Thoughts on <The Thinking Machine>

Opening

Artificial intelligence is a topic that never fails to come up whenever I meet and talk with someone these days. Between conversations with people around me and the steady stream of newsletters and media coverage, I find myself feeling an increasing sense of FOMO (Fear of Missing Out) — wondering whether I am making enough use of AI. After being driven by that FOMO for several months and pushing myself to use AI as much as possible, I sometimes found myself longing to go back to reading and exercising, the things I had enjoyed before AI became such a dominant subject. It felt a lot like how, during exam season, everything except studying suddenly seems incredibly appealing. It was in that state of mind that Jensen Huang's biography came to me. I had vaguely heard from a friend last year that they had read it, and it had stayed somewhere in the back of my mind ever since. So I searched my local library, found it available for borrowing right away, and started reading it immediately. When I began the book, what I was hoping to get out of it was a concrete understanding of how the industries surrounding AI infrastructure — GPUs, memory, and so on — first came to be. I believe that fear comes from uncertainty and unfamiliarity. So rather than simply sitting with FOMO, I thought that if I understood when and how the foundations of AI were laid, my fear would diminish and I would be able to start forming my own picture of the future. This book is not an autobiography written by Jensen Huang himself, but rather a biography assembled by a professional biographer who spent several years interviewing Jensen Huang and key figures at NVIDIA. As a result, even after finishing it, my feeling was less that I had come to deeply know Jensen Huang as an individual, and more that I had just watched a film about the birth and rise of a company. What kept me in a state of excitement throughout was the fact that NVIDIA currently sits at the top of the world's market capitalization rankings — meaning this biography is not merely a story about the past, but a starting point for a future that is still being written.

Passages from the Book

The content from the book that I most want to record here relates largely to the qualities that help paint a picture of Jensen Huang as NVIDIA's founder.

• Jensen Huang's pivotal decision that propelled NVIDIA to its current position as an industry leader
  • At the 2012 ImageNet competition, AlexNet appeared and won — and one of the key reasons for that victory was that its training was powered by NVIDIA GPUs. The training was completed in a matter of days. Jensen Huang began saying this represented "an entirely new way of running software," and NVIDIA made a wholesale strategic pivot toward becoming an AI semiconductor company, staking out an early lead in the market for AI-focused GPUs.
• NVIDIA's "speed of light" scheduling NVIDIA's product lead times were famously capable of compressing work that would take competitors months into just weeks. This was attributed to the scheduling methodology Jensen called "speed of light," a management concept he instilled in his employees with almost religious conviction. The book's author noted that during interviews, virtually every employee mentioned "speed of light" at least once. The idea is to concretely determine the theoretical fastest speed at which a given task could be completed — assuming everything goes perfectly and the budget is unlimited. Managers can then work backward from that unattainable benchmark to arrive at delivery schedules that are both realistic and remarkably fast. One of NVIDIA's managers, Dev, put it this way: "It seems demanding, but it actually reduces pressure. Once you understand the limits of what is physically possible, you also realize that competitors cannot move any faster than that either."
• Jensen's presence on the floor Jensen would roam around the company and frequently stop junior employees to ask them about the work they were doing. He could speak for more than two hours at a stretch without any materials, and at every presentation he repeatedly emphasized the same three themes: the importance of speed-of-light scheduling, pioneering zero-to-billion-dollar markets that others dismissed as fairy tales, and the danger of bureaucracy that quietly creeps in over time.
• NVIDIA's flat and fluid organization Even as NVIDIA grew, Jensen Huang maintained an agile and nimble organizational structure. There were no fixed departments or hierarchies. There were no other C-level titles besides his own, no clear second-in-command, and not even a chief of staff. Instead, more than thirty people reported directly to him. Most of them held the broad title of "Vice President" while taking on fluid, shifting roles.
• Jensen once told a gathering of executives: "I think all of you need to be ready. No one knows when you might suddenly become the most important person in this company."
• Jensen frequently contradicted himself, sometimes offering opposing viewpoints within a single interview. He simply enjoyed attacking ideas from both sides at once. A close associate explained: "He's not trying to be a politician. He's not trying to keep his message consistent. He's trying to process information as it comes in real time. And he's willing to hold contradictory thoughts, even if only temporarily." What might appear on the surface to be a firm declaration was often, in fact, Jensen simply thinking out loud. His morning routine
• At four in the morning, Jensen would rise and begin working. He always started his day with the most important long-term project on his plate. His reasoning was that once that was handled, no matter how the rest of the day unfolded, it could not be called a wasted one.
• NVIDIA's breakthrough emerged at the intersection of two areas that had long been treated as outliers in computer science: parallel processing computing and neural network training.

Rather than directly quoting from the book, I would like to share my own account of the story I found most gripping — NVIDIA's brush with crisis. NVIDIA is said to have faced three near-bankruptcy situations since its founding. Of these, the one I found most riveting was the story of what happened when the NV2 chip was about halfway through development. After launching with the NV1 chip, NVIDIA had gotten partway through developing the NV2 when it became clear that their cost structure was far too high compared to competitors, and their only major customer, SEGA, indicated that NVIDIA's graphics processing approach was not the right fit for their needs. At that point, NVIDIA had enough cash to operate for only about nine months. They needed to bring a new, company-saving chip to market within that window — and had already cut 70 percent of their workforce. For a fabless company, which typically operates on a product development cycle of at least two years, this was an impossible set of conditions.

The gamble NVIDIA chose was to skip the prototyping stage entirely by using an emulator, go straight from design to tape-out, and immediately place an order with their foundry for 100,000 units. The result, delivered in just six months, was the RIVA 128 chip. From that point on, while other chip companies released new products on 18-to-24-month cycles, NVIDIA began shipping new graphics cards every six months. The RIVA 128 sold one million units within four months of launch, and NVIDIA survived.

Of course, the chip did not simply sell itself the moment it hit the market. It was the result of NVIDIA identifying performance-obsessed game developers and the hardcore gamers who sought out those games, and going directly to them to sell. Whenever I encounter a story like this — one that looks easy in hindsight but, under closer examination, was anything but, and was in fact closer to a miracle — I feel a renewed motivation to live and work with greater intensity. Perhaps that is exactly why I love this story enough to want to preserve it here.

Closing

For anyone curious to learn more about NVIDIA as a company and Jensen Huang as a person, I would strongly recommend reading this book yourself. And if you happened to be someone who loved computer games around 1993, when NVIDIA was founded, the atmosphere of that era and the story of NVIDIA's struggle to survive will likely resonate with you on an even deeper level.

Having finished the book and now writing this review, as I look back on what has stayed with me — in both my head and my heart — there are many dry facts and figures, but there are also moments that give you a genuine sense of who Jensen is as a person. A few that personally stood out to me: that Jensen was skilled enough at table tennis to place third in a junior national competition; that he worked part-time at the American diner Denny's as a student, and still periodically goes back to eat his favorite menu items there, as he did around the time he founded NVIDIA; that he was immersed in cutting-edge computing technology in Silicon Valley as far back as the 1980s; and that his career was not that of a chip engineer, but a chip PM. I hope that those who read this book will find their own points of inspiration through their own personal filters, and that it becomes the kind of reading experience that leaves a lasting, positive influence.