YouSaid · the spoken record

Tae Kim

lines on the record: 59
first: 2025-03-07
most recent: 2025-03-07
sittings or episodes: 1
sources: podcast

Every line below is reproduced as it was said and linked to the record it came from. Nothing here is summarised or generated. Directory · Search · Corrections

“I'm first adopter on X. The book is available everywhere, Amazon, local bookstores. I also have a website at take him.com T-A-E-K-I-M.com. So I would love to hear from you guys. And thanks so much for having me, Clay”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“It's kind of amazing. The turnover rate is 3% in industry that's usually 15%. And the last part of the NVA is just extreme velocity of how things get done. People just move fast in the best way possible and beat their rivals. The speed of light, like just get stuff done quickly. Don't get bogged down. You can make mistakes, but learn from your mistakes. So those are three main components of what I term the NVIDIA.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“I think the first part of the NVIDIA is just extreme work ethic. There are no shortcuts. If you outwork your competitor or you outwork your rival, it's going to give you a huge edge. And Jensen just talks about this all the time. No one is going to outwork me. Someone might be smarter than me, but no one's going to outwork me. So if you work really hard, it's going to give you an edge. The second part is just talent cultivation. Jensen knows the way NVIDIA is going to beat this competition is by having and hiring and retaining the best talent. So he looks at stock allocation like his blood from human resources executive told me. Like if he sees an engineer that's a rock star that's doing a great job, that is adding a ton of value, he'll double the sock rent on the spot. And this kind of meritocracy and this culture of winning and just retaining people, there's so many NVIDIA's executives that are there 25, 30 years.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“To serve that demand, and we're going to actually increase what we spend later this year. So all these narratives that people kind of freaked out about just simply aren't true. And it shows by from what the large technology companies are saying with AI demand and the capacity they need going forward.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“So it's just a fact of computing history where the cost of computing will come down and developers will find a way to use that extra computing power to make new AI applications. It will drive AI adoption. There's all the things you can do with extra compute power. And it creates innovation. It sparks innovation. So the second part where people started freaking out that maybe we don't need AI compute power anymore because DeepSeek is more efficient than other models. That doesn't make any sense either because this has been happening for the last few decades. And this is just a continuation of what's going to happen. And what happened is after all this chaos with deep Zeek, every major technology company in the US erased their guidance in terms of their AI infrastructure's capital spending this year. I mean, Google, Meta, Amazon, all of them said that AI demand is off the charts. We don't have enough capacity.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“With the misleading narrative that China somehow figured out a way to recreate open AI with five, six million dollars, which is not even what the paper says. So that's the first part of it. The second part of it, which is actually true, is that deep-seek model is very optimized and efficient. Like compared to the OpenAI model, it's probably used as 90% less compute resources. So it is very efficient in the use of compute. And the way it did that, it kind of combined all these AI innovations that other AI startups and companies have already done in the past and combined it in a very smart way, a novel way, mixture of experts, number precision, all these things, right? But if you just step back, that compute efficiency 90% is completely normal. It's something Jensen talks about in his speeches all the time, how Nvidia and industry has driven down the cost of computing by a thousand times. And he hopes another thousand times will happen over the next decade.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“And then you read this next line and it says this does not incorporate any employee costs, the pre-training costs, all the research and experiments we did before. It was just a final training run. But all the research, all the employees, all the infrastructure costs to lead you to their final perfect training run was not included. So the actual cost was like an order of magnitude higher. The week before all this stuff hit the wires, the deep seek CEO met with the Chinese premier and said, we need more GPUs. If you've magically figured out a way to create cutting-edge models for $5 to $6 million, why would you go to the Chinese premier and say you need more GPUs? It doesn't make sense. And a few days before this deep-sea craziness, the Chinese government announced they're going to invest $140 billion in AI over the next few years. So all these things happen at once. Everyone just went off.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Were a couple of narratives that came out when Deep Zeke kind of first hit the mainstream media. The first narrative was somehow China magically found a way to create a cutting edge model without using a lot of resources. The number everyone threw around was about $5 to $6 million. They figure out, make a model that's almost as good as OpenAI's best model for $5 to $6 million. And that freaked everyone out like, oh my gosh. Like China figured out some magic alchemy. If you actually looked at where that figure came from, it was from a paper on December 26th. It wasn't even the last few weeks. It was like a long time ago. And that $5 to $6 million figure was just on the final training run, on a final theoretical training run if they rented that 2000 GPUs from a cloud provider. It wasn't even real. It was a theoretical rental cost if they did a final perfect training run for that model.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“And if you do FSVO light, your competitor can't beat you, right? If you're doing it at the fastest way possible, your competitor can't beat the laws of physics. So if you have that mentality on everything, right, instead of two years to make an AI GPU, we're going to do it every year. And we're going to figure out a way to make it every year. Your competitors can't compete. And that's what NVIDIA does. That extreme velocity of getting things done at the speed of light is like another huge competitive advantage for NVIDIA.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Most companies, if you say, oh, I did this 10% faster than the competitor, or 10% faster than I did last time. Yay, props to you. You did a good job. If you did that NVIDIA, you would get dressed down, they yelled at because they don't care about how you did first last time. They don't care about how you're doing versus your competitor. They want to know how you're doing versus the physical limits of reality, right? So you break down this process into the component parts. You take out all the possibilities in terms of lag and downtime between each step. And then tell me how close are you to the ultimate speed run only held back by the laws of physics. So if you do stuff at the speed of light, like if you do make a product at the speed of light, you're doing it at the absolute maximum possible, the fastest way possible, because you're taking away all the slowdown, all the cues, all the downtime.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Where he, the executives in the room, including junior employees, they hash it out and make the decision this is what we're going to do. Or it's a problem solving meeting. So they're working on a project. He brings all the top players that are working on that project and they go through the biggest problem, then the second biggest problem, then the third biggest problem. We need to fix that. We're not leaving here until we figure out a way to fix the biggest problem on that project. So these meetings and things that how things are done in NVIDIA, it's just action-oriented, it's process oriented. It's getting to an end result. The other thing, besides mission as the boss, which I already discussed, the second biggest catchphrase that NVIDIA that talks about their culture is speed of light. Speed of light is a mentality of everything NVIDIA has to be done at the fastest way possible. Inequality way, but the fastest way possible.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Executives. One of the AI executives I talked to when I interviewed him, he's like, Yeah, I got 13 emails from Jensen today. Like, he's constantly emailing people left and right up and down all day long, right? So that email culture where he's constantly has his finger on the pulse of what's going on is how he manages NVIDIA. And I don't think most companies are like that. I don't think most corporate CEOs are emailing a dozen emails per executive a day, especially on the top projects. Like that is how he manages NVIDIA just so closely better than anyone else. And then he does these meetings. Like every meeting at NVIDIA is live, right? I've been in big companies. A lot of meetings are completely worthless. There's a lot, like I said, there's a lot of indecision. There's a lot of, you know, people just talking for the sake of talking. It's a complete waste of time. Nvidia, that doesn't happen. It's either a decision meeting.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“So, the board members NVIDIA told me every time a new board member comes, the first thing they say is this doesn't make any sense. Jensen needs a chief operating officer. And then Jensen gets all upset. It's like, no, I don't. This is the way I do things and it's more effective. And like I said before, he just wants his hands and everything. Like he wants to know what's happening everywhere. So that's why he talks, he has 60 direct reports underneath him. One thing that he doesn't do is coddle and do career coaching. So a lot of CEOs meet one-on-one to talk about their careers or how things are going, just a lot of hand holding. He doesn't do any of that. So that aids them a lot of time. The other part of it is just there's this crazy email culture in Nvidia where he's constantly peppering all his executives with emails. What are you doing here? What are you doing that? What do you think about this? Here's a paper. Do this. Do that. He's doing that with all six.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Think right now they're doing what they've been doing like the whole accelerated computing, making all the GPU parallel processing attack all these different non-graphics and non-gaming problems, they're still on that curve up. I mean, the two things I talk about are drug discovery and robotics. These are very early technologies. It really is a computation problem. It's a computation simulation problem. And if we're able to put a thousand times more AI processing power to these problems and simulate how all the drugs and proteins interact with each other, you can literally solve what drug molecule will be most effective at treating this disease or cancer. And I think this is something that really drives NVIDIA employees today. The stakes are so much higher than making nice computer graphics on a screen or playing a video game. And I think they're really excited and passionate about what can happen over and over.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Work with audio and video, just like the AI models are so good with text today. So all these AI innovations are just ramping at the same time. And all that requires mass amount of compute from GPUs. And NVIDIA is really the only one that has these scalable systems that can meet that demand.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“They don't have a 72 GPU AI server that has all the interconnects, all the networking, everything optimized. And then you could stack a row of these together and build out 100,000 GPU cluster, right? So the competitors don't have CUDA. They don't have all the optimized libraries and all the bugs ironed out after 10 years of battle testing. They don't have just a networking hardware and software expertise all optimized and combined at once. So that's why NVIDIA has one and is going to keep winning. It just has the resources and all that foundation built up over the last 10 plus years. It just looks, you know, as long as the AI training laws and scaling laws hold and all these new innovations that require even more compute, I've talked about reasoning models, AI agents, multimodal models that can work.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Yeah, so like I said, the CUDA ecosystem is just so ingrained in all the developers that it's really hard to shake that out. So Broadcom, Marvell, there's all these custom chip programs inside these big hyperscalers. Jensen has said, you know, we'll see what they make in two, three years. That's what they're developing now. But if you're spending a few hundred million dollars doing R&D, NVIDIA is spending $10 billion in R&D. Like it's at a different scale. Nvidia has the networking expertise from the Melanox acquisition in 2019. So right now, in the past, even just a year ago, a typical AI server would have eight GPUs. It was a hopper AI server. Now the current blackwell AI server that's shipping right now is 72 GPUs in the same space. It's like one rack. It weighs one and a half tons. Like this is why AMD can't compete.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“How amazing it is to go from 5 to 30 billion in six quarters. It looks like it's going to be even larger over the next four quarters. So that may report where they gave guidance for the next quarter $4 billion above expectations was the thing that set everything off. And the whole AI arms race was on. I think the next day, the stock went up like $180 billion in market value. It's like one of the biggest one day gains ever. That one day gain alone was bigger than Intel's market cap. So it was a big event and it's going to be looked back on history as a big deal.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“It looks kind of like a gun that went off, and everyone had to play ball and invest. That's exactly what happened. I mean, literally, like I said, NVIDIA's data center revenue went from like $5 billion to over $30 billion in six quarters. I mean, that's like one of the largest kind of technology ramps in history. And this is not like Google or Facebook or software company where you could just copy paste the bits. The software is easy to scale, right? If you have demand for 100 billion of your product, you could just sell your software. It's no big deal. This is hardware. This is stuff that has 35,000 parts. So NVIDIA had to like go crazy with its Taiwan suppliers, talk to them. What do you need to meet this tidal wave of the demand that we're seeing in that pipeline? So it's literally one of the biggest kind of accomplishments in technology industry history. I don't think people really recognize.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“And he just titled his note The Big Bang. This was the Big Bang. And it was off to the races. And it kind of reminds me of Netscape when IPO'd back in 1997. And the stock doubled and tripled the first day. And everyone's like, wow, this is real. This internet thing is going to be huge. And all these internet startups start going public. This was the same effect. So every company was like, oh my gosh, this is a game changer. The whole financial Wall Street industry just opened their eyes saying this is going to be huge. And almost like every company also had to realize, holy crap, every other company is investing in this stuff. This is an existential threat for me. If my competitor incorporates these AI models and has better customer service, is able to figure out better product or better service, I might get disrupted. My competitor might destroy me. So it just”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“So, ChatGPT came out in November or October. And it didn't really affect NVIDIA's revenues for a couple quarters. People were talking about it. Oh, it might help NVIDIA. Maybe it will help a little bit. They reported a quarter. It wasn't anything special, but it was like inline. And then they reported their financial quarter in May of 2023. So this is about six or seven months after ChatGPT first came out. Like I had this thing where a trader is waiting for these results and the Bloomberg headline goes across the screen. And he blinks. He's like, that can't be right. That can't be right. So NVIDIA's, what the street was forecast was like 7 billion. And NVIDIA said, we're going to do 11 billion. Like it was insane. People are just like stunned. Like it just blew everyone's heads off. I quote all these fun managers and they couldn't believe the number. It was like the biggest number they've ever seen. Literally the largest upside, like you said. There's this chip analyst at Bernstein Clustace.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“About all these things, the natural language processing, AI model scaling laws, he actually sees exactly what happens two, three, four years later when the transform architecture paper came out from Google, Jensen and the Nvidia team is all over it. They're like, this is a big deal. This is going to change everything. And they actually built in a transform engine in the Hopper GPU that came out a month before ChatGPT was released. So they're like, these guys are super technical nerds. They're on top of all the technology trends and papers. And they position NVIDIA to be there years before it actually happens. And I think exact timing is impossible to predict, but the technology stuff, like they're on top of it. And they've proven that time and time again.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Stay around and keep investing and be the first guy there when the market was ready to go to internet video streaming. And Jensen does that time and time again, like programmable GPUs, CUDA, the steel for Melanox, which gave NVIDIA the ability to build out these massive AI data center scale networking GPU clusters. That's exactly what's happening today with these 100,000 GPU clusters that are being built going to 500,000 to 1 million. Like all that was enabled by Jensen seeing what's going to happen someday and then positioning NVIDIA to be perfectly positioned. to take advantage of it. So it's a combination of both. He sees the end goal, but he's willing to stick around and stay there until the end state happens. The other thing is like chat GPT took off in late 2022. I talk about this podcast where the head of CUDA, Ian Buck in 2019,”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“I mean, I would push back a little bit. Like, he is a fortune teller in the sense that he sees the end thing happening. But he doesn't know the exact timing. So even like the 2013, the Big Bet on AI, it took another nine years for it to really take off where the data center revenue in the last six quarters went from like $5 billion to $30 billion. But he just foresees what's going to happen. And when he's very confident that that's going to be the end state, he's willing to keep investing for five, ten, fifteen years, ray tracing, DLSS, and obviously AI are examples of this. And I actually compared him to Retastings, like Read Hastings intuitively knew that video would go to internet streaming someday. And Wendy knew this, like the technology wasn't ready. American households didn't have broadband. Everyone was on dial-up. But he knew he could invest, staying on top of the technology, do a DVD rental business by mail.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Line about how hyperscaler is going to make their own AI chip and that a startup is going to come with a better chip that's going to be faster than NVIDIA. And I always roll my eyes because Amazon and Google have been doing this for five, ten years. And then the Amazon Web Services CEO goes on television and says, yeah, NVIDIA still has 95% market share. You've been doing this for five, ten years. And the reason is what I just said. It's just NVIDIA makes the best performing chips. Everyone builds their stuff on NVIDIA, so there's no point in switching and risking your entire business by switching to a different platform. And it's going to continue that way until we go into completely new computing paradigm, which is it's going to move from GPUs to whatever, like it could be quantum computing or whatever. And then that gives an opportunity for another company to be the dominant player in that computing paradigm.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“In the past, you could use going forward. So, a combination of these things there's a technical risk, NVIDIA chips are usually that fastest performing, highest performing chip. And you don't want to rely on another vendor that is either going to abandon you, which Google has done a million times. You're going to build your business on Google platform that might not have a track record of sticking with you in the end, or a startup that might go bankrupt in two years. Like you said, the safer thing is to stick with NVIDIA. They're going to be around. The number one player, and you're not going to have the technical issues that you had because all the bugs and stuff have been optimized from like 15 years of work of these millions of developers going on stack overflow, sharing their trips and tricks. And Nvidia software engineers kind of ironing out and making sure everything works in the best way possible. So that's why every other week there's a headline.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Cerebus, which is an AI chip startup. The other problem is when I talk to these AI startups is it's not easy. There's always technical problems when you port your software from one chip platform to another chip platform. So back when the Mac was running on PowerPC and Windows was running on Intel, porting a Windows application to PowerPC on the Mac was a huge endeavor. And there's always problems after you do it. So that's what the ecosystem, that's what the mode is, right? People aren't going to be inclined to switch over to a chip that might be 30% cheaper because why risk your business on doing that? It's not worth it. I'd rather focus on making the software or making my AI model better. Plus, NVIDIA is making a new chip in 12 months that's going to be much faster. And it's going to be backwards compatible. So every piece of software that you wrote.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“The beginning, they would run these sessions where they invite all their customers and they would stay and talk to all the CUDA engineers and say, what do you need? What can I do to make your life easier? And they would take all the input and then make libraries or make the hardware circuits to be optimized to run their software faster. So this constant process every year and ironing out the bugs is also a huge unlock here. You just develop an ecosystem where all the developers rely on the libraries so they build all their applications on top of the CUDA libraries. And all they do is learn how to program in CUDA. So you have these millions of developers who learn CUDA, rely on all these libraries and it just creates that kind of ecosystem where it's really hard for a developer that learned on CUDA and has built all their programs on top of CUDA to switch and port that to say AMD Rockum now or”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“So, this actually plays into Jensen's kind of long term thinking. Most companies are looking out to next quarter or the next year. Jensen just wants to create the perfect platform, the perfect ecosystem possible for the long-term development system. So what he did was create these math libraries for every single vertical, whether it be science or industrial, medicine. He wants NVIDIA engineers to talk to all the customers and developers in each sector and say, what do you need? How can I make your life easier? So by creating all these hundreds of libraries, whether it be like ray tracing for Hollywood animators to the math libraries I talked about, it just relieves a lot of programming burden for these developers. So they could just focus on fixing the problem instead of creating these things that just take a lot of work and effort that aren't focused on fixing the problem. So he did that.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Thermodynamics of clouds, figuring out stock options pricing, whatever they want to do, Jensen wanted NVIDIA engineers to help make the software libraries to accelerate that process.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Computing power of GPUs using programming extensions on the C language. And that's how things were often running. The AI stuff actually did not really hit till Yes, 15 years after CUDA came out. So if you think about it, CUDA came out in 2006. It didn't really take off for like a decade. But Jensen believed that this was a future computing so much that he dedicated parts of the chip to accelerate the CUDA operations, these things called Kudacores and then later tensor cores. And Wall Street was upset because this hit their gross margins, like their gross margins plummeted in the years after they put the coup de cores onto the graphics chips. But Jensen was just like, no, this is the future computing. I know what's going to happen. I know the world is going to use these GPUs in this way at a certain point. And he told his people to make libraries, these math libraries, these science libraries for every single vertical and to help accelerate these developers to use these GPUs whatever they wanted to do, whether it be for MRI imaging, figuring out the”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Jensen foresaw each graphics chip would have hundreds of processor cores. And then eventually it would go with the thousands and tens of thousands. So they would break down the program and split up the work between hundreds and thousands of process cores. So when you do scientific problems or things like that, you can calculate that all at once across hundreds and thousands, of course. And that would lead to tremendous speedups, 100 to 1,000 times faster than running the same program on CPUs. So he foresaw that the computing whole paradigm would change and the world would move from CPUs to GPUs. And that started with graphics with a programmable shaders. And then that programmable shaders, which was the GPU, people, they thought of the CUDA language in 2006 programming platform that lets non-graphics people use the”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Long term vision that he called accelerating computing that it wasn't just going to be video games, it wasn't just going to be graphics, that this style of parallel computing where a computing workload is split up and all these cores are going to attack that and solve that workload at the same time. He kind of foresaw that that was going to be the future of computing because it was so much more performant and faster at doing this high performance computing. So eventually it did happen. So this is why just a basic overview. Most computer and processors at that time, the microprocessor, which was started by Intel, was a thing called a CPU. And CPUs usually have about four to eight kind of process records, but they do things seriously where they follow the program and then you do one after another, follow that program with about four to eight cores at the time. What a GPU does and what”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“The G4II that allowed developers and PC game makers to have complete artistic freedom how to program all the graphics and art styles. So what that did was give the developers reason to create all these amazing art styles, but also it led to what actually happened later with CUDA when all these non-graphics people start to see all the computing power that's possible in these GPUs and kind of hack into the programmable shader language where you could program the art styles and they would hack that language and use it for non-graphics algorithms applications. So that led to using NVIDIA's GPU computing power in non-gaming, non-graphic simulations work that happened later. And Jensen actually, Curtis told me he knew this from 1983. He had this kind of long”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“So they were scared about Intel coming in and incorporating basic level graphics capability onto the motherboard or onto the chip. Jensen is a student of history. He knows technology companies get disrupted all the time. So he was always thinking, how can I make what we do more performant, higher value that we could protect our margins and not be taken over by a bigger company? And the thing that came up with was a programmable GPU or parammmable shaders. So before programmable shaders, which actually came with the GFOS 3, before that, graphics were a very fixed function.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Like that. Never say die, figure it out. Their first two chips were complete disasters. The MV1, MV2 lost tons of money, got refunded by all the retailers. The MV2 didn't even take off. It wasn't even sold. But somehow he figured out ways he had to lay off half the company, but somehow he figured out a way to raise some money and survive and then launch the ship that did really well. So that kind of resilience and business genius. And then on top of that, his technical expertise is a combination that led to NVIDIA's success.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“That he just licensed the technology from. He's completely ruthless hiring the best talent from the competition. There's this time where they're running out of money, 30 days from going out business. In 1998, Intel was breathing down their neck, they're spreading fear uncertainty about the i740 chip that they told PC makers is going to destroy NVIDIA. He was running out of money. They're having problems with TSMC in terms of yields. Literally, they're weeks away from going bankrupt. And he somehow convinced his three board partners, why don't you invest in us? We'll give you a ten percent discount when we IPO. You know we're good at the technology stuff. Just give us some money and we'll figure it out and you know we're the best. So give us some money right now to tie this over. In the middle of the Asian financial crisis in 1998, but he was able to convince his three main board partners to put in some money when they needed the most. Like most CEOs aren't able to be resilient.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“For really cheap, and that will like block a competitor from coming up from the bottom, right? The big Clayton Christensen and Gang Disrupted by the high volume, low price supplier, he blocked that off with this concept that everyone does now. We have different bins of different chips, a low end part, a medium end part, high end part. So he just create these amazing business strategies that let him survive. And the other thing is he's just super resilient. Like Curtis told me he couldn't believe the rabbits that he would come out of the hat at the last minute. They needed a 2D graphics capability for the Riva 128, which is their third ship, the first one that was successful. And he just got a 2D graphics license from their main competitor. And he's like, how did you do that? Jensen found a way. And then a few weeks later, he hires the top 2D graphics engineer from that competitor.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Curtis Prim, one of the co founders, he is like a technical genius. He created this pseudo operating system that sits on top of the chip that let them put things in hardware and then pull it out if they didn't need anymore and put it and kind of emulate in software. So it allowed them to just do that every six month cycle because if something didn't work, they're developing a hardware feature, they could just put it in software, right? And then if they had enough power, they could put it in hardware. So they gave them a lot more flexibility, these new graphics chip features that other companies didn't have. They didn't have this operating system that sat on top of the graphics hardware. So they just have these like smart advantages that other companies have. Another thing they did was this concept called ship the whole cow. They built in redundancies into the chip where if the yield some of the parts of the chip didn't work, they could sell that chip as a lower end part.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“So you could have like three chips in 18 months. So, one new chip architecture in six months, you have a fastest derivative, and in six months you have another faster derivative. So he figured it out. And then he convinced the PC makers to say, you should stick with us because we're going to keep the drivers the same, the software that you need to run the graphics chips, and it's going to be more reliable. And it's better because everything's backwards compatible. So you don't have to, the drivers are just going to be unified and you don't have to worry about that. And the drivers are a huge headache for PC computer makers. And by executing that six month cycle instead of 18 month cycle, it pretty much blew everyone out of the water. And that kind of thinking that's completely out of the box, no one has done before. Time and time again, Jensen figures something out like that and comes out with a business strategy that is smarter than anyone else and gives NVIDIA a huge advantage. This other thing that”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“And then he realized I figured it out. The PC makers decide which graphic ship to put in their PCs twice a year in the spring and the fall, two seasons. And the reason why graphics chip companies can't stay on top is that the PC makers just choose what's best twice a year and whoever's best at that time, they incorporate that chip into their computers. And he realized the reason why that happens is most graphics chips take about 18 months to two years to make. Like you design the chip and then you kind of tape it out and fix all the bugs. And then at the end of two years, you have a chip that you can sell to ComputerMaker. And Jensen was like, hmm, what if we, instead of 18 months, we make three chips in the 18 months and make a new chip every six months? And then the employees are like, huh? That's impossible. But he figured out a way to make an architecture every 18 months and then two derivatives.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Yeah. I mean, they literally are the only grass company that survived out of, I think someone counted 200 to 300. I mean, the accurate number is probably around 60 or 70. If you think about it, this is a market, a computer industry where Intel pretty much commoditizes any piece of hardware and then incorporates it onto the motherboard or CPU. Microsoft, anytime there's a piece of productivity application, Microsoft would make something that's competitive and kind of bundle it into their Microsoft office. So it's industry that a lot of companies do well for a while and then they get crushed, either by the two gorillas, Microsoft and Intel, or by other competitors. And Jensen kind of had to think about that. Like, why are the PC graphics companies, they always do well for about a year and a half, two years, and then they lose their leadership perch, right? So he had to solve that. And he talked to his executives.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Ranked NVIDIA number two out of the top three, right? And normally at S3, the market manager said, if we got number two, that was like, hooray, great job. We're number two. And there's dozens of graphics companies. That's great. So when he let Jensen know that we were in second place, Jensen got upset, angry, and was like, you know what? Second place is their first loser. And the market manager was stunned. What? Second place the first, like, nothing is acceptable unless you're in first place. Later, I found actually after I wrote the book, that line is from Ferrari, like the founder of Ferrari used that line decades ago. That kind of mentality where you have to be first place, the winner at all times, at all things, I think it's just innate from the beginning. That's just who he is. And some people have that chip on their shoulder and other people don't. And I think that's actually rare. I've talked to probably dozens of CEOs in my life.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“That Jensen has with the business of Midia, that I don't think you can teach. It's just something that he's so passionate and obsessed about. Look, the funny story about one of his early CFOs who used to be a top 50 chess player when he was younger. And he knew this, but he had to beat him. He would study all the chess moves and all the openings. And the CFO would just crush him every time because he knew exactly what Jensen was doing. He would do something a little off. Then Jensen memorized an opening he couldn't react because he wasn't as good as chess. And then every single time he lost, he would flip over the chessboard and knock the pieces off, and then forced to CFO to play him in ping pong because he's better than that ping pong. So he needed to beat him at something. Like that kind of competitive drive. I think it's just innate. It's like people have this drive to win. There's this other story about the marketing manager that came from S3, which was a good competitor. And the first review spread by a major computer PC magazine.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“I think he's just had it innately since the beginning. I mean, when I talk to co-workers and his friends, he's always been lifists. And it's just a competitive drive that's inside him. I think he also has a chip on his shoulder. I compare him to Michael Jordan. Like he would create things that make him angry and then work even harder and then give him more drive. When I talk to him, he brought up this article a journalist wrote like 30 years ago. That person didn't even write NVIDIA on the list of graphics chip companies in the 1990s. Like he's still carrying that slight against him that gives him more a drive to work even harder. I think the work ethic is really important because he talks about how if he goes to a movie theater, he never remembers a movie because he's constantly thinking about NVIDIA and like what he needs to do at all times. So it's this obsession that Michael Jordan had with basketball.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Really real time view of what's happening inside NVIDIA, and that way he can allocate time, energy, and resources almost like an F1 race car driver, perfectly able to steer the race car which is NVIDIA at all times. So this kind of real-time control of NVIDIA is what helps NVIDIA be successful because Jensen has almost like perfect intelligence at all times, and he's able to steer resources, steer the company in a way that is the right thing to do in the current technology landscape and market. And most companies don't do that at all. This is a completely different way of running a company and steering a company.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“The company a day. So it gives them a perfect sample of what's exactly happening inside the company. Jensen has FS, J. Purry, which is executive. Most companies don't do this. If they have these slow status reports where employee sends information or emails to their manager and the manager sanitizes it, takes out all the bad stuff, and sends that status report to his manager. And by the time he gets to the CEO, that's three, four levels up, it's completely useless. First of all, it's too late. Second of all, all the negative things are polished out because the lower manager doesn't want this higher manager to know any of the bad news that's potentially happening. So this top five emails takes care of the slowness of the SaaS reports. And there's a culture of NVIDIA that something bad is happening, like a competitor has a better product and it's doing better or there's a customer that is upset with NVIDIA. You have to share that in these emails. So at all times, Jensen has”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Never happens to NVIDIA. Like, if I'm talking to two people NVIDIA, they have the same tune, the same vision, same strategic direction. And that doesn't happen to other companies. The other thing that's really important is this thing called top five email, which has been around since early 90s of NVIDIA. So basically every NVIDIA employee for every week or every two weeks sends an email to their coworkers on their team, their manager. What's the top five most important things that are happening in my job right now? It might be, oh, I read a paper that is affecting all the AI technology in our area. It might be, oh, I'm falling behind on this project. I need some help. It might be some competitive development or industry development. What are the top five things that are most important right now? And you send that to your team members and your manager. And Jensen is able to somehow in his outlook read 100 of these emails across”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Different general managers, and nothing gets done. Like some times, the managers are incentive to just throw quicksand into the gears and slow everything down. And that's the opposite of the culture NVIDIA. The other thing that is really special about NVIDIA is that I've heard that Apple, everything is siloed and information is guarded and people don't share information at all. Nvidia is the exact opposite. Jensen says he would actually want it to be error on the side of oversharing because if everyone in the company knows the strategic direction where NVIDIA wants to go at all times and you're transparent with information, they're going to make decisions to push NVIDIA in the right direction. There's this large software company executive that I talk to. It's a very large company. And he talked about how when he meets executives at other companies, sometimes those, and they're talking about partnership or deal, those two executives would argue back and forth on what the partner company wants to do. He said, Tay, that.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT
“Just rip you apart and embarrass you in front of everyone. And he told me if you do that once or twice, people will stop doing it. And he does it. So that sniff of internal politics, of meeting after meeting, indecisiveness that is prevalent at every other large company that I've heard when you talk about Microsoft, Google, when I talk to these employees that go to these companies, these former NVIDIA employees, they can't adjust to that kind of company where it's where endless meetings you need to get five stakeholders to say yes. At NVIDIA, Jensen gathers the 20 people that he needs to make a decision. They hash it out. There's a thing called honing the sword where there are, you know, friction brings the best results. So they're yelling at each other, hashing out with data and arguments. And at the end of the meeting, Jensen makes a decision and they go and execute, right? Other companies, they might take five meetings along PowerPoints with five.”
2025-03-07 · We Study Billionaires · TIP704: The Nvidia Way w/ Tae Kim · IDENTIFIED FROM THE TRANSCRIPT