Thank you for joining me here on CNBC and for Fort Knox. We're at CES and you just got off stage and out of that analyst conversation a bit ago. I want to talk about the CES announcements, but first I haven't had a chance to talk to you since the Grok licensing announcement. Lot of talk about inferencing here. Why was that an important deal for you to do given all the advantages you already had in inference?
Well, we don't compete. You're absolutely right that Eva's platform is unique in the sense that we're great at training, great at post-training, and now with reinforcement learning and the scale of computing necessary for post-training, and because of the investments we made with NVLink and all the technologies we created, we're incredibly good at inference. We're the only architecture that really spans all three. We've known about Grok for some time, and the thing I really like about Grok is, although we never saw ourselves competing, they built a very specialized version of a chip that uses integrated SRAM. It can't scale up to the scale we could scale. It can't be used for training and pre- and post-training. But for low latency, high token rate generation, as long as it fits in the SRAM, it's really quite interesting. So I really love their low latency focus. Everything about their programming model, their architecture, their system architecture was designed for low latency. Now, the extreme capability for low latency token generation is also one of the reasons why they had a very hard time addressing the mainstream part of AI factories. But in combination with us, they don't have to address that. They no longer have to address the segment of the market where the scale is. And maybe with us we can go explore the edges, the fringes of where AI factories could be someday. So I really like the team. I really like Jonathan. Really talented team. I love their conviction. I think their architecture has unique strengths. For the mainstream part of the market, some real challenges, but together those aren't their problems anymore. They could really focus on their strengths.
Something that I've noticed that is different to me is how deep you go on the details of other people's businesses. So I was watching the video you did with the poodle talking about runway, and I know Chris a bit. We started talking about three years ago, and you were talking about his business like it was your business. There's so many companies. What's your process in understanding the landscape, the ecosystem that way?
John, there's only one way to do it. You just can't sleep. But Nvidia is a platform company. We sit at the center of this technology shift, this platform shift from classical computing to artificial intelligence. We're fortunate that we've been working on it for a long time, and the way we solved the problem turned out to have been quite useful. Because Nvidia has always been a systems company, we're full stack, meaning we build chipsets, we develop software, we develop algorithms, and we even create systems. But we never wanted to be a vertically integrated company. We wanted to be a platform company where we could work with every ecosystem partner. So enjoying seeing other people's success that we work with is part of our DNA. I'm always thinking about how can I be helpful to this CEO? How can we do things together such that when they grow, they take us with them? What are the things that I could do with this CEO, with this company, that are somewhat unique and we invent something together? I'm constantly trying to figure out how can we be of service to someone? How can we be helpful to someone? How can we partner with someone? That's a very different psychology than a lot of CEOs who are always trying to disrupt someone or take their market share. Not one nanosecond does Nvidia think about taking somebody else's market share or disrupting anybody. We're always trying to think about creating something, and that's just the nature of our company. It's fitting that we're at the center of the ecosystem. We want a lot of friends, we have a lot of friends, we need a lot of friends. I think the DNA of our company that wants other people to succeed so badly is one of the reasons why we learn a lot about them.
So that leads nicely into open, which is one of the early themes that you talked about, and the idea that it's going to take open platforms, open approaches to truly grow AI's potential in the next phase, in the next era. You've been contributing a bunch to various open efforts, and I want to get to Apomeo in a bit on the physical AI side. But percentage-wise or portion-wise, how much of Nvidia's time gets spent on open, and how much of what you saw happen 20 plus years ago in the most recent Linux open source era influenced that?
We have several thousand people working on the open stack, from the infrastructure layer, which is the runtime of the AI factory, the operating system of the AI factory, to the model layer like Neotron, Groot, Alpamo, Earth 2, all of our Clara Biology models. All together, that entire section of our company is probably several thousand people, if I venture to guess something along the lines of 5,000. We are the largest contributor to open source for AI in the world. We build frontier models for physical AI like Cosmos, for autonomous vehicles like Alpha Mayo, humanoid robotics like Groot, protein models like OpenFold 3 and EVO 2 and Laortina, all the R2 models. We're building frontier models across a lot of different domains. We open source it, not just the model and the model weights, but we open source the data. All of the data we use are all permissible for downloading. We make sure that the data we train our models with are permissible. They're so permissible I could even give the data away. All of the scripts we use and all the runtimes, everything's open. It's because of that that we're literally integrated into every company in the world. We're the only AI company in the world working with every AI company in the world. We run every model from OpenAI to this year we announced Anthropic. I'm very excited about that. Of course, we have the benefit of working with Elon and xAI. We've always been working with Google on Gemini. Nvidia is the only company in the world that runs across every domain of science, every AI company, every AI model. The only way to do that is to do it openly. If we were to be proprietary in some way, closed in some way, then we would have the benefit of monetizing the entire layer and the entire stack, but it prevents our ability to be open. I really love that our company has the ability to work with everybody.
Let's talk about physical AI starting with robotics. It's been...
Oh, excuse me, John. If you take all of the open source models in the world, all the open models in the world, and we just glom that together as one, the world's open models is probably the second largest model. The first is OpenAI, but the second largest is clearly open source. That tells you something. The leader in open source is actually in a lot of ways already the second largest model. It's just that there's a whole bunch of them, but that's okay. We don't care that it's only one. But I think this open model strategy really works. It's a big investment for us, and we have to attract some of the world's frontier AI model makers. Nvidia is a great place to work, and we have amazing AI researchers. This is an area we contribute deeply to the world.
And I guess one might argue it keeps the industry safer from getting captured by any one particular proprietary point of view.
The opposite way of thinking about that is it enables innovation at almost every scale. It's because of these open models that we're able to work with every enterprise company in the world. From the work we do with Cadence and Synopsys and Siemens to the ones we do with CrowdStrike and Palantir and ServiceNow and on and on. These are amazing companies we have the benefit of working with, and it's all because the work that we do is open. In a lot of ways, the strategy is not monetized, but it's monetized in a way that our platform has adopted.
Right. When the sun shines, it shines everywhere and things grow. Robotics has been a dream part of CES for a long time, but it's also been elusive. We had iRobot kind of not go in a great direction. Granted, we do have autonomous vehicles now, and those are really robots, I guess, as we're talking about physical AI. What makes this moment different for robotics in particular? And maybe give me humanoid because we tend to get very excited about it, but I wonder too excited.
Excellent. Timing is everything. We've been thinking about this area for a long time. We're in for that moment. As in many of the things we do, whether it's digital biology and turning drug discovery from a discovery process to an engineering process, from a scientific process to a science and engineering process, to self-driving cars, to even the work we do in computer graphics, ray tracing is now completely done in real time. It took us 30 years to do it in real time. You have to pursue something for a long period of time looking for that moment where that enabling technology is discovered. In the case of human robotics, let me first say that a computer doesn't know and doesn't care what kind of tokens it's generating. It could be generating a language token, a video token, a steering wheel activation token, or a finger articulation grasp token. The computer in the final analysis is just a bunch of numbers. The moment I saw generative video, us texting into a prompt, two people sitting having a conversation, the guy wearing a jacket reaches out, picks up a cup of water, and drinks it. I could describe that into a generative AI model today, and you and I both know that Chris, for example, with Runway, I could give him that prompt and he will generate an amazing video. That video has a person reaching out and picking up a cup. Why is that model different than a generative model for a human or robot picking up a cup? The moment I saw that working that well, then the rest of it is still a whole bunch of research, a whole bunch of technology, but you could tell that the enabling technology is just around the corner.
So what you can see happening in one mode, right, in video, it's just innovation happens.
Yeah. That's how innovation has happened. I'm constantly doing this. If that's possible and we break it down to first principles, why can't I do that here? If you could do that, when you see the art of the possible in one domain, as long as you could reason back to first principles, you could reapply that in another domain. Most innovators think in this way.
Okay. So I was talking to Chris Valenzuela over at Runway. He was one of your early partners on Vera Rubin. I'm so proud of them. The next version of the platform which you're delivering to partners second half of 2026. Right now, all six chips have kind of been tested out. Chris will never say this, he'll probably never tell anybody this, but Chris has already had access to Vera Rubin.
Yeah. Chris has already generated a video on Vera Rubin.
That's where I was going. Is that right? Okay. He was telling me that it only took him a day to switch things around from what he was doing. I don't think he said he was on Blackwell. He was on Hopper to Vera Rubin. In the past it probably would have taken weeks to do that kind of change. What kind of a platform impact? What kind of a customer joy impact do you think that upgradability in particular?
