I picked up “The NVIDIA Way” by Tae Kim as I was walking through the stacks of books in my local library and cross-referencing with Goodreads references. This was a fantastic book that I reinforced a lot of business lessons that I have picked up at Ramp. Seeing how Jensen Huang was able to fortify his company across several near-death experiences has made him an even more inspirational figure in my eyes. Without further adieu, here are some of the main lessons I extracted out of the book.
Chips are not what makes NVIDIA great; it’s process
Throughout the book, we see that NVIDIA had made some incredible technical leaps, such as launching the revolutionary GeForce and RDX family of chips, developing CUDA and being a visionary with AI on chips. However, what was more striking to me is how laser-focused the company is on improving their development process. Some examples:
- “Speed of Light”: NVIDIA prides itself on execution speed and will do whatever is necessary to make their development process faster. For example, in the early innings of graphic chips development, Jensen realized that the cycle time of development is often bandwidth-constrained by chip manufacturing. He immediately found an emerging company building chip simulators, which were not that common at the time, and asked employees to focus on driver development first and worry about hardware at the end. This process shift propelled NVIDIA to ship chips at twice the speed as their competitors, overwhelming and dominating them. Ramp similarly really prides itself on speed and does whatever it can do internally to remove roadblocks to help ship and move faster. This in itself is a competitive advantage.
- Insane sense of urgency and paranoia: Throughout the book, Jensen engenders paranoia among all of his associates to help avoid the near-death experiences that the company went through at the earlier stages. Everything is existential in his eyes, which helps avoid Innovator Dilemma’s. I have noticed this same mentality among some of the cofounders in Ramp and with Zuckerberg.
- ”Ship the whole cow”: Jensen noticed how much waster NVIDIA was generating by throwing away components that couldn’t perform at the level their most performant chips demanded. He instead realized there was an opportunity to use these weaker components to create weaker but cheaper graphics chips, like how a butcher tries to sell all parts of a cow instead of only selling the best cuts.
Obsessions and long work hours
One thing that I found interesting is NVIDIA’s approach to work-life balance, or the lack of it. Jensen strongly believes that there is no shortcut to achieving NVIDIA’s mission; long hours and obsession is a must. While NVIDIA tries to avoid employee burnout by giving generous perks, there’s no getting around the late-night and weekend work style. While this is not as performative as the 9-9-6 culture that has gripped many SF startups, I do think its interesting to ponder how much I am willing to work for a company’s mission and what types of people it tends to attract.
How CUDA came into existence
CUDA is NVIDIA’s moat. It allows people to program GPUs to perform parallel computing, which is heavily used in scientific and AI workloads. For the longest time, I had thought CUDA was the brainchild of some chip architect, but actually it has a much more humble origin story.
Mark Harris, a CS researcher, noticed that many of his fellow researchers were using NVDIA GeForce chips for non-graphical use-cases, namely to run simulations. GPUs, by design, are much better at parallel computing compared to CPUs, but using it for non-graphical uses was pretty unheard of. At the time, NVIDIA had some programmable shader technology that scientists could use, but it was a huge pain for scientists to use to properly parallelize their workload. Regardless, Harris started evangelizing the GPU parallel computing gospel, eventually catching the ears of NVIDIA. NVIDIA hired Harris to exploit this new use case and help chip teams design their chips to make parallel computing easier, which eventually led to the development of CUDA. Once Jensen realized the magnitude of importance of CUDA, he pushed for all NVIDIA chips to immediately support it for immediate market saturation. He wanted CUDA to be as easy to use as possible and relegating CUDA to the top-line of chips would forever cement its status as a high-end workload tool, which is what he wanted to avoid. This immediate back-compatability was expensive to develop but was a genius move in hindsight, as it instantly made CUDA popular.
This example is a beautiful illustration of how products should develop. Listen to the customer, bring them on as close partners in the development process, and make products as easy to use as possible.
Jensen’s leadership style
One thing that has been in my mind for a while as Ramp continues to scale is how our organization will change as more people work for the company. What I am most worried about is internal politics and slowdown. Jensen has been thinking about the same problem for NVIDIA, as it’s only true moat is its culture of non-stop, fast-paced shipping. Some ways he counterracts inertia and complacency:
- Manufacturing urgency in everything: this paranoia seeps into everything. He is willing to give tough questions to people and force them to work faster and better. Shock treatment seems to work well in NVIDIA
- He is involved in many small decisions: this was a bit of a surprise to me, but Jensen, if he thinks is important, is willing to go to the smallest meetings and sort out things. For example, there is a whole chapter in the book describing how Jensen would spend hours learning about ray-tracing and understanding its importance
- Flat organizational hierarchy: He has 60+ reports. This is frankly insane to me, but he has a good point: the more V-shaped a company, the more likely internal politics clouds information. He wants the information as raw as possible and has zero patience for sugarcoating. This, paired with his tendency to get into the weeds, gives him an incredible mental model of how the company is moving
- Top 5 emails: He asks employees to send weekly emails describing the top 5 things that they are working on as well as any interesting trends. This enables him to understand trends before they happen and shift capacity to support promising leads
- Public shakedowns: He is very willing to publicly humiliate people. Perhaps humiliation is the wrong word to use, as Jensen uses these shakedowns to help junior employees understand what it takes to be a senior and what good leadership looks like I am quite impressed by how much time Jensen takes on organizational optimization. His paranoia and quest for truth is what makes NVIDIA so great.
Conclusion
Jensen Huang is only now becoming a household name, but what’s remarkable is how little his personality and views have shifted since he first started the graphics chip company in the 90s. His vision, knack for organizing the NVIDIA troops and relentless pace and work ethic is what makes him, in my eyes, one of the best CEOs of the last 20 years.