As Nvidia posts another round of explosive earnings, few people are better positioned to explain the company’s meteoric rise, and the temperament of the man steering it, than Stephen Witt. Witt is the author of The Thinking Machine: Jensen Huang, Nvidia, and the World’s Most Coveted Microchip, the first deeply reported narrative biography of Nvidia’s co-founder and CEO.
In a timely conversation on Kaleidoscope and iHeartPodcast’s TechStuff, hosted by Kaleidoscope Co-founder and CEO Oz Woloshyn, Witt unpacks the forces driving the global AI boom and the unconventional path that led Nvidia from gaming GPUs to the beating heart of today’s AI revolution. Drawing from years of research and firsthand encounters with Huang, he explores the CEO’s relentless personality, his habit of pushing employees to breaking point, and the paradoxical way he behaves under pressure.
Witt also explains how a group of once-fringe researchers using consumer-grade Nvidia gaming cards accidentally sparked the modern AI wave; why Nvidia insiders fear a historic bubble; what it’s like inside a secret Microsoft AI compute facility; and why Huang remains consumed by threats, from Chinese competition to the possibility that AI’s scaling laws suddenly stop working.
Q1. You’ve written about the roots of today’s AI boom. How did Nvidia’s gaming hardware end up powering an AI revolution?
A: “It’s for mad scientists. It’s for scientists who are pursuing unpopular or weird or kind of offbeat scientific projects, but ultimately the key use case turned out to be AI and specifically a branch of AI that most AI researchers thought was crazy… They couldn’t get $5,000 in research funding, but they could get enough money to afford two $500 Nvidia retail graphics gaming cards, which they did. And then they turned the neural net that they ran on these cards into something called AlexNet, which then started to recognize images better than any AI had ever done before. It smashed the paradigm.”
Q2. You’ve spent significant time observing Jensen Huang personally. What does it feel like to sit across from him?
A: “It is like sticking your finger in an electric socket. He’s so tightly wound. He expects so much to happen. In every conversation, just to even start talking to him, you have to be totally up to speed. He’s not gonna waste any time. He’s not gonna suffer fools. And he has what I would describe as somewhat self-indulgent performances of anger from time to time… If you’re not delivering, he’s going to stand you up in front of an audience of people and just start screaming at you. Really, I mean, yelling, and he will humiliate you in front of an audience. People at Nvidia have to develop very thick skins.”
Q3. You’ve even been on the receiving end of that intensity. Can you walk us through that experience?
A: “I presented him with a clip from Arthur C. Clarke… I was hoping to get that response from Jensen. Instead, he just started screaming at me about how stupid the clip was, how he didn’t give a shit about Arthur C. Clarke. He’d never read one of his books. He didn’t read science fiction, and he thought the whole line of questioning was pedestrian and that I was letting him down by asking. I was wasting his time despite having written his biography. Jensen remains a little puzzle and extremely neurotic; by his own admission, he’s totally driven by negative emotions.”
Q4. How does Huang handle high-pressure moments inside the company, especially during crisis periods?
A: “Jensen is actually a lot calmer and more compassionate with his employees when things are going wrong. It’s when the company’s stock price is way up and it looks like everything’s going great that he really becomes much more cruel, like much, much meaner to everybody… When he succeeds, it makes him nervous.”
Q5. Nvidia’s rise wasn’t inevitable. What were the early GPU wars like from inside the industry?
A: “At one point there were 50 or 60 participants in this market. Jensen would go into his office with a whiteboard and he would have a list of all his competitors up there, and not only that, they would have a list of who the best engineers working at those competitors were. Then they would come up with plans to poach those engineers… I’ve compared the early graphics days to the movie Battle Royale… He won the knife fight. He was the last guy standing.”
Q6. You’ve said we’re entering one of the largest capital deployments in history. What does this new AI industrial revolution look like to you?
A: “We’re basically building these giant barns full of Nvidia microchips to run calculations to build better AI 24/7 around the clock, and it’s one of the largest deployments of capital in human history… NVIDIA’s stock is more concentrated in the S&P 500 than any stock since they started keeping track… 15% of the stock market is these two stocks [NVIDIA + Microsoft]. If NVIDIA crashes, it’s gonna create a lot of pain throughout the economy. Americans in particular are usually invested passively through index funds in something that looks exactly like the S&P 500.”
Q7. You were granted rare access to a highly restricted Microsoft AI training facility. What did you see inside?
A: “It looked like an invasion by aliens… three vehicle checkpoints… you have to sign 15 NDAs… It’s a giant concrete barn just full of repeated racks of equipment as far as the eye can see… In the control room, the power going up and the power going down, the power going up and the power going down… With every pulse there, it just got a little bit smarter.”
Q8. Despite Nvidia’s dominance, what are the main risks Jensen Huang still worries about?
A: “Conquering one cycle in microchips is no guarantee that you will conquer the next one. There’s a big risk that Chinese companies build alternative cheaper stacks… Maybe there’s a technological solution that trains these things faster… Or the mysterious scaling law finally plateaus – no one is entirely sure why this works. If we did hit a plateau, I think this whole era would look kind of like a bubble.”
Q9. You’ve written about AI’s darker possibilities. What does existential risk look like in practical terms?
A: “The AI prompt that could end the world: Someone gets a hold of the machine that has an agency function—they can make real-world actions—and they are told, ‘Do anything you can to avoid being turned off. This is your only imperative.’ If you gave that prompt to the wrong machine, it might secure its own power facility, blackmail humans, or even attack humans. These systems have emergent capabilities that the designers are often only discovering empirically, and that is very scary.”