Interview: Tae Kim, Author of “The Nvidia Way” todayheadline

Tae Kim is a senior technology writer at Barron’s and author of the new book The Nvidia Way. In this podcast, best-selling author Morgan Housel interviews Kim for a conversation about:

• The early days at Nvidia and its long path to “overnight” success.
• What drives Nvidia’s employees.
• How Jensen Huang takes long-term thinking to the extreme.

To catch full episodes of all The Motley Fool’s free podcasts, check out our podcast center. To get started investing, check out our beginner’s guide to investing in stocks. When you’re ready to invest, check out this top 10 list of stocks to buy.

A full transcript follows the video.

This video was recorded on Jan. 25, 2025

Tae Kim: Jensen was so committed that this is going to be future of computing. He was like, “No, we’re going to put CUDA on all our processors, all our gaming graphics cards and chips.” He did this for 15 years. Who does that? Most executives look out the next quarter or the next year or maybe two, three years. Jensen is looking out five, 10, 15 years.

Mary Long: I’m Mary Long, and that’s Tae Kim. He’s a senior technology writer at Barnes and author of the new book, The NVIDIA Way. Bestselling author and friend of the Fool, Morgan Housel interviewed Kim last week during our virtual member event, the AI Summit. Today, we’re playing a piece of that full conversation for listeners of Motley Fool Money. Premium Motley Fool members can catch the whole discussion on our site. I’ll drop a link in the show notes for you to do just that if you’re interested, but in the opening segment we’re playing for you now, Morgan and Tae talk about the long path to NVIDIA’s overnight success, how culture differentiates NVIDIA from its competitors, and what keeps Jensen Huang up at night.

Morgan Housel: Tae, I want to start off with what I think is so interesting to me as an outside observer about NVIDIA, which is, it seems like to a lot of people, NVIDIA just hit its stride a couple years ago, but NVIDIA is now a 32-year-old company, and I actually think in the history of business, that is very rare. Microsoft exploded right away, Apple exploded right away, Uber exploded right away, but now you have NVIDIA that’s more than 30-years-old and just exploded in the last two or three years. Maybe especially for those who are not very familiar with NVIDIA, can you take me back to the early days of how it was founded, what Jensen Huang did to create it, and what it was like before it became what it is now, one of the largest companies in the world.

Tae Kim: It’s such a great story. It’s a real life Good Will Hunting story. It’s such an amazing origin. You have three engineers, Jensen Huang, Chris Malachowsky, Curtis Priem, meeting at Denny’s in 1993 with this great idea to make 3D graphic strips. The first market was video games for PC gamers. People say NVIDIA just started on the scene three years ago, but for PC gaming nerds like me, I’ve known them for 30 years. They were making the graphics ships for graphics cards that really boosted the frame rates and made PC games take off. If you remember the early days of Doom, John Carmack created that, and then Quake was the game that Carmack made, which every PC gamer knows is the first multiplayer 3D graphics game. It was just such a great time, and NVIDIA’s revenue just took off like a rocket. Its graphics cards were the best in the business a few years into this revolution.

Those three engineers just figured out a better way to create a technology company. The most important thing, which I quickly realized after I was interviewing the entire early management team. The thing that separated NVIDIA from its competitors, which they are about 100 in its early years, was the distinct work culture that they brought and created. Even more so than technology, which has been great and they’ve been on the forefront of all these innovations, technically, I think the culture is the unique distinctive thing that makes NVIDIA great today, and is the reason why it’s been so successful.

Morgan Housel: I want to talk a little bit about those early days. We now know NVIDIA today as just a company that just has mega success, all the demand that they can possibly take in. Talk about some of the early stumbles, though, when this was not the company that we know it is today.

Tae Kim: The first two chips that NVIDIA made were complete flops. The first one was called NV1, and they built a few hundred thousand, almost 99% of the chips were returned because gamers return them. If you think about that, most companies can’t survive your first two products being total disasters. NVIDIA, because of Jensen Huang and his resilience and his charisma, he was able to convince more people to give him more money at the time, but yeah, the first two chips were complete disasters. The thing that Jensen and NVIDIA does time and time again is when they make mistakes, like the NV1, which was a jack of all trades, it wasn’t good at one thing. It was decent at five different things. Jensen realized that’s not the way you succeed in the marketplace. You have to be really good at one thing and the thing that people want, but yeah, the first two chips were terrible, and somehow NVIDIA was able to survive and be resilient.

Morgan Housel: One of the things that I think about with NVIDIA’s success in recent years as it pertains to AI, two part question. I’m just going to start with the first one here. Did Jensen and the company see AI coming, or was it right place at the right time and they happened to be making these chips that could be used for AI? I remember it wasn’t just a couple of years ago that why NVIDIA stock was surging was because you could use them in a crypto mining facility, and that was a big use case. Then AI just exploded what seems like out of the middle of nowhere when ChatGPT launched about two years ago. Did Jensen see AI coming and prepare for it or was it right place at the right time?

Tae Kim: I think it’s actually both. They were at the right place at the right time because Jensen always believed in parallel computing, which is what their chips are good at, it’s called graphics processing units, GPUs. Basically, the way it works is that GPUs have thousands and thousands of cores, and they split computing workloads into smaller jobs, and they run on those cores at the same time. That compared to traditional processors, which are called CPUs, they might have 4-8 cores, and they do computation workload serially. He always thought parallel computing and GPUs would be the future of computing, and he worked in NVIDIA to innovate and do the technical background for that. They’ve been working on that with CUDA and all that from 2007, 2008, but then when AI took off around 2012, 2013, when a couple of teams figured out you could use GPUs, and it really accelerated the workloads and made the image recognition software work, Jensen saw that and said, this is a game changer, I need to put resources behind it.

He put tons of employees to write software libraries to accelerate the AI stuff, and he invested for like 10 years before ChatGPT came in. It’s a combination of both. They were in the right position with dominating the parallel computing processing technology, but then he saw deep learning and neural networks really take off, that if you add more computation, if you add the algorithms and the libraries, that using GPUs can make AI work 100 times faster than CPUs. Then he told NVIDIA and all his employees that we need to invest in this. Even though some of executives told him, “Oh, this is a fad, this isn’t going to work,” Jensen was like, “No, this is going to be real.” But then again, this is 2012 and ChatGPT is 2022. You’re talking about a 10-year run where they kept investing even though the revenue wasn’t really coming in, wasn’t really taking off. That’s a genius behind Jensen. I actually compare him a lot to Reed Hastings. Reed Hastings had this intuitive sense that someday we’re going to stream video over the Internet, and that was going to be the future, but consumers didn’t have the broadband infrastructure.

The technology wasn’t ready. Reed just basically stuck around, positioned Netflix to be on top of the technology as improved, but he did the DVD by rental business and used that as the revenue to come in, but when technology got ready, Netflix was all over it and was ahead of everyone else. Jensen and NVIDIA have done this time and time again. Every big competing shift in graphics from programmable GPUs to CUDA to data center, full-SAC solutions, he was ahead of the game, he was there early, and then when it took off, he was there. The ChatGPT moment in late 2022, the key architecture there is a transformer architecture. It came out in this famous paper that everyone knows that Google’s engineers wrote. Jensen saw that paper and was like, “This is going to be a game changer. We need to accelerate that and create these tensor cores that make these workloads and transformer models work better.” Actually he saw how important that was. When the Hopper GPU came out a month before ChatGPT was launched in November of 2022, the Hopper GPU that NVIDIA made had a thing called the transformer engine on it already, which is a combination of hardware and software.

They were preparing for that ChatGPT moment like a year or two years beforehand, even before it happened. I remember this Ian Buck, who’s considered the father of CUDA, he had this podcast, which I recommend everyone listen to and NVIDIA has AI podcast, I think it’s just called AI Podcast. While I was researching the book in 2019, he did this podcast interview for NVIDIA’s AI podcast, and he was talking about natural language processing, the need for AI models to look into knowledge repositories. All that stuff that became ChatGPT three years later, he was talking about in 2019. NVIDIA was all over this technology. We can talk about Mellanox which is going to go down as one of the biggest successful acquisitions of all time. Jensen, when he bought Mellanox in 2019, he put in the press release, “Someday, AI workloads are going to be scaled across thousands and tens of thousands of AI GPUs made by us, and Mellanox is going to be the key technology behind how these AI server clusters work,” which is exactly what’s happening right now. This year, you have 100,000 NVGPUs and Elon Musk’s xAI cluster.

Mark Zuckerberg has talked about they have a cluster that’s even larger than that over 100,000. Now Broadcom is talking about in two or three years, you’re going to have one million GPU clusters out there. Jensen has been able to see the future, know how the technology is going to develop years beforehand, and he positions NVIDIA to invest in the technology before it happens. Time and time again, it’s unbelievable how he’s been able to predict the future.

Morgan Housel: There’s a book that Bill Gates wrote in 1994, I think, it’s called the Road Ahead, where he talks about what the Internet would be like. This is 1994, so most people don’t even know what the Internet is yet. I read it a couple years ago, I went back and scanned it. I was expecting to find all the areas that he was wrong, all the false predictions he made. It’s astounding when you read it because he got virtually everything right. In this book in 1994, he perfectly explains what would become Netflix, Facebook, Google Maps. He saw all of that coming. It’s astounding to watch someone who can truly see the next wave of where it is a decade before it happens and a decade before everyone else, and it sounds like that’s what Jensen was as well, but how rare was it? How crazy was it for Jensen to be talking about and focusing on and investing a ton of money in AI 10 or 15 years ago?

Tae Kim: I agree Bill Gates was very pressing about that, but I think the difference with Jensen is he was able to predict the future and then commercialize the technology to win and actually do well. When CUDA came out, which is the critical technology and the platform that enables to parallel computing on NVIDIA’s processors, when CUDA came out in 2006, and they’re investing in the hardware circuitry, Wall Street was like, what the hell is this? It’s not really generating revenue, you’re dedicating die space on the chip toward this CUDA thing that it doesn’t seem like people are accepting. Jensen’s like, “This is the future, I believe in it, and I’m going to do it.” Gross margins plummeted from like 45%-35%, and the stock got punished. The stock was down 80%. People inside NVIDIA, employees were grumbling as why are we doing this CUDA thing? It’s killing our profits. It’s not generating anything, but Jensen was so committed that this is going to be future computing. He was like, “No, we’re going to put CUDA and all our processors, all our gaming graphics cards and chips.” He did this for like 15 years. Who does that? Most executives look out the next quarter or the next year or maybe two, three years. Jensen is looking out 5, 10, 15 years. This isn’t the only example of this. There’s this thing called rate tracing, which does better lighting effects. Jensen invested in this technology for 10 years before it was commercialized on gaming GPUs. Same thing with DLSS. He invented this in the meeting, which is one of those scoops in my book.

Literally, one of his research scientists came to him with this DLAA, anti-aliasing technology, and he’s like, “How about you take some of that transpose it this way?” That will be something that’s very interesting. The research scientist is like, “Maybe that will work. Let me look into it.” Then a few weeks later, “I think we can do this.” This DLSS technology, which basically uses AI to upscale graphics into higher resolutions, they worked on the frame generation version of that for seven years. No other company in my knowledge does that. They invest in stuff that becomes commercialized seven, 10 years later. Xerox PARC is legendary for creating the technology behind the GU to GUI, Graphic User Interface for operating systems and all this stuff, but they never commercialize it. Steve Jobs with the McIntosh, they were able to commercialize it and benefit and take that to market. NVIDIA does the research, commercializes it, and wins five, 10 years later. I think that’s a key differentiator for NVIDIA. They have this long term thinking mentality. Jensen is so paranoid about getting disrupted because he’s a student of history and sees how every major technology company in history gets disrupted by innovation and start-ups. He’s willing to disrupt himself and cannibalize himself at all times. He knows, if you don’t invest, you’re dead in technology, eventually. That’s why he’s willing to invest longer term than anyone else.

Morgan Housel: I want to talk a little bit more about Jensen, who is obviously now the public face, the founder and CEO of NVIDIA. So many of the big, dominant, very successful CEOs in history, whether it’s Henry Ford or Bill Gates or Elon Musk or Steve Jobs, a lot of their business success comes at the expense of their personality and their friendliness. A lot of reason they are so successful is because they are such demanding bosses. You’ve met Jensen, I’m just curious, what is he like in real life when you meet him? What’s his personality like?

Tae Kim: I think this is a key characteristic of personality and the key reason why he’s successful. He’s very blunt and direct. I was asking him questions about his culture and different questions here and there. He was like, “Tae, you’re not doing a good job now. You’re not getting NVIDIA right now.” I’m just taking it back. Most people are nice to each other, they’re friendly. No, Jensen wants you to know what he’s thinking at all times, and he’s blunt and direct to me, he’s blunt and direct to his employees. He doesn’t believe in coddling. I think all of us have been in large organizations where the HR, you have to be professional. If you have a problem with someone, you should take them aside and gently tell them and be constructive. No, he says, “Why should Tae be the only one that gets to learning here? Everyone in the room, more employees across NVIDIA should learn what the problem is so that we could all improve together.” There’s this one story in the book where a manager of the Tegra 3 Project, which is a chip, was late, and he was having problems sticking to his schedule. He dressed him down in front of the entire company, and he got the cameraman to go into his face and says, “Rayfield, you need to get Tegra 3 back on schedule.

This is not how you run a business.” Multiple NVIDIA employees told me, this is the most humiliating thing I’ve ever seen, but Jensen’s thinking is, no, you need to know exactly what you’re doing wrong, how you can learn from it, and I don’t care about your feelings. Your feelings doesn’t help the company move forward. That blunt and directness, they call it intellectual honesty. Be willing to say the unpleasant truth. By doing that, he tells people, you’re not doing a good job. This analysis is not good enough.

You get embarrassed, your feelings are hurt, but next time, those employees are going to work on a different level to make sure they don’t get dressed down in front of everyone again. It’s a strategy that works. I’m sure, NVIDIA employees, it’s a hard driving culture, it doesn’t feel good in the morning, but it develops this winning culture. The track record speaks for itself. Steve Jobs was like this, I’m sure Bill Gates was like this. I don’t think he does it to be mean. People say, “You can’t take it personally because that’s just the way Jensen is,” and I think it works. There’s no question about it.

Morgan Housel: It’s such a fine line for this between a hard driving, motivational leader and just being a jerk. It’s actually a very fine line. I know a lot of people historically, of course, it was an honor and a privilege to work for Steve Jobs back in the day, but incredibly demanding and incredibly difficult and incredibly stressful. What is it about Jensen? We’ve detailed the hard driving, blunt side of him. What is it about his characteristic and his personality that is so motivating to all these employees that are willing to put up with that?

Tae Kim: I think one reason why there’s near universal praise when I talk to former NVIDIA employees is that they know that Jensen does this a level of excellence expectation on himself. He works more than anyone in NVIDIA. In the early days, he was going from 9:00 to midnight. He considers work relaxing. He says, “I might not be the smartest person, but I know no one’s going to outwork me.” He works all day, Saturday, Sunday. I have a funny story in the book where he’s drinking a Scotch on Sunday evenings, reading employee emails. People don’t want to send him emails on a Friday afternoon because they know he’s going to reply back and ask for more work to be done, and it’s going to ruin their weekend. I think that expectation where they know Jensen is working just as hard as they are alleviates that thing where they might get annoyed or whatever. The other thing is, this is such a competitive industry.

Chip companies die all the time. They go bankrupt. No one wants to spend three, four years or lives, working on a chip project that becomes nothing and the company goes bankrupt. You want the work you do every day, you work like crazy hours. You want it to be successful in the market and have an impact on the world. NVIDIA, you’re going to do that. You’re going to work with some of the best and smartest people, and you know the work you do is going to probably win in the marketplace and have an impact on the world. Winners want to stay on the winning team, and winning begets winning, and that’s why people stay at NVIDIA. I think the biggest evidence of that is the turnover rate, NVIDIA’s 3%, which is unheard of. Most technology companies are 13-15%, probably 17% at some chip companies. The reason is, people want to stay at a winning culture in a winning company, and obviously, financially, this is an extraordinary outcome. Literally, NVIDIA is the best performing stock in US history with any company with 20-year returns. If you put $10,000 on the first day of its IPO in 1999, that holding would be worth $40 million today. Financial rewards, the meritocracy are obvious attractors of employees to NVIDIA.

Morgan Housel: I think there are plenty of Motley Fool investors who maybe didn’t get on the IPO day, but got it more than a decade ago and have done quite well. I want to read to you a quote that I heard from Jensen in an interview that gets to his motivation and his drive and his work ethic. He was asked what companies keep him up at night, and his answer was, “That shouldn’t keep me up at night because I should make sure that I am sufficiently exhausted from working that nobody can keep me up at night. That’s really the only thing that I can control.” It’s one of those quotes that you read, and I just think, I would hate to compete against that guy who just has, there’s nothing else in his life going on except his work, and he’s willing to work 24 hours a day to get there. Especially now that he is personally worth over $100 billion, he’s one of the top 10 richest people in the world. What motivates Jensen?

Tae Kim: Money is nice, but the whole executive team, a lot of them have been there for 20, 25, 30 years, and there really isn’t much turnover. All of them buy in into thinking that we are changing the world, and more so than never, this AI revolution is going to change the world, and drag discovery. I’ve had multiple executives tell me they’re so excited about what the AIGPUs are going to mean for healthcare and the capacity and ability to do all the simulations for drug permutations and possibly cure cancer. These are life changing, humanity changing possibilities, and that motivates people beyond the money. I think with Jensen, from the beginning, I actually compare him to Michael Jordan. Michael Jordan has this thing where he would create chips on a shoulder and take things personally because he had this extreme competitive desire to win at all costs. One story that shows a similar extreme competitiveness is Jensen had this early CFO, and the CFO was a grand master at chess. He was one of the top 50 chess players in the country, and Jensen knew this, but he had to beat him. He would learn all the chess openings and ask Jeffrey Bar to play chess. Invariably, Jensen would lose every time because Jeff is a chess master and he knows how to beat a novice.

Then after he lost, Jensen would knock all the chess pieces off the board and then force Jeff to play him in table tennis, where Jensen used to be a table tennis champion when he was a teenager, and he would beat him at table tennis. It would fulfill his desire to win after he lost. That’s the mentality that Jensen is. He needs to win. He’s extremely competitive. If you have that mentality where you need to win, you have a work ethic that’s better than anyone else, and exactly what you said. Multiple NVIDIA employees told me they would never want to join AMD or Intel because they would have to compete against Jensen, and they know that work ethic, that competitiveness, it’ll be almost impossible to beat him.

It’s not just the work ethic, it’s the business genius, which time and time again he creates this amazing business strategies, but he combines that with a deep technical understanding, which I think separates him. One chapter in the book, I go off on all these technology CEOs that have backgrounds in marketing or business management, MBAs. Those people can’t compete with someone like Jensen who reads AI papers as soon as they’re published. He trades the latest trends and articles with all his top AI engineers. The combination of that deep technical expertise and the business genius is what makes Jensen special and successful.

Morgan Housel: I want to follow up on that point that you just said, because I think historically, there are a lot of good technology minds, there are a lot of very talented MBAs, it’s pretty rare that you get a technological genius and a business titan in the same mind. Jeff Bezos was that, I think Elon Musk is that, I think Jensen is that. Talk about some of the business moves that NVIDIA has made. We’ve been talking about their technology and seeing AI 10 years ahead, but talk about some of the business moves they’ve made that have made the company so successful.

Tae Kim: The biggest one that I actually think is happening right now in AI, too, in the late ’90s, he had a meeting with his top executives, and it was like, how come all these 3D graphics companies, these graphic chips companies in PC market, how come they always lose their leadership? Every two years, there’s a new 3D graphics company on top of the heap and doing well. He basically figured out that PC makers decided to pick their graphics vendor every six months, two seasons of the year, back to school and in the spring. Then he thought, most graphics companies, when they created a new graphic chip, it took 18 months to create a new graphic chip, and then they start again, and it takes another 18 months, basically around two years.

The problem was, you would be inside a gateway or car park or whatever, but then in the next two years, another graphics chip company would come out with a better product that was faster. He was like, how do I fix this? How do I keep NVIDIA on top? Instead of 18th month cycle, he decided, why don’t we just do a graphic chip every six months? Who does that? Everyone’s like, “No, it takes two years, it takes 18 months.” He’s like, “We’re going to make three graphics chips in the times it makes one graphics chip.” It wasn’t that easy. They made one 18th month architecture and then created two faster derivatives every six months, but he had the genius to be like, we’re just going to accelerate the product cycle.

If you do that, competitors can’t compete, and all these other graphics chip companies went bankrupt after NVIDIA decided to make a new graphics chip every six months. He’s doing the exact same thing with AI GPUs now. Typically, it took two years to create an AI GPU architecture. Then last October, he’s like, “No, forget two years. We’re going to do this video game playbook we did in the 1990s. We’re going to make a new AI GPU every 12 months.” Think about if you’re AMD. How do you keep up with that? NVIDIA already makes a GPU that’s 5-30 times faster every two years, and now they’re going to do every year. If you’re a competitor, it’s nearly impossible to keep up with that, and that’s just Jensen’s mentality.

It’s one of their key philosophies. It’s called speed of light. We don’t compare ourselves to other competitors, we don’t compare ourselves to what we did last time. If you do that as an employee and tell Jensen, “Oh, I did 10% better than last time. I did 10% better than other competitor.” He will yell at you and dress you down and say, “We don’t do that. We do it compared to what’s physically possible by the law of physics. If you broke down the task by each component with no lag, no queue, what’s physically possible? What’s the best time you can do based on the law of physics? If you do that, if NVIDIA does that, that’s literally the most maximum speed you can do in reality, and if you’re a competitor, you can’t beat that.

Mary Long: As always, people on the program may have interest in the stocks they talk about, and the Motley Fool may have formal recommendations for or against, so don’t buy or sell stocks based solely on what you hear. All personal finance content follows Motley Fool editorial standards and are not approved by advertisers. The Motley Fool only picks products that it would personally recommend to friends like you. I’m Mary Long. Thanks for listening. We’ll be back on Monday.