Nvidia CEO Jensen Huang holds fireside chat on AI

Application end use, whether you're in healthcare, entertainment, manufacturing, self-driving cars, transportation, each one of these industries have AIs that deeply affect them. And these are the five-layer stack. At the lower level, the platform level, the reason why we say that we're an AI company that works with every single AI company in the world is because we're the platform by which we are able to work with all of these technology companies and all these application companies across all these industries. And so that is a platform that we created. The mode that people describe, not so much a mode, but basically the language by which all of these different applications and these different technologies speak to us is an architecture we invented 25 years ago called CUDA. And on top of that, CUDA library is a whole bunch of algorithms that we invented over the years. And that is basically Nvidia's platform. In the end, we don't build self-driving cars, but we work with every single self-driving car company in the world. In the end, we don't discover drugs, but we work with every single drug discovery company in the world. We are the platform by which they build their things. We're a platform company. I had half a dozen young kids come up to me today and say Nvidia gave them a better gaming experience. I mean, you all hear, before we invented the AI industry, the first industry we created is the modern gaming industry. You don't know this, I'm very proud of this. Nvidia is the world's largest gaming platform. And that's probably not known, that's how most of the kids were talking to me about that today. I want to see them, 300 million active users. A hundred million of them Nintendo Switch. And video is in that.

It was a very good headline. It was a great headline. And apparently caught a lot of attention. As you know with headlines, the disclaimer part, the foundation part was left out of the headline. But the way to think about that is, let me just handicap it right now. If you look at AI and go back to the first thing that we said, AI is a five-layer cake. Let's just simplify, it's not quite this simplistic, but let's simplify AI into a five-layer cake: energy, chips, infrastructure, models, and applications. Let's handicap it from bottom to top. At the lowest level, energy, China has twice the amount of energy we have as a nation. I want to ask about that, twice as much energy as we have as a nation, and our economy is larger than theirs, makes no sense to me. We also know that one of the most important initiatives, one of the most important policies of this administration, and it was the first thing that President Trump said to me when we met, is listen, we need to reindustrialize America. We need to unshore manufacturing again. We need to help America make things again. It's going to create jobs. That part of the economy has been offshored and completely gutted the United States. We need to bring that back and he needs my help to do so. So that entire sector of the economy is missing. However, without energy, how do we build chip plants, computer system plants, and these AI data centers we call AI factories? We're building simultaneously three different types of factories in the United States: chip factories, supercomputer factories, and AI factories. They all require energy, every single one of them. So on the one hand, we want to re-industrialize the United States. How do you do that without energy? And so the fact that we vilified energy for so long, President Trump sticking his neck out and taking it on the chin and helping the country realize that energy is necessary for our growth is one of the greatest things he's done right off the bat. So now at the energy level, back to that stack, we're 50% and they're growing straight up. We're kind of flat right now. So number one, energy. Number two, chips. We're generations ahead. We are generations ahead on chips. And I think everybody recognizes that. Number three, infrastructure. If you want to build a data center here in the United States, from breaking ground to standing up an AI supercomputer is probably about three years. They can build a hospital in a weekend. That's a real challenge. So at the infrastructure layer, their velocity of building things because there are builders, their velocity of building things is extraordinarily high. Now, really quickly on chips, we're several generations ahead. But don't be complacent. Remember, semiconductors is a manufacturing process. Anybody who thinks China can't manufacture is missing a big idea. But China discounts energy costs for a chip company by 50%. That's right. They provide free transportation for employees to come out to the factory. That's right. I mean, you can't do that. Our energy cost is more expensive than theirs in the first place and then they discount 50%. And so it's probably four to eight times the cost.

Well, before I get there, don't let me not answer that question. I'm dying to answer it, but let me handicap the next two layers. The large language model layer, the model layer. United States frontier models are unquestionably world class. We are probably six months ahead. However, out of the 1.4 million models, most of them are open source. China is well ahead, way ahead on open source. Now, the reason why open source is so important is because without open source, startups can't thrive. University researchers can't do research. You can't teach AI. Scientists can't use AI. Basically, all of the industry around the economy have no ability to fundamentally advance themselves unless you have open source. Without Linux, where would we be? Without Kubernetes, without PyTorch, all of these different types of technologies that made AI thrive are all open source. They are well ahead of us on open source. And then the layer above that, applications. If you were to do a poll of their society and ours and you ask them, is AI likely to do more good than harm? They're going to say in their case, 80% would say AI will do more good than harm. In our case, it'd be the other way around. And so that tells you something that's very important socially. Socially, we need to be careful not to describe AI in these science fiction movie ways and cause people so much concern. We want to be concerned, but we also want to be practical. AI is about automation. I think that we need to be careful not to fall behind in the application and the diffusion of AI because in the end, whoever applies the technology first and most wins that industrial revolution. As you know, electricity was invented in the UK, but the United States applied it faster and more broadly, and as a result, look where we are. And so we have to be a little mindful. Anyways, I just handicapped that stack. And I don't think it's important when you're looking at AI not to see it as a holistic thing. It's really not about ChatGPT versus DeepSeek. You have to look at it across all of the stacks and across all of the industries. Does that make sense? It's a little bit more complicated than one simple answer.

First of all, America's technology industry, just as our financial services industry, our military, our technology industry, we can all agree, are the mightiest in the world. I am part of one of the mightiest technology industries anyone in history has ever seen. We are going toe-to-toe against anyone. The American technology industry has nothing to fear. We are mighty. We're fast. We're inventive. We'll take anybody on. However, we can't concede the market to them. As you know, at the moment, Nvidia has been banned from going to China. Not to mention, China has banned Nvidia going to China. So we're the first company in history that has been banned on both sides. Whoever banned us going to China, them and China, agree that Nvidia should not go to. Now, of course, I'm being a little cute here, and I'll be a little bit more nuanced in a second, but at the moment, we're simply not competing in China. Now, what's going on? We have conceded essentially the second largest AI market, the second largest technology market in the world. I know China will, somebody has said to me, well, okay, we're not in China, but we'll grow somewhere else. You're not going to replace China. It's just as the world wanting to sell to America and they want to export to America, if they don't export to America, you're not going to replace the United States. We are singular in the world, we are absolutely singular. And so in the case of China, we shouldn't concede the entire market to them. They're formidable, but conceding that entire market, we ought to go compete for it. Having said that, we should also acknowledge that Huawei is one of the most formidable technology companies the world has ever seen. They deserve, although they have a lot of support, whatever support they have, they deserve all of the respect that everybody ought to give them. We compete with this company. They're formidable. They're agile. They move incredibly fast. We said if the United States was not in China, China's AI industry would be set back. Absolutely has not happened. As a result, their semiconductor industry has doubled, doubled, doubled. The semiconductor industry in the west around the world is growing at 20-30% per year. Growing 20-30% per year compounded versus doubling every year compounded, it doesn't take long to catch up. But they're starting on second base and we're just going up to the batter's box. They've got a real advantage, and that means we need to think about this.

Take a step back again. Remember, I think it's like nine out of the top 10 science and technology schools in the world are now in China. They lead in science and technology in many different fields. This has completely flipped in the last half to a decade. We used to lead most of them. Now they lead most of them. They have a large population of highly qualified students. Number one. Number two, 50% of the world's AI researchers are Chinese. Third, 70% of last year's AI patents were published by China. The ecosystem of AI in China is vibrant, rich, incredibly innovative. They work incredibly hard. This is a country with an enormous might. So that is the ecosystem of software developers. Now that layer, as I just mentioned, the model and the application layer, all of these scientists, they're sitting at the model and the application layer. And now they're going to take that capability because the United States is no longer participating in China. We've left China. We've evacuated that market. We've conceded that market. And so now they've got to go build their own. So using these AI researchers, all of this incredible computer scientists that they have, their richness of software capability, and they're going to go build their own complete stack. Once they build that entire complete stack, they'll export it as quickly as you could imagine. And this is the world we will find someday if we don't activate, will be buyers, not sellers.

I do agree that industrial policy should intervene when a dramatic action needs to be taken. I also would say that President Trump walked into a situation where dramatic actions needed to be taken. The first dramatic action that needed to be taken is to reverse the mistakes that have been made in energy growth over the course of the last decade. We have done our country a great disservice. There are no new industries you can grow without energy. Electricity. That's right. We need electricity because otherwise, sure, we could all be in the services industry, and as you know, the service industry only needs calories, but the manufacturing industry needs electricity. So we need energy, number one. Number two, we do need to, if we want to fix our social issues, domestic social issues, we have to create prosperity not just for people with PhDs and college degrees. We have to create prosperity for every segment of the economy. And so the largest segment of the economy is manufacturing. And we've offshored that for too long, for 20 years. We got to bring that back and we have the ability to do so. And this AI industrial revolution, this flashpoint, is precisely when we should do it because it created a company called Nvidia, made it possible for us to have such a large ecosystem of suppliers. We can encourage them to partner with us. We can encourage Taiwan to partner with us to help us reindustrialize the United States. And they've done so with great support. Taiwan really needs to have some acknowledgement for the incredible effort that they're putting in place to help us re-industrialize the United States. If not for the work that they're doing, the work that Nvidia has done in Arizona wouldn't have made the progress it has. I recently gave a congratulation speech at TSMC in Arizona. And when I looked out into the audience, it was two-thirds Taiwanese and one-third American. Arizona is practically, well, the quality of Taiwanese food in Arizona, let me just put it that way. The quality of Taiwanese food in Arizona has increased tremendously. You want to get a bowl of beef noodle soup, you're going to do just fine. But all these young families from Taiwan came to help us stand up our factories. And so we should acknowledge that. And South Korea helping us stand up our memory manufacturing. We should acknowledge that. The companies Foxconn, Wistron, Amkor, SPIL helping us set up systems manufacturing. They came from Taiwan. Really important. So the second industrial policy is to reindustrialize the United States. That I think is a fantastic initiative and I'm all behind that. I was probably the first CEO to jump behind that and take advantage of Nvidia's capabilities and this flashpoint of AI industrial revolution to help bring all of that supply chain. I committed to my customers, my partners that we're going to build in this administration, within President Trump's term, half a trillion dollars of AI supercomputers. And so that's the second part: energy growth, re-industrializing the United States. The third part that I think required a fair amount of discussion to help policymakers understand that technology leadership, American technology leadership and American national security goes hand in hand. Our nation's extraordinary technology industry is part of our national security. The fact that we have our technology all over the world that the world relies on to build their industries, their ecosystem, their economies is an advantage for the United States. It's a strength of the United States. It helps keep the United States safe when everybody works with us.

Number one, Nvidia was birthed in the United States. We are a proud American company. We're inventive. We're vibrant. We're at the center of the single most important industrial revolution in human history. This is an industrial revolution in every single way as important as electricity. We are going to impact every single industry. Every single company, every country will build it. Every company will use it. We export American technology wherever the United States would like us to export the American technology. This is an extraordinary opportunity for us to make a substantial contribution to our national security. We also know that national security and economic security and economic prosperity go hand in hand. The wealthier our nation, the more we can fund the mightiest military on the planet. And I do believe that it is the case, because of this new industrial revolution that we've started, we are creating new factories in America. We're creating new jobs in America. And somebody recently told me that we contributed more to economic growth singularly than just about any company in the world today to the American economy. And I believe that that's probably true. And the reason for that is Nvidia is a multi-hundred-billion-dollar company supporting multi-hundred-billion-dollar companies going after trillion-dollar industries. And so the economic prosperity, the technology leadership unquestionable, the economic prosperity that we can contribute to unquestionable. Now the question then becomes how do we think through the diffusion of the export of the proliferation of American technologies and standards. We should, of course, number one, safeguard our national security, the little N little S, to ensure that adversaries don't have access to sensitive technology or advanced technology that we don't need them to have access to. We should, number two, ensure that American companies, American technology companies through partnership with us have the benefit of the best and first. But then after that, after number one and number two, we should also proliferate American technology standards, compete around the world, fuel this flywheel of funding our R&D so that we can continue to be the mightiest technology industry in the world so that we can fund the mightiest military in the world. And all of that, I think, goes hand in hand.