know, how we develop technologies, um, it is one of those things where, you know, safety has to become a first class citizen when you're doing development. And, you know, we've seen that in our own work, when we think about all of the security and the confidentiality things that you have to do around highest, the highest performing processors. I think we're going to have to bring that into the I ecosystem. And I know that that's, you know, front and center in the conversation. Um, I am actually an I optimist. I believe that we will put the right procedures, processes, guardrails. Um, I think an industry wide effort where we actually share best practices. And, um, that's one of the things that we do on the hardware side today, for example, around security. Those are all important aspects of it. But I think we need to be unequivocal to those who are the users of AI. We absolutely are going to deliver great technology, um, that is safe, that is aligned with its purposes, and it requires a tremendous amount of work. So it's not going to be easy, but it is definitely something that, you know, we are committed to as an industry, and we must work together to make that happen. When we spoke last week, Feifei, um, I listed to you regulation third party evaluation, the frontier labs taking responsibility in your your response with all of the above. Um, but how do you see your work within AMD on this issue of AI safety? Also, bearing in mind your long history with this industry and this field of research? Yeah. First of all, I totally agree with Lisa. And like I said, this entire happening is because of the shared, uh, shared vision and shared mission. So safety is shared responsibility. AI is a civilization. No technology. And like all civilization, all technologies, we start with the intention of making human lives and work better. But along the way we learn that we have to use it, develop it, and deploy it responsibly. Because every technology is a double edged sword. So I do think whether it was at War laps or Stanford or now joining AMD. I really want to work with Lisa. Her team and the rest of the ecosystem to see how we can bring out this shared responsibility at individual level, in individual, company level, industry level, as well as societal level. I find this fascinating because I want to go back to, okay, AMD's Acquiring World Labs. That gives deep insight into the development of industry and the models. The capabilities of the models and training runs improve also because the compute its capabilities is improving generations generation. I guess the root of the question is, you know, what responsibility then the compute platform providers take for how that training is conducted safely. And then on the inference side, the running of those models that they are aligned, they behave as we intended them to to do. Well, it's a great point, and I think every one of us in the ecosystem has a responsibility and we are accountable for our our actions so far from us as a hardware platform. We need to make sure that our hardware is secure. That, uh, you know, it does what it's meant to do, that when models run on it, um, that the data is secure. That is something that is very much front and center and center in what we do. We need to work with the model companies in terms of, you know, how the hardware and software interact. Um, I think, you know, Fairfax point up here is, you know, the thought leadership aspect of it is. We must step up as an industry. And, you know, I do think this is an industry led thing in terms of, uh, you know, we have the accountability to put out safe technology, but there are ways for us to work more cohesively across the industry, because this is kind of new. And whenever you have something that's kind of new, you're actually learning. Each party might be learning, uh, at a different pace. And so what I really welcome is the opportunity for more industry collaboration, as well as some third party evaluation to ensure that every part of the stack is working together towards this objective of, you know, phenomenal technology that is safe and aligned to its objectives. Lisa, this is a massive moment for AMD. Actually, it's a moment after some sustained momentum. $1 trillion of market cap, a big moment for the company. You have been very visible alongside the president of the United States, a leader in industry. What happens next? AMD I appreciate that that World Labs is still to close, but what you see is is AMD in the future. Well, first I would like to say I love this industry. I love what we're able to do. You know, I love the comments that Feifei made about, you know, first and foremost, we're here to make technology, you know, better for society, better for people. and that's what I get excited about every day. So the fact that, um, you know, our execution has been strong, that we've continued to really work with the leaders of the industry, uh, both the large companies as well as, you know, we've done, uh, a number of acquisitions that have made a large difference to AMD, and, and World Labs will be, uh, the next in an extremely, uh, strategic and important one. I think this is all part of recognizing that technology is such an important part of society today, whether you're talking about the economy or you're talking about national security, and AMD has a chance to, uh, you know, truly influence and direct where, you know, the future of computing is going. So no better place to be. And Feifei, to finish our conversation with you, what happens next? Um, you know, you and I have discussed this in various mediums. The idea that you're the godmother of I, I'm going to keep it very busy, You know, you have been researching in this space for a long time now. This is a next phase for World Labs as part of AMD that I want that that sort of crystal ball of what you think will the next major development in AI will be something that you stay up late thinking about, that you're writing about, that you pitch Lisa back on saying, I think this is what's going to happen. What happened next is Lisa and I AMD and we're labs will build the future together. And I believe so strongly this future has to be benevolent to the human society and human lives. And this is our shared responsibility and it's what makes us so excited. AMD CEO Lisa Su, World Labs Founder CEO Feifei Li, soon to be chief Scientist of AMD and a part of the models teams at AMD, among other things. Thank you both very much. Thank you. Congratulations. Thank you. Thank you.