The Nvidia Way
Tae Kim's terrific new book on the story of Nvidia and Jensen Huang
On the most superficial level, Nvidia’s rapid rise to vie for the title of the world’s most valuable company is easy to explain. AI is emerging as a technology that will be both pervasive and revolutionary. Nvidia makes the best chips for AI. QED.
Dig a little deeper, though, and there are lots of questions. How did Nvidia, a designer of graphics chips, make its products essential as the engines of AI? Was it luck or foresight? Why are other companies struggling to compete? And perhaps the most crucial question of all: Will Nvidia be able to maintain its lead?
At the end of his new book on the GPU-turned-AI chip designer, Tae Kim says he had assumed there must already be several books describing Nvidia's rise and the story of its now-famous CEO, Jensen Huang. It turns out that Kim’s new book, ‘The Nvidia Way,’ is the first.
On reflection, it’s perhaps not surprising, as Nvidia’s ascent and appearance in the wider public consciousness have been sudden and dramatic. As recently as 2019, Nvidia’s stock was trading below $4, a far cry from the $130+ where it stands, with a market capitalization of more than three trillion dollars, at the end of 2024.
But Nvidia is a remarkable company with an eventful history. It has a culture and approach to business that are highly distinctive, even compared to its closest peers. At the center of that culture is its founding CEO, Jensen Huang, still leading the firm after more than three decades. It’s a firm that deserves closer inspection, independent of its recent success.
So does Kim’s book do justice to Nvidia’s remarkable story? Does it help to answer the questions we posed above?
The Nvidia Way is two, intertwined, books in one. The first is a straightforward retelling of Nvidia’s history since it was founded by Huang, Curtis Priem, and Chris Malachowsky at a Denny’s in 1993. There are triumphs and there are (near-fatal) disasters.
The disasters came early.
Nvidia’s very first design, the NV1 a graphics chip for the PC, was a story of a firm that had badly misjudged what its customers wanted. Put simply, they wanted to play ‘DOOM’ on their and DOOM used VGA graphics:
John Carmack, the game's designer and cofounder of its publisher, id Software. Carmack built the game using the 2-D Video Graphics Array (VGA) standard and leveraged every hardware-level trick he knew for maximum visual impact.
but
… the NV1 chip only partially supported VGA graphics and relied on a software emulator to supplement its VGA capabilities-which resulted in slow performance for gamers playing DOOM.
With audio support for DOOM’s soundtrack that was arguably even worse, gamers shunned the NV1 in favor of one of Nvidia’s competitor’s products.
The company fared no better with its next product, the NV2 which was abandoned by Sega after a seemingly promising start.
Nvidia was now quickly running out of cash and had a track record of two failed products. Many CEOs would have been deterred and retreated into, likely fatal, caution.
Not Huang. Nvidia’s next product, the RIVA 128, would be even more ambitious and risky than its predecessors:
"I wasn't worried about my cost," Jensen said years later, when asked to explain his decision-making process. "I built a chip that physically was as large as anyone could build at the time. We just wanted to make sure this is the most powerful chip the world's ever seen."
And the bet wasn’t just on the size of the chip. The company also needed to take huge risks with its schedule:
Given its financial position, Nvidia would have to make the RIVA 128 in record time, and without the safety net of multiple quality-assurance runs. Standard chip development usually spans two years, involving multiple revisions to identify and fix bugs after a chip "tape-out," when a finalized chip design is sent for prototype manufacturing. The NV1, for example, had three or four physical tape-outs. Nvidia could afford just one physical tape-out for the NV3 before the company had to send it to production.
The answer was a novel and expensive machine that enabled the Nvidia team to emulate their new design in software. Using the machine was a painful process:
The emulator did not produce any bug reports automatically. Instead, when a program froze, all Levin could do was take a screenshot and call over one of the hardware engineers to figure out what happened or where the corruption occurred. If it was a significant problem, the engineers would go back to redesign a part of the chip.
But it enabled Nvidia to get the chip working, and to market, in record time.
Huang would later say that the RIVA 128 was a ‘miracle’. But he was determined to turn it into a repeatable miracle:
“There's got to be a way to solve this problem of the design cycles.”
The answer was again software, but this time running on Nvidia’s chips. Curtis Priem developed the idea of the ‘resource manager’ which was ‘a miniature operating system that sat on top of the hardware itself’:
The resource manager allowed Nvidia's engineers to emulate certain hardware features that normally needed to be physically printed onto chip circuits. This involved a performance cost but accelerated the pace of innovation, because Nvidia's engineers could take more risks. If the new feature wasn't ready to work in the hardware, Nvidia could emulate it in software. At the same time, the engineers could take hardware features out when there was enough leftover computing power, saving chip die area.
The approach gave Nvidia a decisive advantage. Rivals, including the once-market leader 3dfx, struggled to keep up. 3dfx made several mistakes and in 2002 Nvidia was able to buy its patents and other assets out of bankruptcy and hire around a hundred 3dfx employees.
Even in Nvidia’s early years, the characteristics that would form the foundation of its recent success were visible. The relentless execution, the risk-taking, and the way it combined hardware and software to give it a decisive advantage over its, perhaps more hardware-focused competitors, were all essential components in its later successes.
A few words about how ‘The Nvidia Way’ handles the most technology-intense parts of Nvidia’s story. Kim does a masterful job. Whilst accessible to the general reader, there is enough detail on Nvidia’s products and technologies to ensure that more technically oriented readers remain engaged.
As ‘The Nvidia Way’ continues the second intertwined book comes to the fore. It’s a book that focuses on Nvidia’s culture and Huang and his approach to management.
Many aspects of Huang’s unique management style are now well known: his 60+ direct reports; the ‘top five’ emails; the long-hours culture; and the public way he gives negative feedback.
If you aspire to reproduce ‘Jensen’s Way’ then ‘The Nvidia Way’ might almost be read as a manual for his management style.
But you’ll probably run into several problems. The approach must be uniquely demanding, both for Huang and his wider Nvidia team. It must take unusual stamina to endure the intensity of the approach over several decades as Huang and several of his senior team have done.
Then, and it’s a cliche but here it’s true, Nvidia has been ‘an overnight success that has been thirty years in the making’. Huang might have benefited from the scale of the ‘AI boom’ launched by ChatGPT and other LLMs. But it’s taken thirty years to build a company with the characteristics that have made Nvidia uniquely well-positioned to cash in. Who else has the patience and resilience - or the opportunity - to shape an organization over many decades?
Still, there is more than enough material for readers to understand how Nvidia and Huang operate.
If ‘The Nvidia Way’ has a gap it’s on Nvidia’s recent corporate strategy. There is nothing on the attempted acquisition of CPU designer Arm. Why did Huang want Arm? What was his reaction to the takeover being thwarted? What is his relationship with Arm owner and Softbank CEO Masayoshi Son? We’ll have to wait for a second book on Nvidia for insights into each of these points and more.
Let’s return to two of the questions that we started with. How did Nvidia make itself essential to AI and was it luck or foresight?
Huang has admitted that he didn’t foresee the current LLM-focused AI boom. But Nvidia’s success isn’t luck either. The company positioned itself as the leader in massively parallel computing and ensured that its products were accessible to anyone who wanted to use those capabilities. That positioning has involved heavy investment over almost two decades.
And then, will Nvidia be able to maintain its lead?
Here there is a simple point to be made. As the book makes clear, Nvidia gained leadership in the brutally competitive graphics card market against numerous bigger competitors. It did so on the back of a culture that combined risk-taking with a relentless focus on execution. During its early years, it had no moat and often had limited financial resources.
Today, that culture seems to have largely endured under Huang’s leadership. It has built a strong moat around its ecosystem in the form of CUDA and it now has a scale and resources that no rival can remotely match.
In these circumstances, it would be brave to bet against Nvidia.
This takes us to one final question, which, as you read The Nvidia Way, comes to mind repeatedly. What will happen to Nvidia when Huang leaves?
Huang is 61 and seems as fit and energized by his job as he ever was. Morris Chang was 55 when he founded TSMC and finally retired as CEO at 86. I suspect that a lot will happen between now and Huang’s retirement. Perhaps it’s one question we can leave aside for a few years yet.
The Nvidia Way is highly recommended. It’s an outstanding book that provides a thorough overview of Nvidia’s history and culture whilst also being a great read.
For readers who want more than the book, Tae Kim’s recent interview with
in Jon’s Asianometry Newsletter is terrific.
Author
himself is on Substack:
With links to more on the Nvidia way here:
Finally, if you like your history in podcast rather than book form, the Acquired team did a great job with their three-part history of Nvidia. The Acquired telling of the Nvidia story is, I think, highly complementary to Kim’s book as it focuses more on strategy than culture. The three parts (with transcripts) are linked below:
Plus a great interview with Jensen Huang himself.
My copy just arrived!
With 3dfx, the secret weapon was the Glide API. Nvidia had no recourse but to adopt the OpenGL and eventually the DirectX API. By then 3dfx was buried. There were no others.