Numlock Sunday: Tae Kim on Nvidia's rise to the top

Photo by Harrison Broadbent on Unsplash

By Walt Hickey

Welcome to the Numlock Sunday edition.

This week, I spoke to Tae Kim, author of The Nvidia Way: Jensen Huang and the Making of a Tech Giant.

We spoke about how Nvidia got started, the company culture that differentiates it from rivals like Intel, and how the American tech industry has been trending away from research and development.

Kim can be found at Barron's, and the book is out this week and can be found wherever books are sold.

This interview has been condensed and edited.

Tae, thank you so much for coming on.

Walt, thanks for having me.

It is my absolute pleasure. You are the author of a brand-new book all about what has become one of the most — if not the most — interesting companies in America within the past year. You wrote a whole book all about Nvidia and specifically how they do business, how they operate internally, and what they've done to rise to the top of the heap. Just backing out a little bit, do you want to talk a little about your career and what first drew you to covering this company?

Just starting off, I've always been a video game, computer gaming nerd. I've been buying and incorporating Nvidia graphics cards for 30 years now. That's how I knew of the company: growing up with computer games starting in the 1990s.

As for my career, I graduated with a history degree from Brown in Providence. I did management consulting for a while; I had a few stints on Wall Street, hedge funds, fund of funds. I wound up in media about 10 years ago. My first job in financial journalism was at Yahoo Finance. I went to CNBC, then Barron's. I was a tech columnist at Bloomberg Opinion for a while. Now I'm back at Barron's writing about technology.

I love that you shouted out gaming. Nvidia first got on my radar because they made the good graphics cards. When you were building a PC, you wanted Nvidia. Then you started hearing a lot about them during the crypto moment, I want to say, because all of a sudden graphics cards became very useful for that. Now, obviously, they're the pickaxe in the gold rush that is AI. This company is super interesting. Do you want to walk us through it? It's not a new company — it's been around for quite some time.

It started, actually, in 1993. The famous story is the three cofounders sat in a Denny's and figured out the business plan for making a graphics company that focused on 3D graphics. This was one of the reasons I was so honored to write this story: I couldn't believe no one has written the history of Nvidia before. Maybe not for the mainstream, but for technology enthusiasts or computer enthusiasts, this is a big company that's been around forever and has done really well.

In the 1990s, they had their ups and downs. They almost went bankrupt two or three times, but they wound up being the leader of PC 3D graphics. The thing about Jensen Huang, the cofounder and CEO, is that he's always looking at the next thing, and he doesn't want to get commoditized. He always has this foresight to invest in R&D aggressively and think long-term. That really differentiates Nvidia and Jensen from its competitors. He was on the forefront of programmable shaders for graphics chips, CUDA, AI. Every time there's a major computing shift in the chips industry, he's there investing years ahead of time.

I like the analogy of Reed Hastings. Reed Hastings had this intuitive sense that video streaming over the internet would happen someday, but the technology wasn't there. Consumers didn't have broadband. So he stuck around. He positioned Netflix to work on the technology, do DVD rental by mail, build revenue, and then keep investing along the way. Then when it actually was viable, he was there to pounce on it and become the dominant market leader.

Nvidia did that time and time again throughout its history, and this latest wave is the biggest wave. There are so many stories. He just sees the stuff before it happens, and that's because he has great technical acumen, great knowledge of technology, more so than any other CEO.

That's a really cool insight. You've described his management practices as unorthodox. I've worked in corporate America; I know a lot of folks have. There is that kind of focus, like the next quarter is the one that counts. This kind of multi-year, sometimes even multi-decade time horizon is very unorthodox. What are some other ways that Jensen ran this company in an atypical manner for, obviously, pretty significant gain?

So, the name of the book is The Nvidia Way, and I would probably break it down to three of the biggest components. The first one is incredible work ethic. From the beginning, he was working from 9 a.m. to past midnight, and his employees follow that work ethic.

There's this focus on talent cultivation. He's constantly recruiting the best talent. And if you're doing well and he sees promise in you, he'll give you spot bonuses, spot refreshers. One thing you never hear at Nvidia is people complaining about their pay. Maybe one or two people complain, but the vast majority of people feel like they're paid very well.

The last and maybe most important thing is this focus on speed and velocity. They call it the speed of light. At most companies, as they grow larger, they become more dysfunctional. People are gaining metrics. There are all these endless political games, endless meetings, indecisions. I hear these employees who work at Google or whatever have to get sign-off from five different stakeholders and have meeting after meeting after meeting. There's all this indecision as the bureaucracy grows.

At Nvidia, there's none of that. You have one meeting with Jensen, maybe 20 people in the room, and they just hash it out. The decision is made, and you go and execute. There's just constant speed and extreme velocity in the decision-making, and things just move faster.

Interesting. It seems like it's a combination of both moving fast but also having a long view on that kind of thing that's compelling.

I remember from reading the book, you interviewed over 100 people for this thing. What was that like?

I did this in about a year, and it's such a short time period, but I benefit a lot from modern media and the internet. LinkedIn was fantastic. YouTube was great. Once you start talking to the first five people, you look into their networks and you can reach out to 10 more people.

Once I got the momentum going, all these early Nvidia employees wanted to tell the story. They are so excited that someone is actually writing the book on it. They almost feel like, why haven't we had multiple books like other tech companies? It just turned out to be really great, and people were more than willing to talk to me. Some of them spent four to five hours with me. That part of it was amazing, being able to compile the computer history for the first time, especially the '90s stuff.

Nvidia is an American tech company, and I think it's actually been a little more under the radar than even some of their most persistent and now fairly stuck-in-the-rearview-mirror rivals. Intel has a pretty widespread and well-known reputation, and even AMD, to a lesser extent. They've been around for three decades, and it feels like they're less known, but they've been in competition with those folks for years.

Intel is almost like Goliath versus David with Nvidia. It's the rise and fall of two kings. In the 1990s, Intel was the dominant chip company. They pretty much created the semiconductor industry, and they almost crushed Nvidia in the '90s. It was almost like Microsoft in the '80s. They would tell the PC makers, "We're coming out with something better. We're going to crush Nvidia," and it really dried up Nvidia's sales pipeline. Jensen had to inspire his troops, saying Intel is out to get us, they're out to kill us, and we need to fight back and beat them. And it took a few decades, but they definitely beat Intel.

Intel has missed the mobile wave. They completely missed the AI wave. Part of it is first they had CEOs that didn't really know the technology, and then they got caught up in the indecision and bureaucracy. I have this great quote from a senior head of engineering at Nvidia who said that Intel's AI strategy just doesn't make any sense because they bought multiple startups. They started and closed, started and closed the internal GPU program. You can't do that kind of portfolio strategy. It's almost like Intel's just throwing darts, he called it. To win in tech, you need a singular focus, make a big bet, and drive that through. That's what Nvidia does. Intel had Nervana, Habana, two startups, Larrabee, internal GPUs. You can't do that. You need to just make a decision and drive it through.

Fascinating. You've reported this book on a really tight timeline, which is very understandable given just how significant this company has become. Do you want to talk about the last two years of Nvidia? Again, I feel like they went from being a company that was known among certain crowds, whether it was gamers or crypto folks, to unquestionably the single most interesting and important company in America.

The last two years are going to go down in history as completely unprecedented. The company's been growing at triple-digit percent growth rates at an enormous size. We're talking tens of billions of dollars. They perfectly positioned themselves to take advantage of this AI, large language model wave that was sparked by ChatGPT. They launched the Hopper GPU that had the transformer engine, which is optimized for large language models, a month before ChatGPT came out.

They've always been looking at this area and saying, "This is going to be a big area. We're going to invest for years in making the best hardware, software, and networking optimized for this area." They were perfectly there. They had all the software libraries, this great ecosystem in CUDA, and now they're powering the AI wave that is just taking off like a rocket the last two years.

How on earth did they do that? It seems like everyone else was either caught flat-footed or is now spending a fortune to try to catch up. They were the only ones really ready for the starting gun of this scientific discovery.

Jensen has his finger on the pulse of technology. He is passionate about this stuff. Anytime there's a big AI paper, he's reading it, sharing it with his top engineers. They're on these email lists, passing articles back and forth, discussing the latest trends. He's on top of the technology. If you actually look back, they have this AI podcast at Nvidia, and Ian Buck in 2019 was talking about natural language processing, large language models, scaling. All this stuff that just blew up in the last two years, he was talking about in 2019. They're on top of the technology, and they're building the hardware and software. They don't know exactly when the market is going to take off, like I keep saying about the Netflix analogy, but they're there, prepared. Maybe it's five, 10, 15 years out, but they know they're prepared.

If you actually go back to when they bought Mellanox, look at the press release. I think this was also in 2019. Jensen says, I need to buy Mellanox here for about $7 billion because someday the future is going to be AI. It's going to be AI workloads and data centers, and these AI workloads are going to work across tens of thousands of AI servers, and Mellanox is going to be a critical component because these AI servers need to be networked with the best technology. And Mellanox had the best networking technology.

He wrote that in 2019, and that's exactly what happened in the last 12 months. With that kind of foresight, seeing where the market was going to go, he prepared for that five years ago. You can even say that with CUDA in 2006. CUDA, which is the programming platform for parallel computing that works on Nvidia GPUs, didn't really take off for five or 10 years. Wall Street was all over him saying, why are you investing in this CUDA stuff? No one is really using it. Why are you crushing your gross margins having hardware circuits that run the CUDA stuff?

And he just said, this is going to be the future. The world is going to move to "accelerated computing," he calls it, where you could break up workloads across tens of thousands of cores in these GPUs instead of the serial processing that happens on CPUs, which is only four to eight cores. He just saw that someday we're going to distribute the workloads, and GPUs and accelerated computing are going to be the future. He was the only one willing to invest in it and take the hit from Wall Street and say, no, no, no. I need to invest in this, even if it hurts my near-term results.

That's why Intel fell apart: They weren't willing to invest R&D in long-term projects and were looking out for their P&Ls and stock buybacks and dividends, looking out one quarter, two quarters, one year at a time, while Jensen's looking at five, 10, 15 years. He's a student of history, he says, and he is so paranoid of getting disrupted that he says if you don't invest, you're dead. He took that to heart. And it's not just about investing; he's the person smart enough with the business knowledge and the technical acumen to make the right decisions.

I actually want to spend a little bit of time here, just because you've hit on what I think is something deeply fascinating about not just Nvidia, but the technology industry and where American corporations are right now.

It seems like a big change that happened is that American companies don't do internal R&D like they used to. In lieu of that, they'll allow venture capital to develop startups that will come up with new technologies that they can then acquire — which absolutely does happen in Nvidia's case, like with Mellanox, as you said. But it seemed that video games, as I was reading the book, was a way for Nvidia to get a bunch of money that they could throw at R&D, which other companies were too cowardly to do. What's your read on that, and how does R&D fit into the Nvidia equation here?

The Netflix analogy works again, right? They were able to make money by renting DVDs to do the R&D into the internet streaming technology. You're exactly right. They were able to make money dominating the PC 3D gaming graphics market, and they made tons of money. They were the fastest-growing chip company in the first 10 years after they IPO'd, and they were able to use that money to invest in new graphics technology.

There's also this thing called ray tracing, which does realistic lighting effects. They invested in that for 10 years before it became a part of their gaming GPUs. Who does that? Who invests in something for 10 years and then incorporates it? But Jensen's willing to keep investing. If it shows promise, even if it didn't work out in two years or four years, he'll keep investing because he sees this technology as being very important. Now, if you don't have Nvidia ray tracing, that's why gamers are willing to pay up to $500 or $600 for their high-end cards. They want that ray tracing and this other technology called DLSS — these special Nvidia technologies that took six to 10 years to develop.

Reading the book, I just kept coming away with actually a bit of sadness, because other companies can't work this way. If you invested in something with no results for 10 years, Wall Street would have your head on a pike. It's fascinating that they got away with it, and my god, have they been rewarded for that risk.

It's not like they're not doing well in this process. Maybe for the five years, they had a bit of a downturn, but they're selling graphics chips and having that data center business going while they're doing this R&D.

One other thing I do want to mention that's very different about Nvidia is that he has this thing where he asks people, what are you optimizing for? Are you optimizing for making the company better and making the right decisions, or are you optimizing for coddling someone's feelings? He's willing to dress people down in public when they make a mistake or aren't doing well. In most companies, the CEO, when they have a problem with one of their executives, they'll take them aside in a one-on-one meeting and try to coddle them and say, well, can you do this? Can you do that? To not embarrass them in public. That takes a lot of the CEO's time, and it just takes longer. But Jensen's point is, I'm not optimizing for an executive's feeling. I'm optimizing for the company's future, for faster decision-making. I'll call the person Bob: Why should Bob be the only person that benefits from learning from his mistake? Everyone in the room, all the employees, should benefit. When I teach Bob what he's doing wrong, everyone in the room can learn from that.

That kind of mentality and being direct and blunt is not really a common practice in American companies, and that's why Nvidia is able to move faster and people are able to work harder to make Nvidia successful. I do think that's really a key cultural component.

Fascinating. The book's a real delight. Before we wrap things up, is there any favorite thing you learned over the course of reporting this book out?

A lot of the playbook of how they won the video game graphics business — accelerating the product cycle from 18 months to six months for a typical chip to come out, investing in the software ecosystem, the software drivers for the gaming chips? That playbook that he used to win the 3D graphics race where they're literally, after 100 companies, the only company remaining standing that hasn't either gone bankrupt or been acquired? He's using that same playbook again.

When he has something that's successful or when he makes a mistake, he pockets that learning and uses it again. That's what he's doing with AI. He just accelerated the product cycle. A typical AI GPU used to take two years; he changed that to one year. He's investing aggressively in the software libraries in CUDA and evangelizing all the teaching of the programming.

So, the amazing thing about Jensen is that he developed these really successful business strategies and he just plays that playbook. And once he does it, it's really hard for anyone to catch up. If you're used to making a chip every 18 months and Nvidia decides to do it every six months, it's almost impossible to catch up unless you have a similar culture of work ethic and speed and velocity like I talked about.

All right. This is really great. I'd love to give you the floor for a sec to tell people a little about the book, a little bit about you, and where they can buy the book.

The Nvidia Way comes out on December 10 at every local bookstore. It will be out on audiobook, on Kindle and at all the major bookstores, so please buy it. It's been the honor of a lifetime to be able to bring this story together and talk to all these early computer industry employees.

Amazing. The book's available this week and folks should get it as soon as they can. Thank you so much for coming on, Tae. I really appreciate it.

Thanks, Walt. Thanks for having me.

---

Edited by Susie Stark.

If you have anything you’d like to see in this Sunday special, shoot me an email. Comment below! Thanks for reading, and thanks so much for supporting Numlock.

Thank you so much for becoming a paid subscriber!

Send links to me on Twitter at @WaltHickey or email me with numbers, tips or feedback at walt@numlock.news.