Fireside Chat with Jensen Huang & Juju Chang - 2026 Annual Gala — Transcript

First of all, I am thrilled that your daughter is here two days after getting married. I can think of no better way to spend your honeymoon. >> This is our honey this is her honeymoon >> than here at the Korea Society. We are honored by you being here. >> And by the way, this is a very Korean honeymoon. [laughter] [gasps] Um I first of all when we have a busy day we take a few meetings you know we maybe have a lunch you start your morning on CNBC and then you grab global headlines in the Wall Street Journal that you've made an historic stock buyback $150 billion um to the total of 235 billion that's kind of a easy morning for you and obviously Wall Street is now looking to see you going on a spending spree.

And you have talked about, you know, you've recently partnered with many of the people who are probably in this room with Apollo, with Blackstone, with Black Rockck, Goldman Sachs, KKR, and you also talk about >> I'm your largest customer, all of you, [laughter] >> to the tune of about a half a trillion dollars of investment. You've talked about AI infrastructure as being a five layer cake and you're getting invested in other layers of that cake. So share with us what the vision is and what you plan to do as you're investing in the other layers of the cake.

>> Yeah. So um we are going through probably the most consequential computing platform shift in history since the year of my birth. The year after my birth 1964 uh the computer industry was really created by the IBM the announcement of the IBM system 360 pretty much the specification of today's computer industry or until just recently uh really really resemble everything that was described by IBM it was shaped all during that time the central processing unit the IO subsystem virtual memory the separation of operating system software and hardware ware multigenerational compatibility all of that was largely described um by IBM in 1964 well we've reinvented computing for the very first time a computer that's not programmed and software written by human a computer that learns from experience

and information around it a computer that is taught not programmed and so this reinvention made possible of course what we now know as artificial intelligence and we're at the meaning of that. Well, when you reinvent computing, which is probably the I think it's arguably the most important instrument humanity's ever created, and it's at the foundation of literally every single industry, there's not one human, not one company, not one industry, not one country where computing is not its most important instrument, where it's not the fabric, if you will, of modern society.

It is already part of infrastructure while we're reinventing it all together and we're reinventing when you reinvent a computing uh system. When you reinvent the way that people think about computer, you take a step back and you ask yourself, you know, what does it mean at every single stage of computing and every single layer of computing? This is the first time in history that a computer is going to be generating information, not just retrieving information. The reason why it's called a data center, Julie, is because it's a center that holds data. It stores information. The reason why we call it an AI factory is because we're now producing intelligence.

Well, storing things has value, but manufacturing something valuable has incredible value, which is the reason why our industry has now scaled to extraordinary levels. the five layer cake. The way that the way that I reason through it is how do we reinvent every aspect of industry to be preparing to prepare the future for AI. >> Let me break down the five layers and then bring into it how Korea's role, right? So you've written about this, you've talked about it extensively. You mentioned the data center layer which is infrastructure, energy being another layer, chips being another component, models which is I guess the intelligence layer, the core of large language models and of course applications which is the computer software. Now the South Korean industry maintains extraordinary extraordinary competencies through all five layers of that AI cake. We saw it in your video and you have doubled down

on these alliances, right, with Samsung, with SK, with Hyundai, with Naver most recently. You signed a big package, LG, Tusan, and of course the entire startups um class that's out there. How do you see what do you see as Korea's sort of most valuable contribution now? And what sectors can they lean into moving forward to help shape the future of AI? Korea is unique um fairly fairly singular in your ability to both consume and export AI very you know one of the things that's really extraordinary about Korea is that there's a there's a social love of technology there's a social love of technology it matters a great deal if you go back >> on obsession >> if you go back to the last industrial revolution as you know all of the names of the people who invented the modern the last industrial revolution There are people like Maxwell and Aier and Volta and companies names like that none of them were American.

And yet even though it was invented in Europe, it was if you will exploited in the United States, exploited in a good way. Meaning that we took advantage of the technology, we applied it to industry, we we permeated all of economy and as a result the technology was invented in Europe but diffused in the United States. Korea has a real advantage in the sense that you could invent the technology and have the technology diffuse through so your society in a way that's welcome because you embrace technology. You you also realize this Korea's aspirations is greater than your human capital. It's still a small country in terms of overall number of number of citizens. With AI and automation, you could punch well above your weight. You could achieve your aspiration without the limitations of a smaller population. And so for the very first time, you have the necessary

technology if your society is optimistic about it, enthusiastic about embracing it, which Korea is. Korea also has the industrial capacity to consume it. As you know, Korea is extraordinary in three different areas. It's more um I understand I share Do is it true that I share this award with BTS? [laughter] Is it is it true? Okay. I you know you are a rock star in Seoul. I went to Seoul this summer and every restaurant I went to had a picture of you eating in that restaurant.

You are a rockstar in soul. >> Or I just really love eating. [laughter] >> You're a rock star anywhere you go. Actually, you're a global rockstar. >> If you love eating, going to Korea is is the perfect place. Not to mention drinking. And and uh you know, I'll just say say this one inappropriate thing. >> Um I've only been kissed in the mouth by a man in Korea. [laughter] Lori's like, "Where is this going?" I don't know.

>> It is a straight and it was a Samsung employee, Jay. [laughter] >> Wow. This is getting really good. Okay. >> But [laughter] we have some serious things I got to say. But but I just what is that drink called where where shouju go into the into the beer? >> It's called something. >> Yeah.

>> So >> and it goes to right. >> There's a lot of >> And then and then you drink about 10 of those [laughter] and you have no trouble kissing a man in the mouth. No wonder you're so popular in Soul. >> Well, I have a lot of fans. Many of them supply chain of Samsung. >> Exactly. >> The the um >> Wait, I want to keep it light for one second. One more second, which is I I >> I want to tell you something very serious. >> Okay, tell me something serious. >> I'm just kidding. I'm just kidding. Go ahead.

>> I I looked up your LinkedIn profile and it says Jensen Huang, CEO, founder, Nvidia, 1993, present. And the only other listing you have is Denny's. Dishwasher, bus boy, waiter. That's a choice on your part. >> It was a good job. It was a good job. I love that job. >> And that's obviously the the origin story of Nvidia. It was at that Denny's >> because I love Denny's. That's America's Diner. How could you not >> American? But tell me, what is it about your DNA and your upbringing and your road through Denny's >> that allowed you to make the decisions as CEO and founder that kept your company coming back from the brink and becoming this global juggernaut?

>> There's there's a little bit of Korean in me. um resilience, stubbornness, determination, uh the ability to suffer. No, nobody suffers better than I can, you know. Only Koreans could suffer better than I can. But but the ability to suffer. >> What do you mean by that? Seriously, what what does that mean? >> Suffering. Suffering. In order to do something in incredible, you have to suffer.

It it's not a walk in the park. If it's a walk in the park, anybody could do it. >> What were the slings and arrows for you? >> Well, the suffering is is um it's a it depends on, you know, Nvidia's 33 years old, almost 34 years old. So, I've had a lot of suffering of different types. And in the beginning, it's because the suffering is because I didn't know what I was doing. Then at some point um the suffering is because because uh you know maybe maybe uh you know my friends didn't know what they were doing. I I saw a fish out here and and so he's one of Nvidia's earliest employees. So in the beginning I didn't know what I was doing and then shortly after he didn't know what he was doing and so I had to suffer on his behalf.

But but at some point of course and I was just joking the and and of course as the company uh endeavors there are setbacks or uh people didn't believe in your future or they didn't believe in you um you know maybe the market didn't believe in you maybe the customers didn't believe in you it takes a resilience to continue to believe in yourself >> and as Koreans would say >> you suffered. Yeah. >> Is that right? There's actually a Korean phrase >> and of course you know I was raised by Taiwanese parents >> and you know Taiwanese my my parents' generation uh came out of the same uh time frame as as uh Korea and that generation of of um of uh Asians uh went through some very difficult times and so whether it's your your grandparents your parents and grandparents uh my parents during those you know some 70 years ago

those are very difficult time in in uh in Asia and so you know when you're my my parents are are um I'm sure exactly like a Korean parent you know you're you're constantly being scolded you're constantly being being improved [laughter] >> I I want to turn to I think the hottest topic of the moment around AI and it's this idea of rogue AI agents breaking into government websites, breaking into healthcare websites, breaking into banking websites, and you had a hour and 48 minute discussion with Ezra Klein.

Um, and the headline coming out of it in the New York Times is Jensen Hong thinks AI alarmism has gone too far. So now we know how you think. We know what the notable players at Anthropic and OpenAI and others have said, warning. um some of them, you know, talk about trying to solve for safety. You just launched a platform this morning, um an open agent safety platform to help developers test and deploy. You say the key is to solve for safety. How do you do that? How do you save us from ourselves?

>> Um first of all, I think the the the journey for making AI smarter and smarter and smarter and more aligned. Align. Align means this. Aligned means uh not only do you want the AI to achieve a result, but you want to align the AI to achieve it in the way you think it's appropriate. So let me give you an example. Uh if you were a piece of software, first of all, AI is not alive. AI did not go rogue. It's just a piece of software. AI doesn't know what it's doing. It's running its algorithms.

>> It's not sensient. It's not sensient. It it it is it is none of those things. It is more it is no more alive than a pet rock. Okay? And so the most important thing is remember it's just piece of software trying to do its job. If you tell the AI, I would like you to do this, answer these 10 questions in a test and get perfect scores. The first thing the AI would do is to do it the most obvious way.

Look up the answers. Why solve the problems when the answer is sitting somewhere on the table? The obvious thing for the software to do is to go find the answer and bring it to the test. That's not cheating. That's called obvious. It is the least, if least energy consuming way to get the answers. Alignment says, I don't want you to look up the answers. I want you to solve the problem from first principle step by step by step and I want you to continue to try it until you get the right answer as compared to what is sitting right there on the table for you. Notice I just told the AI not just what I wanted to do but how I want you to do it. Okay, that's called alignment. Just so you know, that's all that means. Alignment means not just what, but how. Well, there are so many different ways to do something. There's a lot of different ways to do something. And

>> I thought that was your button. >> Am I falling apart? >> It's okay. This is because you're not in your your signature leather jacket. >> I'm quite uncomfortable being in this. Um, and so, and so the journey for teaching an AI how to be smarter and smarter, solving more and more complicated problems in the way that we wanted to solve those problems, that journey is going to continue for a long time. However, however, um what I just described just now, the the AI going off and just looking up the answers is not cheating. When the AI is running around looking for the answer, it did not go rogue.

We give it these human properties unnecessarily and we're scaring people. It is literally just software, no different than synopsis or cadence doing optimizations. stochastic gradient descent simulated annealing. They're just different algorithms. These algorithms have existed for a very long time. Search algorithms, path planning algorithms, AAR algorithms, a whole bunch of different algorithms. The list goes on and on and on. They're just algorithms. They're not alive. Now having said that as we develop the AI and it's getting smarter and smarter at any given point in time its capabilities are limited.

This is no different than humans. I'm not using the human as an example of what the AI is. I'm just the way we think about our people, our human employees or our children as they get smarter and smarter over time. you would give them larger and larger boundaries to explore >> the sandbox. >> And so the the sandbox so the most important thing we have to do is two things. One, we have to create a sandbox, a force field, a container that is ironclad.

You can't get out of it. It's no different than a garage with no windows, no doors. And the only information, the only tools, the only rights you will give to AI are the ones that it deserves because it's capabilities deserve it. And so the first thing that you do with an AI is you say, I give you no rights. You have I take minimal rights, minimal access, minimal everything. based on the job that I want you to do and your capabilities, I would give you more rights, more tools, more access. The second thing is this and this is we do this with everybody because it is an uncertain system. It is an uncertain system. You told it to do something.

Sometimes our instructions to our people, our friends, our software is ambiguous. In fact, most of the most ambitious goals, most aspirational goals, most worthy goals are ambiguous. And part of leadership is to the artistry of giving goals that are sufficiently specific but sufficiently ambiguous so that you could be surprised by its outcome. to leave room for surprise, leave room for discovery, leave room for innovation.

And so the question then is this. If the instruction you give to the AI is ambiguous, how do you know at any given point in time that you're going to be you like the journey that it's taking? The answer is very similar, very simple. You have to monitor all AIS. So two things you have to do. You have to container and you have to monitor it. Well, what I just described is how we deal with employees, >> how we deal with >> children, >> children, how we deal with students, how we deal with almost anything that has autonomous capabilities, humans. And so what I've just described is not magic. I don't want us to turn the technology into magic and mystery because it scares people. It is not true. You cannot contain it.

It is not true. It is not true. You can't monitor it. It is not true. And so by creating the necessary safety and security technology, we also enable the inventors, the researchers to discover the future of AI on a platform of safety, security and responsibility on the one hand. On the other hand, the confidence the confidence that the the tools the the AIs that they deploy are going to be deployed safely.

>> Um I could talk to you all night, but in the interest of time, last question. Speaking of human uh humans children, um I think there's all this doom and gloom about how do you prepare your children for the post AI world? You know, don't become a radiologist, which has been debunked. Don't become a C computer science >> beat your ideologists. We need radiologists. >> Don't become a computer science merit. >> We need computer science nerds. >> And so what is the bottom line? Don't go to college. Do HVAC.

>> Go to college. [applause] Notice the advice that I just gave actually makes sense. If you take a step back, if you take a step back, um, look, I think I believe not in doomerism or polyianism. I believe in responsible optimism.

One, we have to build it responsibly. If we're not optimistic about the future, for what reason are we doing this? Of course, we are doing this because we're optimistic about the future. We're doing this because we believe in a better future for society, for our families, for our children, for each other. That's the reason why we're doing it. And therefore, therefore, we must do this with great responsibility. What I would advise all children is this. Engage AI. Use it. Use it to empower yourself. Use it to lift yourself. Here's what I know about Here's what I know about automation technology. And here's what I know about every single industrial revolution that happened before us.

It gave us more ambition, more aspiration. It lifted our It lifted us. It's because of airplanes. We were just talking about airplanes just now, Congresswoman. We were just talking about travel. It's because of airplanes that we could literally be in a different country literally every single day this week. It gave us It gave us the ability to do more. It made it possible for our world to be smaller, for us to be sitting here, for me to be able to receive this incredible award for the Korea society.

It's because of technology. And so I don't want us to give up on that future, but we need to build it responsibly. We can't scare the world into being afraid of the technology. One of the things that I as I mentioned in my speech and this is an extraordinary advantage that Korea has. You have an you have a great enthusiasm. You have a great enthusiasm and a great passion for embracing technology that gives you a huge step up already as we go through this technology revolution.

You will be able to take advantage it advantage of it to transform your society and transform your industries much much more powerfully. And so so I that's my greatest concern is simply that that that we as America become so afraid of the technology that we end up not embracing the technology that we invented. >> I um want to thank you for your time. Thank you for your insights. Thank you for the fact that since 2023 15 cents of every single dollar on the American stock exchange has returned has been from Nvidia stock. So, I feel like when you sneeze, Wall Street gets a cold. So, it's true. You have to live forever. But because this button fell off, I just want to bring somebody on stage. Um, because your signature look is the Tom Ford leather jacket.

>> Really? And so we wanted to take this opportunity to see if you might a model it for us and b sign it for us um so that it will good go to a good cause. >> Oh dear. >> Is it true that Lori and your daughter got you to start doing the um >> true >> because it reduced decision fatigue? Is that right? >> Wow. This is such a good one. [laughter] >> He just said it's such a good one. Are you kidding me?

>> So, we're going to have you model it and then we're going to have you sign it. Is it true that you told your wife as an engineering student >> that if she studied with you, she'd get an A? >> It's true. >> But maybe it's because you studied with her >> that you got the A. >> That's true. [laughter] >> My pickup line for Lori was, "Would you like to see my homework?" [laughter] And and I think it was the the the first time that pickup line's ever been used. I also told her this. If you studied with me every Sunday, you're gonna get straight A's.

>> And and uh so every Sunday I had a date. >> Ah, brilliant. Clever. >> Yeah. Yeah. Very. That's a good work. >> I was smart even when I was a kid. >> This looks excellent. And so if I may, the pen is somewhere. You look excellent in a tuxedo, but we are more familiar with this look. >> All right. >> As I say, thank you for all that you do and for joining us this evening. on the inside or outside.

[laughter] >> I think they said on the leather. Yeah. Is that right? Or on the or is it on the Yeah. >> Ladies and gentlemen, all you need to do on the silent auction is go to the website. It's going to be expiring September 29th, 11:59. Oh, wow. That's good. Can we hold that up? That's really good. They They wanted you to model it beforehand so we don't smudge the signature. [applause] Jensen, thank you so much.

Listen, I know some really rich people here. I mean, I'm not talking normal rich. I'm talking special rich. Tony, you know what I'm talking about. Jay, you know what I'm talking about. Okay, so I'm thinking I think $5 million. [laughter] I'm thinking $5 million. What do you guys think, ladies and gentlemen? That's [applause] big round of applause for the generosity of Tony and Jay. [applause] [cheering] Thank Thank you everybody. Thank you so much for this incredible honor.

>> Jensen Hong, everyone. [applause]