원문 보기
Some call it a vision. Vision is a very big word for me , because I believe, above all, that vision matters. We've stopped the weekly show, and there are only three people we're doing it for. President Trump, Jesus, and Jensen. The number one podcast in the world. This is Jensen Huang. He is the founder, president, and CEO of Nvidia.
Whether you know it or not , his decisions shape your future. Nvidia is the most important stock in this market, and Jensen is probably the best CEO in history. Revenue increased by 97% year-on-year. Demand is not only high , it is actually accelerating. Nvidia is the only computing platform that is a full-fledged AI factory.
GPU is like a time machine because it allows you to see the future earlier. And if we can see the future and predict it, then we have a better chance of making that future the best it can be. Welcome Jensen Huang. Oh, we need to stand up to say hello. Oh, well, that's it. We get up. We stand up to greet him.
This is our boy. Ladies and gentlemen, GPU Jesus. They love you. Thank you. I love you too . We love the number one podcast in the world. world. Undoubtedly. Wow. We like the new jacket. So, did you sell the old one at auction? I felt like you guys needed some energy. Yes. This is the right jacket. We are talking about serious things, but we need to talk about it with energy.
Yes. Let's start with this essay from the weekend. From which one exactly? Let's start with Dario's essay, because I think... Hemingway was involved? In general, did anyone chase him through Pangram? I don't even know how much AI helped there, but it was something incredible. And then, I think many were surprised by how Frontier Labs rallied around this essay.
Jensen, tell us what happened, how you read it, how you interpreted it, and then we'll get into the details inside. But, perhaps, to start, let's share some general thoughts. Well, first of all, there was a lot going on. Yes. And first there is the part about security, which we have to take very seriously.
Safety is the most important thing, obviously. Security and leadership are not wrong choices. You are able to innovate quickly. You are able to act quickly. And America can lead while doing so safely. I think this is a wrong choice, but safety is certainly important. Here is the issue of internal control, which I think he was talking about.
Of course, the situation with whistleblower Koxin is very serious. When a whistleblower comes forward, it should be taken very seriously. I think Koxin showed great courage in voicing his concerns. And even then there were some issues that were somewhat mixed up . I think exposure is normal. But scientific predictions about the future are less normal because they are clearly not based on science.
This was expressed by scientists, but clearly had no scientific basis. And that's why I don't agree with this. But of course, the part about exposure...I think there's a lot of stuff in there. Pauses, slowing down—are these all voluntary steps they can take if they feel the company is losing control?
If Koxin saw something, we don't know what exactly, but if he saw that the company was out of control, perhaps it was a shift from research to engineering. As you know, these labs are moving from research to engineering. Extraordinary talents, extraordinary engineering. But, of course, engineering is different from research.
Maybe this transition is awkward, we don't know what he saw, and ultimately only he knows. But if it's about lack of control, that's another topic. How should the government deal with this? Suddenly, regulation appears that covers everything in one fell swoop. Can you help us figure this out somehow?
We tried to play this game this week on the podcast, and it was hard: how do you describe, for example, when my mom calls me and says, "Dude, what's this story about the end of civilization?" I don't know how to explain this to her . So when very smart people try to quantify and quantify it. I think this is what worries many people.
They ask, "What does that mean? A 10% chance of extinction?" No one knows how to explain to the average person how this is even possible. Well, first of all, we shouldn't do this. Because it's a fiction. First, I think We shouldn't, because it's fiction. And these... these are highly educated people.
They are called researchers. They, of course, work in a laboratory. So, combining these words, and then, and then, and then, this prediction is alarming and worrying, and it should not be done. This is irresponsible. Now, essentially, let's go back and look at the real facts. The facts are as follows: there was a prediction that in 5 years radiology would be completely replaced by artificial intelligence and there would be no radiologist left in the world .
In fact, everything turned out to be exactly the opposite. Today, the world needs more radiologists than ever before. However, AI has completely taken over radiology, and that's great. This is automated scan reading, and it's great. There was a prediction that within 6–12 months, maybe last year, within 6–12 months, 90% of the code would already be generated by AI.
This turned out to be a mistake. Within 6–9 months, as predicted last year, 50% of the initial jobs will be destroyed. This turned out to be untrue. Let's see what else, what else turned out to be wrong? I mean, all of these predictions were wrong. True. And if you're on the fence about GPT-2 being too dangerous to release, that Llama 3 is too dangerous to release.
Oh, one. Oh, yes, we heard those Half of office jobs will disappear next year. Apocalypse, yes. Yes. We must be responsible. We must be held accountable for all the stupid predictions made. True. Someone has to Someone has to take charge Yes. And so we just have to keep track of all of this, and of course people do, and remind us that these predictions are ultimately incompatible with America winning the AI race.
In short, some people say, you know, they say " trust the experts ," and they used the analogy with COVID, which again started with people who were researchers, educated people who had an asymmetrical understanding that we didn't have, and said things that we later found out were not true. And so there's this war going on right now between the " trust the experts" movement and, well, let's just look at the real history of these predictions and just think more methodically.
Where does this come from? Because it comes from within the very places where it is created. For example, what do you think is the psychological component or what is the real incentive? Maybe it's a business incentive. Perhaps this is a political incentive. Can you guess, or what do you think, why they do this?
Well, first of all, I have to say, these are some of the most influential companies in history. Extraordinary engineers , extraordinary researchers, truly fantastic work. On the one hand, I work very closely with them as a company with companies. On the other hand, we should have such conversations publicly.
And that's very unfortunate. And I think these companies should actually be built the way we built companies before, which is in silence . True. You know, and So, wait. Jensen, you don't let anyone in your organization speak for the entire company, especially when they've had a bad weekend and are leaving with a scandal.
Are they not allowed to tweet on your behalf and on behalf of the organization? No, because, well, that's what they decided when they came to work for us. And we told them: this is how you should behave when you work for our company. And if you like our company culture, and as you know, Nvidia's culture and our employees are incredibly happy.
Yes. They like that the company is consistent, that we are stable, that our core values are taking care of families and creating an environment in which they can do their life's work. That we are doing meaningful work. We make it as quiet as possible. And we contribute to the success of others, which we are very proud of.
So people are attracted to such core values. But when you come to work for our company, there are also things we don't welcome. For example, we do not welcome political discussions within our company. Leave that at home. You can talk about politics outside the company. We are Yes. The company is apolitical.
You know, we are bipartisan. We want America to succeed. And we want to make sure that, no matter what government is in power, we do everything in our power to help America succeed. So conversations about race, religion, politics and all that stuff, we tell people: do it outside the company. This is not for us.
on maybe regulating AI, more narrowly, Satya was here this morning and he said that before we talk about regulation that could slow down the process, why don't we get the basics right? Why not set up a measurement? Why not establish standardization? What do you insist on regarding the engineering component?
To fix the engineering , of course. Implement research more predictably so that we don't spread panic. Keeping everything inside until we are ready to make it public. What do you think would be the correct reaction? Demis had a proposal for creating an organization similar to FINRA. It is not entirely clear what Demis wants from this transnational structure, which has some control.
What is your opinion on this ? What do we need now? Regulation should solve real problems. The question is , what real problems have we already faced? That's right. And if you look at the real facts, all the problems so far have come from laboratories. And the reason for this, it is worth noting in their defense, is that they have the greatest computing power.
True. And the reason for this is that they are trying to solve problems at the limit of their capabilities. So, it is logical that it is from advanced laboratories that the greatest danger comes. It is unlikely that a high school student will do anything because he simply won't have enough computing resources.
Yes. And it is unlikely that the startup will be the cause, because they also won't have enough capacity. In fact, if you look at the entire planet, no one will have enough computing resources except for advanced laboratories. So now the question is, if you look at the real events, they are doing groundbreaking work.
This is really very difficult. They move from research to engineering. I can imagine, they are obviously creating some of the most significant technologies and companies in the world. They build the company, culture, technology, engineering, and products all at the same time. So I can understand that the situation is a bit critical.
However, four incidents in one lab, one major incident in another— the first thing to do is find the root cause of the problem from an engineering perspective. What happened? What could we have done differently? And what are we going to implement and institutionalize , whether it's technology, whether it's methods, whether it's processes, to ensure that we don't let this happen again.
I can bet you money that in each of these cases it is within their power to prevent this in the future. Because the alternative, if it's not in their power, and I'm sure they are... I'm sure these four incidents won't happen again. Um, I'm sure they found the root cause and fixed it. I'm sure they have the technology now, much better technology for, you know, sandboxes, runtime environments, monitors, and constant monitoring, and I'm sure they have much better technology now.
The alternative is also unlikely, which is for them to say, " Look, we've had these incidents. After we've analyzed everything, we've come to the conclusion that we do n't understand anything, we have no idea how to control this, and we're asking the public for help." Yes. Well, if that's the case, then we should, you know, a bunch of companies with engineers should send their specialists there.
I mean, we should advise them if we can, but I doubt it. I think they have excellent specialists, they will handle this . Well, we don't work in a vacuum. David, last night you informed me that there is a Chinese lab, the developers of GLM, who are going to allocate 3 billion for recursive self-improvement.
So maybe you could prepare this for Jensen. That's what was announced. So, at zeku.com, the founder just raised $5 billion and said that one of their priorities will be to try to create an AI that trains the next AI and automate this process as much as possible. Um, yeah, I think what I mean is... Well, it's a new buzzword, but as you know, RSV is a combination of a system of ideas.
It all starts with contextual things. It starts with skills. It starts with reflection. It starts with, you know , reinforcement learning and synthetic data generation. And these are all very sound ideas that make AI better at solving problems over time. And you can also use low-rank adaptation. You know, all of this doesn't include weight.
You can actually improve the scales, and it's called LoRA. Laura, this can be improved by generating synthetic data and doing reinforcement learning without retraining the base model itself, and over time, you can retrain the base model on all this experience. Therefore, I believe it makes perfect sense to use this technology to improve productivity in all kinds of tasks, including creating AI.
I think this is a very logical idea, and I'm sure everyone uses it to one degree or another. It's just that this phrase is now being used to weaponize technology in some way and maybe set it up as if it's going to get out of control—that's the impression they're trying to create , but you don't believe it's real.
No, of course not, and the reason is that you can do this all day long in your company, but when you release a product, you have to evaluate it, right? You should test it again, make sure there is no regression, right? So, the basic control process: these labs, moving from research to engineering, will have much better control.
Yes. And when they have much better control, that control will come from methods, knowledge, practice, tools, and technology. All this leads to better control, verification and evaluation, Yes. This will allow for development within the company and production of quality products outside of it . About open source for a moment.
I mean, with Hugging Face, we talked about it, and I said, "This is going to be one of the most significant acquisitions." I do n't even want to call it a deal because I think that's more important. Give us your explanation of the basic principles of open source vs. closed source and open weights, and how the ecosystem should evolve over time.
The world needs both closed and open models. You want to use—I use as many closed models as possible. I used four of them this weekend. And they work great. They are advanced . This is a wonderful experience. They work incredibly well. They are getting better every day. And the way I think about closed models is kind of like bottled water, you know?
Water is free, friends. I don't know if I told you, but the water is free. I don't want to, you know, destroy your illusions, but water is free. And this morning I used a lot of free water while showering. And so you use the right water in the right places. And it's no different from electricity. This, you know, is no different from all the kinds of goods that we use in the world.
You need both . So, in the case of open models, the reason why they are needed is due to considerations of sovereignty, privacy, or technology ownership. Look at the facts. The facts are as follows: in the last 6 months, $400 billion in venture capital investment has gone into companies specializing in AI.
80% of them use open models. If there were no open models, how would they be able to build their dream? That's right. Because their dreams may be different. Of course, they will differ from the dreams of advanced laboratories. And in America, there are so many ways to innovate. This is one of our main advantages.
Great ideas just flow, and that's why open models facilitate this. Open models empower everyone. If we want to win the AI race, it's not about a few tech companies winning it. We're talking about every company in America. Every company, every industry, every researcher, every teacher, every student, every startup—everyone wins.
Some of them will use closed models. Many will use open models. There are 10 million startups. Does it matter what the distribution is? Let me ask, does it matter whether the open models come from China or the US? Well, we do our best to contribute to open models. However, at the moment you download, for example, the vast majority of open source contributions today come from China.
They just have a lot more engineers. They produce everything on a large scale because it's a big country. And so they train students in science and mathematics in large numbers. Yes. This is one of our weaknesses. Yes. They train them through amazing universities like Tsinghua University in large numbers.
Well, they are contributing to open source today. We load Linux, we load Kubernetes, we load all this software. Chinese experts contributed to a significant part of it . And once you download it, it's yours. We create forks, improve it, make it our own. So when you download one of these Chinese models, it was just created by really great researchers in China, but now it's yours.
Do whatever you want with her . Is this a race? Race? Yes. I think this is a very apt remark. My point is that the race is really about who uses technology best. You know, during the last industrial revolution, all the inventors were like Maxwell, Volta, Ampere. None of them were American. They were Europeans, right?
The last industrial revolution came from Europe. But we used it. We took advantage of it socially better than anyone else in the world, and look what it did for us. I want to make sure the next generation goes through the same thing. Yes, yes. So why are communists so successful in getting their ideas across now?
You know, I think, first of all , their narrative is much more practical. This narrative is much more practical. In China, no one is saying that this or that is coming to an end, or about some catastrophes or anything like that. About everything being lost . That everything is gone. They are much more pragmatic about it.
They see AI as a technology that will advance their economy and society. And they don't have these groups that are actually saying that this will lead to the end of civilization. And we're making all this up. The most annoying thing is that if this were true, if it were truly true, we would have to discuss it and do something about it.
True. Even if it's true . We should spend more time doing something about this, not scaring a bunch of people who can't do anything. Our task is to develop it correctly. Has there ever been a time in history when so many people so passionately claimed something so false? And this is measurable, this is actually demonstrably false.
And it's even logical that this is not true. It's not based on science, it's not based on research. Everything based on science and research suggests otherwise. on the border of the unknown. People have never been there, we have never seen it, so we are afraid of it. And so it's easy to tell everyone to be afraid of it.
It could also be a life experience, David. Well, let me give you an example. When I graduated, I was an engineer and didn't type very much. The reason is that I belonged to the first generation, before software became popular. We had to build computers to make software possible. Can you imagine that in this generation, every engineer who comes into the world of engineering spends all their time typing.
Literally, that's what they do. When you get a job, you are given a laptop, a chair, and you start typing. You type all day long. You've been typing since the moment you wake up. Well , engineering existed before typing. True. So, can you imagine that there's a mountain of engineering work in the world where most of it no longer requires printing?
Of course. We were We had busy engineers even before printing. I think we'll do a lot of great engineering work after printing. Yes. When I say " typing," I mean programming. I mean Yes. So, even at Nvidia, when software engineers talk to me, I tell them: you just type. I say this forever, but of course, jokingly.
And I tell them that my favorite key is backspace. And the reason for this is that the best software is the least software. Yes. I want you to use backspace as software. actually let's talk about Nvidia. I Let's do a little Nvidia analysis. So, analysis simply means explaining the details, because there is a grand strategy involved here.
Let's start from the very bottom . So, oh no. It wasn't planned, but we know who it is. Oh no. No. Mr. President. Oh, yes, sir. Um, I have something to tell you. I would be on stage with the " best" if it weren't for your call. I'm on stage with the "best". I'm on stage with the " best". I'm on stage with Sax.
Yes. You know, you know the whole group. Yes, Jason is here, Chamat is here, David and David is here . Yes, I'm sitting in front of several thousand people. And we were talking, as it turned out, we were talking about you. Good job, sir. Good job. The fact that you've figured all this out, I mean , there's a lot of complexity here, and the fact that it's important, you've figured all this out, and I, you know, we're all very grateful.
Give him my regards. So, do you want to say hello to the crowd? Jason would like Jason would like to put you on Even Jason. Loud speakerphone . How do we How do we connect the president? Can we put him on speakerphone? Loudspeaker. Yes. Straight into the microphone. We will go on air now. Wait a second, sir.
Wait, sir. We take the microphone. Mr. President. You are now talking to the entire planet. You see, the great thing about life is that Jensen can design the most complex computer chip in the world that no one can copy for 10 years, but he can't figure out how to put me on speakerphone. Completely. It's almost a conspiracy.
And the happiest group is China, and China is very happy. And I can even say that there are many states in the country that are happy, even though they should n't have gotten anything, because they are overflowing with people who want to be there. But now suddenly you see what they are building in Finland; they want to build one.
Google wants to build a big center in Finland, which I'm not happy about because they couldn't get permission. And I'm telling you, it's all a hoax. Data centers are great, they make people rich, they make states rich, and they are the oil of the next 20–25 years. This is more than the internet and AI, you know, much more.
And they just play into the hands of a lot of people who don't want this to happen, it could be politicians, it could be China. And we will not let this happen. This, this is a hoax, and. You're right, we won't let that happen, sir. No, we, we're going to let this happen, and robots won't take over the world, and, uh, that's not going to happen.
You know, my uncle was the best, probably the best of all time, frankly, a professor at MIT for 41, 42 years, known as one of the most brilliant people, and he was there for 41 years at the top. He was at the very top, at the peak of his career, he had done many things. Jensen knows all about it , he's done a lot of things.
So I have a little bit of genetic, uh, a little bit of genetic power, if you believe in the resource theory, which I do. I have genetic strength . Robots won't be why you know so much about AI. Well, I know about AI. I know, and I also have common sense about AI. Uh, robots won't take over. Uh, AI won't take over the rest of the world.
All this is a hoax. Now, with that in mind, we have to be a little careful. We have to be, you know, we have to act sensibly, but that doesn't mean we're going to stop the industry while we work on how to destroy it over the next 10 years. So I'm with you until the end. I didn't even know how you felt about it .
I assumed you felt the same way I did . And if we want to be leaders, I have this saying: whoever wins in AI wins everything. That's how important it is. This is bigger than the internet. And whoever wins in AI wins, and we cannot allow this to happen. And this largely applies to data centers. There are communities that were in decline, but now have data centers and have become wealthy communities.
Truly wealthy communities. We're going to make sure that we're going to make sure that everyone...we're going to make sure that everyone wins the AI race in America. Every industry , every company, every state, every person. Good. Well, I have strong beliefs about this and I am in a position that allows me to do something about it.
We will not allow this to happen. So, I have no idea who is at this meeting. I have no idea who the hell I'm talking to, but I... Did you hear that? Did you hear that? Did you hear that? Thousands of people applaud you, sir. All I know is that if you're there, they'll listen to Jensen, but he did an incredible job and David did an incredible job.
And good luck to everyone, and we will hold on to the future. The country has never been in better shape. We are bringing $20 trillion in investment into the country , compared to less than $1 trillion under a sleepy Joe Biden. And that was four years ago. This is in one year. So the country has never seen anything like this, and we're going to continue to do this .
So, thank you all very much . Thank you, Mr. President. Mr. President. I'll call you back later. Thank you, Mr. President. Thank you. Um, I thought this was something unique. I thought it was a prank. Did you know this was happening? I thought it was a prank. You know what it was ... You thought it was...
No, it really was. At first I thought it was a prank and said: put it on speakerphone. Wow. And he calls you... Well, how are you...how are you... He calls you at all hours of the night, right? We were, we were in, uh, we were in the Oval Office when he called you. I was sleeping, I was sleeping. I was in bed.
You were sleeping, and he said, "Wake him up." And I felt so terrible because he asked, "Who's coming to this dinner?" We're looking at the list and he says, "Well, what about Jensen?" I replied: "No, sir, we...he's on vacation because he had to postpone it for 5 years ." And he says: "Put him on the phone." What kind of vacation?
But why do you think he sees it as a deception? This...this is exactly what I was asking people about. This is an extraordinary thing. This is...a rating of minus 80. Yes. So anyone else sitting in the Oval Office will do what is popular. You represent the people. This is what everyone wants . They want to close data centers, and AI seems to be trending right now .
But he says it's a hoax. And he calls things by their proper names. How does he do it? I have to tell you, I'm not sure. And the reason is that many people are drawn to it. So the fact remains : this is all complicated. You know, first of all, if you look at the history, if you look at the materials, everything rests on two things.
The first thing it was based on was national security. And recently it was all shattered to pieces. Right. And so, this story is no longer based on national security . Now it is based on security. So, if you want AI to be safe, first of all, we have to make sure that the labs that develop it have everything under control.
So that they have good tests . If we want to involve third parties to make sure that independent evaluators are available. This is no different from financial control . You know we have auditors. And the auditors are quite...they don't have to be as experts in our business as we are, they just have to ask the right questions, and I think someone said it's good to have independent auditors or appraisers, but there should be several of them.
I agree with that too. Just as there are many appraisers and auditors, this ensures that no company becomes dependent or in any way bribed. So there are many different ways that this issue can be addressed. So I think the main thing is: let's develop technologies safely, let's make sure that their testing is safe.
And I fully realize that what is being created is extraordinary, but these are extraordinary companies and we have to make extraordinary demands of them, and they want it and I want it. I wanted to go back to open source for a second: a year ago we weren't taking it seriously, it was two years behind , 18 months behind.
The thing is, as you know, one of the problems when talking to President Trump is that it's hard to say anything. I'll definitely get a kick out of this, I'm sure. He will remind me of that. But anyway, what I was going to tell him and all of you is that artificial intelligence is creating a huge number of jobs.
What he wanted most at the beginning of the administration, and what was discussed in my first phone conversation with him when we first met, is that he wants to create jobs in America. He wants to reindustrialize the United States. He wants to make sure the United States has the energy to support the next industrial revolution.
Without energy, there is no industrial growth. So he wants to be sure that energy is growing , jobs are growing, and that we are reindustrializing the supply chain. Look at everything we're doing now. All of this is happening right now as we speak. We are creating more jobs than ever before. We create jobs in the software industry .
We just talked about this earlier: $400 billion in venture funding has been poured into the AI industry for absolutely no reason. Six months. Well, it created a bunch of jobs. This created a lot of jobs. This has , of course, created a huge demand for computing power, which I am very happy about. Which also creates a high demand for data centers.
And we had to talk about it . I think I just spoke with Texas Governor Abbott, and he wants to appeal to the industry to be empathetic to small communities as we build data centers across America, just to hear them better. You Let's actually talk about this for a minute. What's amazing about Nvidia, if you break it down into its components, is that you actually had to become a bank for AI to get this ecosystem going.
And you had to do it at all levels. You know, you just did this with Cloverleaf, where you're in land, energy and construction. You did this great thing with BlackRock, Goldman, and others to essentially create financial opportunity . Tell us about your capital allocation strategy. What needs to happen for the broader ecosystem to be able to get involved and fund this next phase?
Well, we are creating, as you know, this is a new industrial revolution , and every aspect of it is true. This new industry requires manufacturing just like electricity, the internet, and now AI ; we feed everything, we can find everything, now with AI we can ask and know everything. Is n't that right?
And here is our future. We connect to the airwaves and can ask him anything, and he can explain it to us. Now, for this to happen, he must create intelligence. So, this is a production process, which is why this infrastructure needs to be built. But once you build the infrastructure, the question becomes, what about all the other levels across the United States?
This industry is not just a model. It's not just chips. These are mostly applications on top of them. This is mainly the infrastructure level, data centers, and all the infrastructure, construction, electricity, energy generation. All of this is involved. So I look at the entire ecosystem and look for " bottlenecks." And if there are places where extraordinary companies are built , restrictions.
The constraints, the extraordinary companies that are being built, maybe it's the supply chain that needs to be scaled so that when we're ready to deploy computing, they're ready for us: land, energy, construction. So it's no different from forward supply chain analysis. You know, I probably think about the long-term supply chain more than others because our company is very large, and for our success, I need to have a whole range of companies supporting me.
You know, it has to be, Corning has to, you know, Corning has to support me, Lumentum and of course TSMC and also the memory manufacturers. And so we started working with all these companies long before the revolution, before the growth came, so that it could even happen. Now, I'm moving on to... Now I'm moving down the chain.
I have an observation that profits over time, over long periods, tend to move up the stack, right up to the application level, where you can earn excess profits for a longer period of time. Um, I mean , you bought Hugging Face. Now you can say that you are actively involved in the service business. It seems only natural to me that products like Open Router make a lot of sense.
It seems pretty obvious that there are better versions of how to create things like Bedrock. I'm sure you're thinking about it . What is the natural conclusion? Because it seems like the guys at the top have no problem moving down. Mhm. And you have the best balance, incredible engineers, and you have the proven experience to do everything right, design a product, and release it.
So how do you look up and say to yourself, "I could probably do that"? The reason Nvidia launches every model in the world, we...we are the only ones...It's incredible. Last year, about a year and a half ago, the only thing we launched was OpenAI . Yes. Now, look at the amazing models available. Meta Muse is available.
You have Grok, it's available. The Grok bot is incredible. Um, now we're launching Gemini. And Anthropic also scales on our platform. Since a year and a half ago, all these advanced AI models have appeared, which are now open and accessible. So, the number of models is growing. There are a whole bunch of companies that I won't name that also make advanced models.
And the number of AI labs is growing. Yes, Ineffables, Reflections, Really? The list goes on. Physical Intelligence, the list goes on. Of course? And that's why all these labs build their solutions on Nvidia. And the reason for this is that as a company, I prefer to help everyone succeed, instead of taking a share for myself.
So we go as high as we need to, but stay as low as possible. Ahem. Our strategy: go as high as you need to and stay as low as possible. And the reason for that is, if I had done this, if I had solved this...if it weren't for Nvidia, who created CuDNN, none of the frameworks would exist. If it weren't for us , who created Megatron, Megatron core, then no large-scale training would have taken place.
Education would simply not exist. So we go and invent all the necessary technologies , as much as is needed, and then we let thousands of flowers bloom. And this position allows us to be, frankly speaking , the only ones, let's be frank. Me, I agree with you. The objection would be that it would be great to have more competition at the hyperscaler level.
And I think you've done a great job supporting neo- clouds. There are some...And by the way, I think you introduced me to Nvidea's superb. Perfectly. Everything there is incredible. But we need about 50 of these guys. We need a hundred of them. We need a thousand of them. And they can just take some Yes.
Yeah, you know, Chamate, I'm just surprisingly uncompetitive. Really? Yes. This is not mine. You know, my—this is, for example, I would be more than happy to have five hyperscalers. However, the reason... pay attention, the early customers of all neo- clouds, all those that we call NCPs, were hyperscalers.
Exactly. And the reason is that hyperscalers plan once a year, but the market dynamics are so volatile right now. Right. that they are almost always wrong. So with all these regional clouds that are agile and can act quickly, they know their state, their country, their region, they provide land, energy and shell in a way that's hard for someone sitting in Seattle or Palo Alto to see.
M-hm. Yes. So we now have, essentially, a large-scale distributed network of companies that build and provide land, energy, and shell for us. And now countries understand that this is strategic. Yes. Many countries say, " I'll take my energy and give it only to my companies." Right. Well, Nvidia is also operating in this country and we could help Neo Clouds in this country to grow and.
So whether it's Fermas in Australia, we just did a lot there, brought in another two gigawatts, in Southeast Asia, of course, IOH and others are bringing in a few gigawatts. So, we are building gigawatts. We're building, you know, we're scaling. Pretty good, though. I, I want to add one thing. It's understandable that you're climbing pretty high and focusing a lot on open source.
Obviously, your NeMo Trons are showing exceptional results. I use them often. Hugging Face, PoolSide, and Laguna—a very, very solid product that you are currently acquiring, leasing, or whatever it is called. And then you have the open source stack for autonomous driving, also very breakthrough. So We are leaders in creating frontier models in five domains.
Yes. Yes. It sounds like you don't create products to get a silver medal. You seem to be aiming for gold. So are you aiming for gold and will you have the best, arguably, open source model? And part B of this question: can open source catch up with frontier models and are you the person to do it? The logic, Jason, is that we will build this because, first, we can, we have the skills to do it, and because our customers need it.
Of course. For example, AlphaMayio is the world's first " thinking" autonomous car. And by " thinking" and reasoning, I mean you don't need that much data, you don't have to train on billions of hours of traffic data, because you can reason about it. Break the problem down into: “Oh, I’ve seen this before.
It's not exactly the same, but it's very similar ." Of course? So, AlphaMayio, why is this necessary? Well, there are a number of car companies. Every car in the world will become autonomous. But beyond that, every agri-tech company, every truck, every van, and most of them don't have enough scale to build this entire stack.
So I will build an extraordinary stack for them. They will adapt the last mile for their application. Now everything that moves in the future can be autonomous. If we hadn't created some biological models, the world wouldn't have them. We created the ESM 2 language model, based on proteins. ESM fold, open fold, AlphaFold 2—all this co- equivariance technology wouldn't exist if we hadn't built it.
One of my favorite projects is protein complexes, you know, next-generation protein synthesis and binding. These are innovative things. We created it. And so we will build it, because Eli Lilly needs it, Merck needs it, others need it, and they don't have the capabilities to do it, or they have n't gotten to that point yet.
So we can make a real contribution. So I do everything out of necessity. I'm not trying to undermine anyone, I mean we don't wake up in the morning with the idea of oppressing anyone. We just wake up to help everyone. Jensen, what would you say about competitive threats to your core business? Can you comment?
How cute. Yes. Well, I know that's... actually, I just want to know your opinion. What do you think about the 100 million square foot Terrafab facility that Elon announced? And if anyone can do it, it's only him. He and I flew to the same country together. And, um, along with the person who sometimes calls you.
It was a beautiful plane. And we had, you know, Elon likes to talk about these kinds of things, so we spent a lot of time talking about it. I...I... It works like that because you design chips, but you don't manufacture them? Could your chips be manufactured there, or is that... We know a lot about manufacturing technology because we're pushing the boundaries in everything.
And because we operate at such a scale, we have incredible memory technology, we are the best company in the world in this field. We have many amazing things. So your point is that you've talked about it a lot . We could just, yeah , we could talk about it , but Ilona is unstoppable, that's his superpower, and if he decides to do something, it's hard to stop him.
And so I... Can you give your opinion on where China is at with its own advanced lithography systems? They will achieve this by 2030. By 2030. Yes, and 2030 is just around the corner. Yes, and most of us will be dead by then, so does this mean for China that the switch has been flipped and all of this will immediately go to mainland manufacturing capacity?
You know, you should think of China this way: they are truly masters of mass production . And it's only a matter of time. Yes. And I, you know, I think in decades. I've been in this business for a long time, and for Nvidia, I have to think about what's going to happen in the next decade and the decade after that.
So 2 or 3 years is just a moment. This is nothing. And so, in their opinion, they are already there. They are already there. Yes. Jensen, Elon and Gwynn. We have to run. America, we must run. Yes. Speedrun. We have to organize a speed run. Slowing down is definitely the wrong strategy. Well, I mean, it's obvious to most of us in the industry that we're in the AGI moment, and that definition, of course, is being smarter than any other human.
We are already there. We're already there, aren't we? And then superintelligence is the next milestone. Given what you see, your customer base and your history here. And Jenson, I think we're already there. Do you think we have already achieved superintelligence? Yes, yes. When you take a narrow segment, like my self-driving car, I don't want it to make me an omelet.
I just want it to drive the car. Right. This is superintelligence. Superintelligence. He is better, he is better than a human. Yes, yes. Oh, yes. The accident rate is ten times lower. That's right. Yes. Protein synthesis, virtual protein screening. We are already there. Yes. Yes, yes. Do you like being on the frontier of humanity?
I like it. Yes. Ladies and gentlemen Ladies and I like it. And guys, this, this, this is great. The future is great, and we want to get there. Listen to Bright and Light. Many of us don't have to work, but I have to say, it's too good not to be a part of it. So fun. Is n't that right? So I want us all, I want to be there.
I want you all to be there with me. We will all be there. Together we will achieve tremendous success as humanity. In the meantime, we must encourage them, push them forward . They do very, very important work, as you know, and I want them to succeed. I also wish we could reduce the drama level. And most importantly, we need all of America to be with us.
Here's how we'll achieve it. Ladies and gentlemen, Jensen Huang. Thank you, friend. I appreciate it . Thank you. Thank you. Thank you. It was wonderful. It was impressive. At your party. impressive. It was wonderful. Thank you, guys. Thank you. That was great, right? It was a great time to be
Comments 0