The Thinking Machine: Jensen Huang, Nvidia, and the World's Most Coveted Microchip by Stephen Witt

AI is hard to miss these days. A ranking of the most valuable public companies in the world right now shows that nine of the eleven are directly involved in AI technologies, while the other two are state-controlled oil companies (Saudi Aramco and PetroChina, respectively). At the top of the list, having quadrupled in value in the past two years alone from a $1T to $4T as of July 2025, sits Nvidia.

It is perhaps odd that a single chip company, one that doesn’t even manufacture its own chips, can command such a lofty valuation. Yet, the more you learn about the current generation of AI–most notably Large Language Models (LLMs) like ChatGPT–the more you see how integral Nvidia is, for now, to the process.

In this timely book, Journalist Stephen Witt sets out to profile both the company and its high-profile CEO, Jensen Huang. Witt does an admirable job describing Nvidia’s history, from its founding in a Denny’s Restaurant in 1993, fighting off threats of bankruptcy and activist investors over the years, all the way to its stratospheric heights of today.

Huang’s intelligence and ambition were clear even at an early age. At age 10, he immigrated from Thailand to the United States, eventually settling in Oregon, where he was a star student and nationally-ranked Ping-Pong player. He studied electrical engineering in college, where he met his wife, and immediately went to work in the microchip business, where he made a distinct impression on all those around him as someone smart, hardworking, and destined for success. A former manager of his, tasked with giving Huang an employee evaluation, left the grades blank. “At the bottom, the manager wrote, ‘Jensen is an excellent employee. I look forward to working for him some day.’”

By his late twenties, working at LSI Logic, “he was in charge of a division of $250 million in annual revenue and with many older and more experienced employees answering to him.” But this early success also caused some friction and a senior director from Intel was brought in to comanage the product line. When two brilliant colleagues, Chris Malachowsky and Curtis Priem, approached Huang to run their graphics accelerator startup, he accepted and, at age 30, became one of the three founders, President, and CEO of Nvidia, all titles which he retains to this day.

Joining a graphics company startup was a risky bet. As Witt notes, “Competitive threats were inherent in any capitalist enterprise, but in the microchip sector those threats were of a different order. For a business like Coca-Cola, once you established a winning formula, the product sold itself–your job was not to tamper with success. The microchip industry was more like the fashion business–if your product today resembled your product from yesterday, you had made a terrible mistake.”

The relentless rate of change in the chip industry meant that even the biggest incumbents had to reinvent themselves time and time again to stay relevant. In the early 1990s, Intel and Sun were two of the giants in the general computer chip space. Rather than try to compete with them directly, Nvidia focused specifically on PC video games, a smaller but growing space driven by popular titles such as Myst and Doom at the time. The developers behind these titles needed ever more processing power, which would come in handy later on.

Over the next two decades, Nvidia weathered ups and downs, reaching massive heights during the dot-com craze and then crashing, along with everyone else, for many years thereafter. Throughout this, Huang remained in charge. Witt notes, “As Nvidia grew, Huang maintained an agile corporate structure, with no fixed divisions or hierarchy. The C-Suite was essentially just him, with no COO, no CTO, no CMO, and no obvious second-in-command. Huang didn’t even have a chief of staff.”

This approach worked in part because Huang was a workaholic who regularly did twelve-hour days, six days a week, and showed no signs of letting up. Despite his outwardly genial demeanor in public settings, he was notorious for screaming outbursts at employees that would shock newcomers. Witt theorizes that Huang is intentional about these outbursts, noting that they almost always happened in public settings and were meant to send a message. They stand in stark contrast to his otherwise staid personal life, where close friends recount never seeing him lose his temper in such a fashion.

Despite such outbursts, Huang was incredibly loyal to his employees–even though he publicly castigated them. In sharp contrast to other tech CEOs like, say, Elon Musk, who are famously socially awkward and seem to relish firing entire divisions on the spur of the moment, Huang comes across as much more personable and humane. If you see him present on stage, as he does often, he now wears the same outfit of dark shirts and pants along with a black leather jacket.

What stands out about Huang is his willingness to take risks, his deep technical understanding, and indeed his passion for chips, combined with a relentless work ethic. He is, by far, the longest-tenured CEO of a technology company at this point.

Back in 2014, Huang made perhaps his most dramatic move yet, reinventing Nvidia as an AI company focused on GPUs far before anyone else. “Neural nets” were an AI technique that was not new, but in 2012 AlexNet, an image classification system built using neural networks, shocked the academic image classification space. As Fei-Fei Li, the Stanford academic behind the competition, noted in her autobiography, The Worlds I See, “It was like being told the land speed record had been broken by a margin of a hundred miles per hour in a Honda Civic.” Academics didn’t believe the results were correct at first.

“Krizhevsky [the lead author] pioneered a number of important programming techniques, but his key finding was that a GPU could train neural networks hundreds of times faster than a CPU could.” He had trained it quite literally in his childhood bedroom with two Nvidia chips.

In 2017, Google researchers introduced the now-famous Transformer deep learning architecture that demonstrated a new approach, powered by GPU chips largely made by Nvidia, to machine learning. Many of the recent AI advancements, such as ChatGPT, are built directly on top of it.

The large tech companies were well aware of these trends and started investing heavily in researchers and the computing power required for these new and ever-larger models. As Witt notes, “Up until around 2018, AI had flourished in a spirit of open academic collaboration. Now the innovations were coming from skunkworks R&D.”

These days, frontier models can easily cost north of $100 million to train, such as GPT-4, with much of the money making its way to Nvidia through Microsoft. Witt also highlights that, along with continual hardware advancements, it is Nvidia’s focus on software that might be its key advantage, given that they have built an entire AI ecosystem around these chips and optimized ways to maximize performance.

One of the unusual things about the microchip space is that many leading companies don’t do the manufacturing themselves. In the past, companies like Intel were both foundries (making the chips) and designers. But the rise of Taiwan Semiconductor Manufacturing Company Limited (TSMC), in particular, has made that approach seem old-fashioned. The technological advancements and investment made by TSMC, itself now the 11th most valuable company in the world, have resulted in major tech firms, including Apple, Nvidia, and AMD, outsourcing their chip manufacturing and focusing instead on the design and related software. Samsung is the world’s second-largest foundry, but it is dwarfed in size by TSMC.

This concentration of power and capital has resulted in outsized gains for Nvidia and TSMC, as well as added to geopolitical risks given China’s regular saber-rattling about Taiwan. A Chinese missile strike on TSMC would not just spark a war with Taiwan but also disrupt the supply of microchips used in supercomputers, the military, and even the Nintendo Switch, which are all made there.

Near the end of the book, the author tries to draw out of Huang some predictions on the possible negative outcomes of AI advancement. Huang replies, “Well, ask yourself, when the marginal cost of doing math goes to zero, then what do you do?’ he said.”

“We invented agriculture and then made the marginal cost of producing food zero. It was good for society! We manufactured electricity at scale, and it caused the marginal cost of chopping down trees, lighting fires, carrying fires and torches around to approximately zero, and we went off to do something else. And then we made the marginal cost of doing calculations–long division! We made it zero!” He was yelling now. “We make the marginal cost of things zero, generation after generation after generation, and this exact conversation happens every single time!”

In short, whatever detrimental side effects arise out of AI’s growth are not Nvidia’s or Huang’s direct concern.

Personally, I found the book to be a very accessible look at the microchip industry in general and Nvidia in particular. There are occasional deep dives into technical descriptions that might cause non-technical readers’ eyes to glaze over. Still, for someone like me, I found it refreshing that the author delved into the underlying technological advancements that spurred this innovation.