Jensen Huang: NVIDIA - The $4 Trillion Company & the AI Revolution | Lex Fridman Podcast #494

Yeah, thanks for that question. So first of all, the reason why extreme co-design is necessary is because the problem no longer fits inside one computer to be accelerated by one GPU. The problem that you're trying to solve is you would like to go faster than the number of computers that you add. So you added, you know, 10,000 computers, but you would like it to go a million times faster. Then all of a sudden you have to take the algorithm, you have to break up the algorithm, you have to refactor it, you have to shard the pipeline, you have to shard the data, you have to shard the model. Now all of a sudden when you distribute the problem this way, not just scaling up the problem, but you're distributing the problem, then everything gets in the way. This is the Amdahl's Law problem where the amount of speed up you have for something depends on how much of the total workload it is. And so if computation represents 50% of the problem, and I sped up computation infinitely like a million times, you know, I only sped up the total workload by a factor of two. Now all of a sudden, not only do you have to distribute the computation, you have to, you know, shard the pipeline somehow. You also have to solve the networking problem because you've got all of these computers are all connected together. And so distributed computing at the scale that we do, the CPU is a problem, the GPU is a problem, the networking is a problem, the switching is a problem. And distributing the workload across all these computers is a problem. It's just a massively complex computer science problem. And so we just gotta bring every technology to bear. Otherwise, we scale up linearly or we scale up based on the capabilities of Moore's Law, which has largely slowed because Dennard scaling has slowed.

That's why my staff is so large.

Yeah. There's the first question, which is: what is extreme co-design? You're, you, we're optimizing across the entire stack of software from architectures to chips, to systems, to system software, to the algorithms, to the applications. That's one layer. The second thing that you and I just talked about is goes beyond CPUs and GPUs and networking chips and scale up switches and scale out switches. And then of course, you gotta include power and cooling and all of that because, you know, all these computers are extremely, extremely power hungry. They do a lot of work and they're very energy efficient, but they in aggregate still consume a lot of power. And so that's one. The first question is, what is it? The second question is, why is it, and we just spoke about the reason, you know you want to distribute the workload so that you can exceed the benefit of just increasing the number of computers. And then the third question is, how is it, how do you do it? And that's the, that's kind of the miracle of this company. You know, when you're designing a computer, you have to have operating system of computers. When you're designing a company, you should first think about what is it that you want the company to produce. You know, I see a lot of companies organization charts, and they all look the same. Hamburger organization charts, soft organization charts, and car company organization charts. They all look the same. And it doesn't make any sense to me. You know, the goal of a company is to be the company is to be the machinery, the mechanism, the system that produces the output. And that output is the product that we like to create. It is also designed, the architecture of the company should reflect the environment by which it exists. It almost indirectly says what you should do with the organization. My direct staff is 60 people. You know, I don't have one-on-ones with 'em because it's impossible. You can't have, you can't have 60 people on your staff if you're, you know, gonna get work done and—

More, yeah.

Almost all of them. There's experts in memory, there's experts in CPUs, there's experts in optical. All, all—

Yeah, GPUs and— Architecture, algorithms, design, um—

And no conversation is ever one person. That's why I don't do one-on-ones. We present a problem and all of us attack it. You know, because we're doing extreme co-design. And literally, the company is doing extreme co-design all the time.

Yeah, exactly.

Exactly.

Exactly. And whoever wants to tune out, tune out. You know what I'm saying? And the reason for that is because the people who are on the staff, they know when to pay attention. There's supposed... You know, it's something they could have contributed to, they didn't contribute to, 'I'm going to call them out.' You know? And so, 'Hey, come on, let's get in here.'

Uh, I could, I c- you, you could- I could reason through what just systematically. We started out as a, as an accelerator company. But the problem with accelerators is that the application domain's too narrow. It has the benefit of being incredibly optimized for the job. You know, any specialist has that benefit. The problem with intense specialization is that, of course, your market reach is narrower, but that's, that's even fine. The problem is, the market size also dictates your R&D capacity. And your R&D capacity ultimately dictates the influence and impact that you can possibly have in computing. And so, when we first started out as an accelerator, very specific accelerator, we always knew that had- that was going to be our first step. We had to find a way to become accelerated computing. But the problem is, when you become a computing company, it's too general purpose and it takes away from your specialization. The tur- I connected two words that are actually have fundamental tension. The better computing company we become, the worse we became as a specialist. The more of a specialist, the less capacity we have to do overall computing. And so, that... And I connected those two words together on purpose, that the company has to find that really narrow path, step by step by step, to expand our aperture of computing, but not give up on the most important specialization that we had. Okay, so the first step that we took beyond acceleration was, we invented a programmable pixel shader. So, that was the first step towards programmability. You know, it was our first journey towards moving into the world of computing. The second thing that we did was we created we put FP32 into our shaders. That FP32 step, IEEE-compatible FP32, was a huge step in the direction of computing. It was the reason why all of the people who were working on, on stream processors and, you know, other types of data flow processors discovered us. And they said, 'Hey, all of a sudden, you know, we might be able to use this GPU that's incredibly computationally intensive, and it's now, you know, compliant with IEEE.' I can take my software that I was writing, you know, previously on CPUs, and I can, you know, see about, you know, using the GPU for that. And which led us to create, put C on top of FP32, what's called, we call Cg. The Cg path took us to eventually CUDA. CUDA, step by step by step. We... Well, putting CUDA on GeForce, that was a strategic decision that was very, very hard to do, because it cost the company enormous amounts of our profits, and we couldn't afford it at the time. But we did it anyways because we wanted to be a computing company. A computing company has a computing architecture. A computing architecture has to be compatible across all of the chips that we build.

Yeah, excellent. That was, that was the first... I would say that that was the first strategic decision that is as close to an existential threat.

Thank you.

Yeah, it turned out to have been a good decision. I think the... So, here's the way it went. So, we invented this thing called CUDA, and it expanded the aperture of applications that, that we can accelerate with our accelerator. The question is, how do we, how do we attract developers to CUDA? Because a computing platform is all about developers. And developers don't come to a computing platform just because, you know, it could perform something interesting. They come to a computing platform because the install base is large. Because a developer, like anybody else, wants to develop software that reaches a lot of people. So, the install base is, in fact, the single most important part of an architecture. The architecture could attract enormous amounts of criticism. For example, no architecture has ever attracted more criticism than the x86.... you know, as a less than, less than elegant architecture, but yet it is the defining architecture of today. It gives you an example that in fact so many RISC architectures which were beautifully architected, incredibly well-designed by some of the brightest computer scientists in the world, largely failed. And so I've given you two examples where one is, you know, one is elegant, the other one's barely aesthetic, and so yet x86 survived and the reason for—

Install base defines an architecture. Not... Everything else is secondary, okay? And so there were other architectures at the time. CUDA came out, OpenCL was here. There were... You know, there's several other competing architectures. But the thing that... The decision that we made that was good was we said, 'Hey, look, ultimately it's about install base and what is the best way we could get a new computing architecture into the world?' By that timeframe, GeForce had become successful. We were already selling millions and millions of GeForce GPUs a year, and we said, 'You know, we ought to put CUDA on GeForce and put it into every single PC whether customers use it or not, and use it as a starting point of cultivating our install base.' Meanwhile, we'll go and attract developers, and we went to universities and wrote books and taught classes and put CUDA everywhere. And eventually people discover... And at the time, the PC was the primary computing vehicle. There was no cloud, and we could put a supercomputer in the hands of every researcher in school, every scientist, you know, every engineering school, every... or every student in school, and eventually something amazing will happen. Well, the problem was CUDA increased our cost of that GPU, which is a consumer product, so tremendously, it completely consumed all of the company's gross profit dollars. And so at the time, the company was probably, you know, worth, I don't know, at the time, eight... Was it like $8 billion or something? Like six, $7 billion or something like that. After we launched CUDA, I recognized that it was going to add so much cost, but it was something we believed in. You know, our market cap went down to like one and a half billion dollars. And so we were down, we were down there for a while and we clawed our way way back slowly, but we carried CUDA on GeForce. I always say that NVIDIA is the house that GeForce built, because it was GeForce that took CUDA out to everybody. Researchers, scientists, they discovered CUDA on GeForce because they were all, you know... Many of 'em were gamers. Many of them built their own PCs anyways. In a university lab, many of them built clusters themselves, you know, using PC components. And, and so that, you know, that's kind of how we got going.

That was also another great, great observation. Yeah.

Well I had to make it clear to the board what we're trying to do, and the management team knew our gross margins were gonna get crushed. So you could imagine a world where GeForce would carry the burden of CUDA and none of the gamers would appreciate it and none of the gamers would pay for it. You know, they only pay certain price and it doesn't matter what your cost is. And so the... You know, we, we increased our cost by 50% and that con- consumed... And we were a 35% gross margin company, and so it, it was a... It was quite a difficult decision to make. But you could imagine that someday this would go into workstations and it would go into supercomputers and, and in those segments, maybe we can capture more margin. so you could, you could reason your way into being able to afford this, But it still took... It took a decade.

Well, first of all I'm informed by a lot of curiosity. At some point, there's a reasoning system that, that convinces me, so clearly this outcome will happen. That this will happen. And so I believe it in my mind, and when I believe it in my mind, you know, you know how it is. You manifest a future and that future is so convincing, there's no way it won't happen. There's a lot of suffering in between, but you've gotta believe what you believe.

Yeah. And you, you reason about how to get there. You reason about why it, it must exist. And, you know, I reason... We all reason it here. The management team would reason about it. All the people that I... We spend a lot of time reasoning about it. The thing, the thing that... The next part of it is probably a skill thing, which is, you know, oftentimes in leadership the leadership stays quiet or they learn about something, and then they do some manifesto, and it's a brand-new year, and somehow at the end of the year, next year, we're gonna have a brand-new plan. Big huge layoff this way, big huge organization change this way, new mission statement... brand new logos you know, that kind of stuff. Um, we've just never, I never do things that way. When I learn about something and it's starting to influence how I think, I'll make it very clear to everybody near me that, you know, this, this is interesting. This is going to make a difference. This is going to impact that. And I reason about things step by step by step. Oftentimes, I've already made up my mind, but I'll take every possible opportunity, external information, new insights, new discoveries, New engineering, you know, revelations, new milestones developed, I'll take those opportunities and I'll use it to shape everybody else's belief system. And I'm doing that literally every single day. I'm doing that with my board, I'm doing that with my management team, I'm doing that with my employees. I'm trying to shape their belief systems such that when I come the day I say, 'Hey, let's buy Mellanox,' it's completely obvious to everybody that we absolutely should. On the day that I said, 'Hey guys, let's go all in on deep learning,' and let me tell you why. I've already been laying down the bricks to different organizations inside the company. Every organization and everybody, many of the people might have heard everything. Most of the company hears, of course, pieces of it. And on the day that I announce it, everybody's kind of bought in to many pieces of it. And in a lot of ways, I like to announce these things, and I imagine, that the employees are kind of saying, 'You know, Jensen, what took you so long?' And in fact, I've been shaping their belief system for some time, and therefore leadership. Sometimes it looks like you're leading from behind, but you've been shaping their, you know, to the point where on the day that I declared it, 100% buy-in. But that's what you want. You want to bring everybody along. You know, otherwise, we announce something about deep learning and everybody goes, 'What are you talking about?' You know, you announce something about let's go all in on this thing, and your management team, your board, your employees, your customers, they're kind of like, 'Where's this coming from? You know, this is insane.' And so, so GTC effect, if you go back in time, you look at the keynotes, I'm also shaping the belief system of my partners in the industry and I'm using that to shape, you know, the belief system of my own employees. And so by the time that I announce something, like for example, we just announced Grok. We've been late... I've been talking about the stepping stones for two and a half years. You just go back and go, 'Oh my gosh, they've been talking about it for two and a half years.' And so I've been laying the foundation step by step by step, so when the time comes you announce it, everybody's saying, 'You know, what took you so long?'

We don't build computers. We actually don't build clouds. We don't... As it turns out, we're a computing platform company. And so nobody can buy anything from us. That's the weird thing. You know, we vertically design, vertically integrate to design and optimize, but then we open up the entire platform at every single layer to be integrated into other companies' products and services and clouds and supercomputers and OEM computers and so the amazing thing is, I can't do what I do without having convinced them first. And so most of GTC is about manifesting a future that by the time that we... My product is ready, they're going, 'What took you so long?'

Yeah, yeah. Yeah, we have more scaling laws now.

Well, we can go back and reflect on what people thought were blockers. So in the beginning, we were the first... The pre-training scaling law. You know, people thought well rightfully so, that the amount of data that we have, high-quality data that we have will limit the intelligence that we achieve. And that scaling law was an important, very important scaling law. The larger the model, the correspondingly more data results in a better... With a results in a smarter AI. And so that was pre-training. And Ilya Sutskever, Ilya said, 'We're out of data,' or something like that. 'Pre-training is over,' or something like that. The industry panicked, you know, that this is the end of AI. And of course, of course that, that's obviously not true. We're gonna keep on scaling the amount of data that we have to, to train with. A lot of that data is probably gonna be synthetic, and that also confused people, you know? And what people don't realize is they've kind of forgotten that most of the data that, that we are training that we teach each other with, inform each other with, is synthetic. You know, I... It's synthetic because it didn't come out of nature. You created it. I'm consuming it. I modify it, augment it, I regenerate it, somebody else consumes it. And so, so we've now reached a level where AI is able to take ground truth, augment it, enhance it, synthetically generate an enormous amount of data. And that part of post-training continues to scale, and so the amount of data that we could use that is human generated will be smaller, and smaller, and smaller. The amount of data that we use to train model is going to continue to scale to the point where we're no longer limited... Training is no longer limited by... Data is now limited by compute. And the reason for that is most of the data is synthetic. Then the next phase is test time, and I still remember people telling me that, 'Inference? Oh, yeah, that's easy. Pre-training, that's hard.' These are giant systems that people are talking about. Inference must be easy. And so inference chips are gonna be little tiny chips, and you know, they're not, they're not like NVIDIA's chips. Oh, those are gonna be complicated and expensive, and, you know, we could make... And this is— in the future, inference is gonna be the biggest market, and it's gonna be easy, and we're gonna commoditize it. You know, everybody can build their own chips. And, and that was always illogical to me because inference is thinking, and I think thinking is hard. Thinking is way harder than reading. You know, pre-training is just memorization and generalization, you know, and looking for patterns in relationships. You're reading and reading, versus thinking, reasoning, solving problems, taking unexplored experiences, new experiences, and breaking it down into... Decomposing it into, you know, solvable pieces that we then go off, either through first principle reasoning, or, you know, through previous examples, prior experiences. You know, or just uh, exploration and search and, you know, trying different things. And that whole process of, of test time scaling, Inference, is really about thinking. And it's about reasoning, it's about planning, it's about search, it's about... And so how could that possibly be compute light? And we were absolutely right about that. You know, so test time scaling is intensely compute intensive. Then the question is, okay, now we're at inference and we're at test time scaling, what's beyond that? Well, obviously, we have now created, you know, one agentic person, and that one agentic person has a large language model that we've now now, you know, developed. But during test time, that agentic system goes off and does research and bangs on databases, and it goes out and, you know, uses tools, and one of the most important things it does is spins off and spawns off a whole bunch of sub-agents. Which means we're now creating large teams. It's so much easier to scale NVIDIA by hiring more employees than it is to scale myself. And so the next scaling law is the agentic scaling law. It's kind of like multiplying AI. Multiplying AI, we could spin off agents as fast as you want to spin off agents. And so, you know, I... You know, I have four scaling laws. And as we use the agentic systems, they're gonna create a lot more data, they're gonna create a lot of experiences. Some of it we're gonna say, 'Wow, this is really good. We ought to memorize this.' That data set then comes all the way back to pre-training. We memorize and generalize it. We then refine it and fine-tune it back into post-training. Then we enhance it even more with test time, you know, and the agents, agentic systems, you know, put it out to the industry. And so this loop, this cycle, is gonna go on and on and on. It kinda comes down to basically intelligence is gonna scale by one thing, and that's compute.

Perfect.

Perfect.

So good.

For example, These AI model architectures are being invented about once every six months. Right? And system architectures and hardware architectures kind of every three years. And so you need to anticipate what likely is going to happen, you know, two, three years from now. And there's a couple ways that you could do that. First of all, we could do research internally ourselves, and that's one of the reasons why we have basic research, we have applied research. We create our own models. And so we have hands-on life experience right here. This is part of the co-design that I'm talking about. We're also the only AI company in the world that works with literally every AI company in the world. And so to the extent that we can, we try to get a sense of what are the challenges that people are experiencing.

That's right. You got to listen and learn from everybody. And have a... And then the last part is to have an architecture that's flexible, that can adapt and move with the wind. And one of the benefits of, of CUDA is that it's, you know, on the one hand, an incredible accelerator. On the other hand, it's really flexible. And so that balance, incredible balance between specialization, otherwise we can't accelerate the CPU, versus generalization, so that we can adapt with changing algorithms, that's really, really important. That's the reason why CUDA has been so resilient on the one hand, and yet we continue to enhance it. We're at CUDA 13.2, and so we're evolving the architecture so fast that we can stay with, you know, with the modern algorithms. For example... When mixtures of experts came out, That's the reason why we had NVLink 72 instead of NVLink eight. We could now take an entire four trillion, 10 trillion parameter model and put it in one computing domain as if it's running on one GPU. Um, people probably didn't notice, I said it, but if you look at the architecture of the Grace Blackwell racks, it was completely focused on doing one thing, processing the LLM. All of a sudden, one year later, you're looking at a Vera Rubin rack. It has storage accelerators. It has this incredible new CPU called Vera. It has Vera Rubin and NVLink 72 to run the LLMs. It also has this new additional rack called Rock. And so this entire rack system is completely different than the previous one, and it's got all these new components in it. And the reason for that is because the last one was designed to run MoE large language models, inference. And this one is to run agents and agents bang on tools, and—

No.

No, it's easier than that. You just reason about it. First of all, you just reason. no matter, no matter what happens, at some point in order for that large language model to be a digital worker... Let's just, let's just use that metaphor. Let's say that we want the LLM to be a digital worker. What does that have to do? It has to access ground truth. That's our file system. It has to be able to do research. It doesn't know everything. We don't have... And I don't wanna wait until this AI becomes, you know, universally smart about everything, past, present, and future before I make it useful. And so therefore, I might as well let it go do research. It's obviously, if it wants to help me, it's gotta use my tools. You know, a lot of people would say, 'You know AI is gonna completely destroy software. We don't need software anymore. We don't even need tools anymore.' That's ridiculous. Let's use the... Let's use a thought experiment. And you could just sit there, enjoy a glass of whiskey, and, and think about all these things, and it would become completely obvious. Like, if I were to create the most amazing- the most amazing agent that we can imagine in the next 10 years. Let's say it'd be a humanoid robot. If that humanoid robot were to be created, is it more likely that the humanoid robot comes into my house and uses the tools that I have to do the work that it needs to do? Or does this hand turns into a 10-pound hammer in one instance, turns into a scalpel in another instance, and in order to boil water, it beams, you know, microwaves out of its fingers? You know, or is it more likely just to use a microwave, you know? And the first time it goes up to the microwave, it probably doesn't know how to use it. But that's okay. It's connected to the internet. It reads the manual of this microwave, reads it, instantly becomes an expert. And so it uses it. And so I think the... I just described, in fact, almost all of the properties of OpenClaw. You know, that it's gonna use tools, that it's gonna access files, it's gonna be able to do research. It has I/O subsystem. And when you're done reasoning through it, reasoning about it in that way, Then you say, 'Oh, my gosh, the impact to the future of computing is deeply profound.' And the reason for that is, I think we've just reinvented the computer. And then now you say, 'Okay, when did we reason about that? When did we reason about OpenClaw?' If you take the OpenClaw schematic that I used at GTC, you'll find it two years ago. Literally, two years ago at GTC, I was talking about agentic systems that exactly reflect OpenClaw today. And, of course, the confluence of many things had to happen. First of all, we needed Claude and GPT and, you know, all of these models to reach a level of capability. So their innovation and their breakthroughs and their continued advances was really important. And then, of course, somebody had to create an open source, you know project that was sufficiently robust, you know, and sufficiently complete and that we can all put to work. And I think OpenClaw did for, did for agentic systems what ChatGPT did for generative systems. And I just think it's a very big deal.

Because consumers could reach it.

Yeah, no doubt.

Yeah, we jumped on it right away and we sent a bunch of security experts this way. And we did this thing called OpenShell. It's already been integrated into, into OpenClaw.

Yep, exactly.

We give you two out of three rights. Agentic systems can access sensitive information, it can execute code, and it can communicate externally. We could keep things safe if we gave you two out of those three capabilities at any time, but not all three. And out of those two out of three capabilities, we also give you access control based on whatever rights that you're given by enterprise. And then we connect it to a policy engine that all these enterprises already have. And so we're going to try to do our best to help OpenClaw become a better claw.

Power is a concern, but it's not the only concern. But that's the reason why we're pushing so hard on extreme co-design, so that we can improve the tokens per second per watt orders of magnitude every single year. And so in the last 10 years, Moore's Law would have progressed computing about 100 times in the last 10 years. We progressed and scaled up computing by a million times in the last 10 years. And so we're gonna keep on doing that through extreme co-design. So energy efficiency, perf per watt, completely affects the revenues of a company. It affects the revenues of a factory. And we're just going to push that to the limit so that we can keep on driving token costs down as fast as we can. You know, the our computer price is going up, but our token generation effectiveness is going up so much faster that token cost is coming down. It's just coming down an order of magnitude every year.

We should also get more power.

All the time, and we're working on it all the time. No company in history has ever grown at a scale that we're growing while accelerating that growth. It's incredible. And it's hard for people to even understand this. In the overall world of AI computing, we're increasing share. And so supply chain, upstream and downstream, are really important to us. I spend a lot of time informing all the CEOs that I work with, what are the dynamics that's going to cause, The growth to continue or even accelerate? It's part of the reasons why to the entire right-hand side of me were CEOs of practically the entire IT industry upstream and practically the entire infrastructure industry downstream. And they were all... There were several hundred CEOs. And I don't think there's ever been keynotes where several hundred CEOs show up. And part of it is, I'm telling them about our business condition now. I'm telling them about the growth drivers in the very near future and what's happening. And I'm also describing where are we going to go next so that they could use all of this information and all of the dynamics that are here to inform how they want to invest. And so I inform them that way like I inform my own employees. And then of course, then I make trips out to them and make sure that, 'Hey, listen, I want you to know this quarter, this coming year, this next year, these things are going to happen.' And if you look at the CEOs of the DRAM industry, The number one DRAM in the world was DDR memory for CPUs in data centers. About three years ago, I was able to convince several of the CEOs that even though at the time HBM memory was used quite scarcely, you know, and, and barely by supercomputers that this was going to be a mainstream memory for data centers in the future. And at first it sounded ridiculous, but several of the CEOs believed me and decided to invest in building HBM memories. Another memory was rather odd to put into a data center, is the low power memories that we use for cell phones. And we wanted them to adapt them for supercomputers in the data center. And they go, 'Cell phone memory for supercomputers?' And I explained to them why. Well, look at these two memories, LPDDR5, HBM4. The volumes are so incredible. All three of them had record years in history, and these are, these are 45-year companies. And so, you know, I... That's part of my job, is to inform and shape, inspire, you know.

Upstream, downstream.

GEV, Caterpillar. Yeah, that's downstream from us. Yeah, yeah, there you go.

Exactly, the deep science. The deep engineering, the incredible manufacturing, and so much of the manufacturing is already robotics, but we have a couple of hundred suppliers that contribute the technology that goes into our 1.3 million component rack. Each rack is 1.3, one and a half million components. There are 200 suppliers across the Vera Rubin rack.

But I'm doing, I'm doing all the things necessary to—

... yeah, see? I can go to sleep because I checked it off. I said, okay, you know, I go, I yeah, I can go to sleep and I go, well, let's see, what re- let's reason about this. What's important for us? Um, because okay, let's reason about this. Because we changed the system architecture from the original DGX-I that you remembered to NVLink-72 rack scale computing— ... what's gonna... What does that, what does that mean? What does that mean to software? What does that mean to engineering? What does that mean to how we design and test? How, and what does that mean to the supply chain? Well, one of the things that it meant was we moved supercomputer, supercomputer integration at the data center into supercomputer manufacturing in the supply chain. If you're doing that, you also have to recognize you're gonna move one... And if, if you're, if you're, you know, total footprint of whatever data center you're gonna build, let's say you would like to have, you know, 50 gigawatts of supercomputers that are running simultaneously, and it takes one week to manufacture that 50 gigawatts of supercomputers, then each week in the supply chain, the supercomputers are gonna need a gigawatt of power. And so, so we're gonna need the supply chain to increase the amount of power it has to build, test, to build and test the supercomputers in the supply chain before I ship it. Well, NVLink-72 literally builds supercomputers in the supply chain and ships 'em two, three tons at a time per rack. It used to be—they used to come in parts and we used to assemble 'em inside the data center. But that's impossible now because NVLink-72 is so dense. And so that's an example. And I would have to go and to, you know, I've... Fly into the supply chain, go meet my partners saying, 'Hey,' I said, 'guess what? So here's what I'm going to do with... This is the way we used to build our DGXs. We're gonna build them this way. This is gonna be so much better because we're going to need 'em for inference.' The market for inference is, you know, coming. The inflection point for inference is coming. It's gonna be a big market. And so I first explain to them what's going on, why it's gonna happen, and then I ask 'em to make several billion dollars of capital investments each. And because they, you know, they trust me and I'm very respectful of 'em, and I give 'em every opportunity to question me and I spend time to explain things to people and I reason about it. I draw on pictures and I reason about it in first principles. And by the time I'm done with them, they know what to do.

No.

Because, because I told 'em what I needed. They understood what I need. They told me what they're gonna go do, and I believe them what they're going to do.

One of the areas, Lex, that I would love us to talk about and just get the message out. Our power grid is designed for the worst case condition with some margin. Well, 99% of the time we're nowhere near the worst case condition because the worst case condition is a few days in the winter, a few days in the summer, and extreme weather. Most of the time we're nowhere near the worst case condition and we're probably running around, call it 60% of peak. And so 99% of the time, our power grid has excess power, and they're just sitting idle, but they have to be there sitting idle because just in case, when the time comes, hospitals have to be powered and infrastructure has to be powered and airports have to run and so on and so forth. And so the question that I have is whether we could go and help them understand and create contractual agreements and design computer architecture systems, data centers, such that when they need the maximum power for infrastructure in society, that the data centers would get less. But that's in a very rare instance anyways. And during that time, we either have a backup generator for that little part of it, or we just have our computers shift the workload somewhere else, or we have the computers just run slower. You know, we could degrade our performance, reduce our power consumption and provide for a slightly longer latency response when somebody asks for an answer. And so I think that way of using computers, of building data centers, instead of expecting 100% uptime and these contracts that are really, really quite rigorous, it's putting a lot of pressure on the grid to be able to... Now, they're gonna have to increase from their maximum. I just wanna use their excess. It's just sitting there.

I think it's a three-way problem. It starts with the end customer. The end customer puts requirements on the data centers that they can never not be available, okay? So that the end customer expects perfection. Now, in order to deliver that perfection, you need a combination of backup generators and your grid power supplier to deliver on perfection. And so everybody's gotta have six nines. Well, I think first of all, right now, we ought to have everybody understand that when the customer asks for these things, you have somebody in your data center operations team disconnected from the CEO. I bet the CEO doesn't know this. I'm gonna talk to all the CEOs. The CEOs are probably not paying any attention to the contracts that are being signed, and so everybody wants to sign the best contract, of course. And they go down to cloud service providers, and the contract, the two contract negotiators that are... You, I could just see them now. You know, negotiating these multi-year contracts. Both sides want the best contract. As a result, the CSPs then have to go down to the utilities, and they expect the six nines. And so I think the first thing is just make sure that all of the customers, the CEOs and the customers realize what they're asking for. Now, the second thing is we have to build data centers that gracefully degrade. And so if the power, if the utility, if the grid tells us, 'Listen, we're gonna have to back you down to about 80%,' we're gonna say, 'That's no problem at all.' We're just gonna move our workload around. We're gonna make sure that data's never lost, but we can reduce the computing rate and use less energy. The quality of service degrades a little bit. For the critical workloads, I shift that somewhere else right away so I don't have that problem, and so, you know, whichever data center still has 100% uptime, and so...

As soon as you could specify, you could engineer it. Beautifully put. So long as it obeys the laws of physics on first principles, I think we're good.

So the second thing is the data centers. And the third thing is we need the utilities to also recognize that this is an opportunity and instead of saying, 'Look, it's gonna take me five years to increase my grid capability,' if you're willing to take power of this level of guarantee, I can make them available for you next month and at this price. And so if utilities also offered more segments of power delivery promises, then I think everybody will figure out what to do with it. Yeah, but there's just way too much waste in the grid right now. We should go after it.

First of all, Elon is deep in so many different topics. Yet he's also a really good systems thinker. And so he's able to think through multiple disciplines, and he obviously pushes things, questions everything, where they're, number one, is it necessary? Number two, does it have to be done this way? And then numbers, you know, does it have to take this long? And so he has the ability to question everything, to the point where everything is down to its minimal amount that's necessary, you can't take anything else out. And yet the necessary capabilities of the product remains, you know? And so he is as minimalist as you could possibly imagine, and he does it at a system scale. I think... I also love the fact that he is represented. He is present at the point of action. You know, he'll just go there. If there's a problem, he'll just go there and then, 'Show me the problem.' You know, when you do all of this in combination, you overcome a lot of previous, 'This is just the way we do it.' 'Um, you know, I'm waiting for them.' You know, I mean, it's just, everybody has a lot of excuses. And so, and then the last thing is when you act personally with so much urgency it causes everybody else to act with urgency, you know? And every supplier has a lot of customers going on. Every supplier has a lot of projects going on, and he made it his... He makes it his business that he's the top priority of everybody else's, you know, projects. And so he does that by demonstrating it.

Yeah. That's right. That's right.

Well, first of all, co-design is an ultimate systems engineering problem. And so we approach the work that we do from that first. The other thing that we do and this is a philosophy, a thought, a state of mind, I guess, a method, that I started 30 years ago, and it's called the speed of light. The speed of light is not just about the speed. The speed of light is my shorthand for what's the limit of what physics can do. And so every single, everything that we do is compared against the speed of light. Memory speed, math speed, power, cost, time, effort, number of people, manufacturing cycle time. And when you think about latency versus throughput, when you think about cost versus throughput, cost versus capacity, all of these things, you test against the speed of light to achieve all of these different constraints separately. And then when you consider it together, you know you have to make compromises because a system that achieves extremely low latency versus a system that achieves very high throughput are architected fundamentally differently. But you want to know what's the speed of light of a system that achieves high throughput, what's the speed of light of a system that achieves low latency? And then when you think about the total system, you can make trade-offs. And so I force everybody to think about what's the first principles, the physical limits, for everything before we do anything. And we test everything against that. And so that's a good frame of mind. I don't love the other methods, which is continuous improvement. The problem with continuous improvement, first of all, you should engineer something from first principles at the speed of light thinking. Limit it only by physical limits, and physics limits. And after that, of course you would improve it over time. But I don't like going into a problem and somebody says, 'Hey, you know, it takes 74 days to do this today. And we can do it for you in 72 days.' You know, I'd rather strip it all back to zero and say, 'First of all, explain to me why 74 days in the first place. And let's think about what's possible today. And if I were to build it completely from scratch, you know, how long would it take?' Oftentimes, you'd be surprised. It might come to six days. Now, the rest of the 74, could be very well-reasoned and compromises, and cost reductions, and all kinds of different things. But at least you know what they are. And then now that you know that six days is possible, then the conversation from 74 to six, surprisingly much more effective.

That's just one pod.

Yeah, that's just one pod.

And Lex, we're probably gonna have to crank out about 200 of these pods a week, just to put it in perspective.

You know, the phrase that I use most often is, we need things to be as complex as necessary, but as simple as possible. And so the question is, is all that complexity there necessary? And we ought to test for that. And we got to challenge that. And then after that, everything else above it, you know, is gratuitous.

It is the most complex computer the world has ever made.

Well, it's gold medalists in every single sport, all assembled right here.

A whole bunch of reasons for, well, first of all, let's start with some facts. 50% of the world's AI researchers are Chinese, plus or minus, and they're mostly in China still. We have many of them here, but there's amazing researchers still in China. Their tech industry showed up at precisely the right time. At the time of the mobile cloud era, their way of contributing was software, and so this is a country with incredible science and math. Really well-educated kids. Their tech industry was created during the era of software. They're very comfortable with modern software. China is not one giant economic country. It's got many provinces and cities with mayors all competing with each other. That's the reason why there's so many EV companies. That's the reason why there's so many AI companies. That's the reason why there's so many, every company you could imagine, they all create some of them. And as a result, they have insane competition internally. And, you know, what remains is an incredible company. They also have a social culture where it's family first, friends second, and company third. And so, the amount of conversation that goes back and forth between... They're essentially open source all the time. So the fact that they contribute more to open source is so sensible because they're probably, 'What are we protecting?' You know, my engineers, their brothers are in that company, their friends are in that company, and they're all schoolmates. You know, the schoolmate concept. There's a, you know, one schoolmate, you're brother for life. And so they share knowledge very, very quickly. And so there's no sense keeping technology hidden. You might as well put it on open source. And so the open source community then amplifies, accelerates the innovation process. So you get this rapid, incredible great talent, rapid innovation because of open source and just, you know, the nature of friends, and insane competition. Among the company, what emerges is incredible stuff. And so this is the fastest innovating country in the world today, and this is something that has everything that I've just said is fundamental to just how the kids were grown, the fact that they have excellent education, the fact that parents want them to do well in school, the fact that their culture is that way. These are just the thing about their country, and they showed up at precisely the time when technology is going through that exponential.

It's a builder nation.

Yeah, it's a builder nation. Our country's leaders, incredible, but they're mostly lawyers. Their country's leaders, and because they're trying to keep us safe, rule of law, governing, their country was built out of poverty. And so most of their leaders are incredible engineers. Some of the brightest minds.

Love it, yeah.

First, if we're gonna be a great AI computing company, we have to understand how AI models are evolving. One of the things that I love about Nemotron 3 is it's not just a pure transformer model, it's transformer and SSMs. And we were early in developing the conditional GANs, which led step-by-step to diffusion. And so, the fact that we're doing basic research in model architecture and in different domains gives us visibility into what kind of computing systems would do a good job for future models. And so it is part of our extreme co-design strategy. Second, I think we rightfully recognize that on the one hand, we want world-class models as products, and they should be proprietary. On the other hand, we also want AI to diffuse into every industry and every country, every researcher, every student. And if everything is proprietary, it's hard to do research and it's hard to innovate on top of, around, with. And so open source is fundamentally necessary for many industries to join the AI revolution. NVIDIA has the scale and we have the motives, not only skills, scale, and motivation to build and continue to build these AI models for as long as we shall live. And so therefore, we ought to do that. We can activate every industry, every researcher, every country to be able to join the AI revolution. There's the third reason, which is from that, recognizing that AI is not just language. These AIs will likely use tools and models and sub-agents that were trained on other modalities of information. Maybe it's biology or chemistry or laws of physics, or fluids and thermodynamics, and not all of it is in language structure. And so somebody has to go make sure that weather prediction, biology, AI for biology, physical AI, all of that stuff can be pushed to the limits and pushed to the frontier. We don't build cars, but we wanna make sure every car company has access to great models. We don't discover drugs, but I wanna make sure that Lilly has the world's best biology AI systems, so that they can go use it for discovering drugs. And so these three fundamental reasons, both in recognizing that AI is not just language, that AI is really broad, that we wanna engage everybody into the world of AI, and then also co-design of AI.

Yeah, I appreciate you were saying that. We open sourced the models, we open sourced the weights, we open sourced the data, we open sourced how we created it. Yeah, it's pretty amazing.

You know, first of all, the deepest misunderstanding about TSMC is that their technology is all they have. That somehow they have a really great transistor, and if somebody shows up another transistor, game over. It's the technology and of course, I don't mean just the transistor, the metallization systems, the packaging, the 3D packaging, the silicon photonics, all of the technology that they have. That technology is really what makes the company special. Their technology makes the company special. But their ability to orchestrate the demands, the dynamic demands of hundreds of companies in the world as they're moving up, shifting out, increasing, decreasing, pushing out, pulling in, changing from customer to customer, wafer starting, wafer stopping, emergency wafer starts, all of this dynamics of the world's complexity as the world is shape-shifting all the time, and somehow they're running a factory with high throughput, high yields, really great costs, excellent customer service. They take their work, they take their promises seriously. When the wafers were promised to show up, the wafers show up, so that you could run your company appropriately. And so their manufacturing system is completely miraculous, I would say. Then the second thing is their culture. This culture is simultaneously technology focused on one hand, advancing technology, simultaneously customer service oriented on the other hand. A lot of companies are very customer service oriented, but they're not very technology excellent. They're not at the bleeding edge of technology. There are a lot of companies who are at the bleeding edge of technology, but they're not the best customer service oriented company. And so it just depends on somehow they've balanced these two and they're world-class at both. And then probably the third thing is the technology that I most value in them that they created this intangible called trust. I trust them to put my company on top of them. That's a very big deal.

Three decades, I don't know how many tens, hundreds of billions of dollars of business we've done through them, and we don't have a contract. That's pretty great.

Story is true. I didn't dismiss it. But I was deeply honored and, of course, I knew then as I know now, TSMC is one of the most consequential companies in history. And Morris is one of the highest regarded executive and business and personal friend that I've had in my life. And for him to ask is, I was humbled and really honored. But the work that I'm doing here is really important, and I've seen, in my mind's eye, what NVIDIA was going to be and what the impact that we could have. And it was really important work. And it's my responsibility, my sole responsibility to make this happen. And so I declined it, not because it wasn't an incredible offer. It's an unbelievable offer but I simply couldn't take it.

Yeah. Yeah, no doubt. So now I can help both companies.

Our single most important property as a company is the install base of our computing platform. Our single most important thing today is the install base of CUDA. Now, the reason why, 20 years ago, of course, there was no install base. But what makes... And if somebody came up with a GUDA or TUDA it wouldn't make any difference at all. And the reason for that is because it's never been just about the technology. The technology, of course, was incredible visionary. But it's the fact that the company was dedicated to it, stuck with it, expanded its reach. It wasn't three people that made CUDA successful. It was 43,000 people that made CUDA successful. And the several million developers that believed in us, that trusted that we were going to continue to make CUDA 1, 2, 3, 13, that they decided to port and dedicate their software on top of it, their mountain of software on top of it. And so the install base is the number one most important advantage. That install base, when you amplify it with the velocity of our execution at the scale that we're talking about, no company in history had ever built systems of this complexity, period. And then to build it once a year is impossible. And that velocity combined with the install base, in the developer's mind, you just go now, take the developer's mind. From the developer's perspective, if I support CUDA, tomorrow it'll be 10 times better. I just have to wait six months on average. Not only that, if I develop it on CUDA, I reach a few hundred million people, computers. I'm in every cloud, I'm in every computer company, I'm in every single industry, I'm in every single country. So if I create an open source package and I put it on CUDA first, I get these both attributes simultaneously. And not only that, I trust 100% that NVIDIA is going to keep CUDA around and maintain it and improve it and keep optimizing the libraries for as long as they shall live. You could take that to the bank, and that last part, trust. You put all that stuff together, if I were a developer today, I would target CUDA first. I would target CUDA most. And that's the reason that I think in the final analysis is our first, that's even our first core advantage. Our second one is our ecosystem. The fact that we vertically integrated this incredibly complex system, but we integrate it horizontally into every single company's computers. We're into Google Cloud, we're into Amazon, we're in Azure. You know, we're ramping up AWS like crazy right now. We're in new companies like CoreWeave and Nscale. We're in supercomputers at Lilly. We're in enterprise computers. We're at the edge in radio base stations. You know, I mean, it's just crazy. One architecture is in all these different systems. We're in cars, we're in robots, we're in satellites, we're out in space. And so the fact that you have this one architecture and the ecosystem is so broad, it basically covers every single industry in the world.