Why Nvidia could continue to win and why semiconductor design companies are low quality businesses

Disclaimer: Nothing on this site, including this page, constitutes investing advice. Please do your own due diligence.

In 2010, Jensen Huang, CEO of Nvidia stood in front of me and ~50 other young, naive engineers, foot on a chair. He looked like a genius, swashbuckling, Asian Captain Morgan and proceeded to recount Nvidia history. There are likely very few founding teams that love their companies like the Nvidia founding team. As anyone who’s started a company would tell you, that’s saying something. All startups require blood, sweat, and tears. But Nvidia was born in a competitive era like no other. The future of graphics was being invented. By ATI, S3, Rendition, Chromatic Research, Cirrus Logic, 3dfx, you get the picture. It was an absolute bloodbath of some of the world's most brilliant people, all vying for a piece of the pie. There were really only two champions who won the war: ATI and Nvidia. The James Bond Skyfall speech describing Silva and Bond comes to mind.

How many cofounding teams do you know of tattoo their arms with the company logo?

During that time, I wanted to be a chip designer. However, as this was shortly after the Great Recession, the launch of the iPhone (2007), and the maturation of the Internet, software engineers were in demand. Also, quite frankly, I’m an awful electrical engineer but a very good software engineer. So I ended up interning at Nvidia’s working on the then-newish Tegra chip on software infrastructure. The Tegra was Nvidia’s answer to processor demands in the mobile boom as well as the evolution of vehicular software needs. That time at Nvidia, to this day, remains an ephemeral memory, my favourite professional experience that probably can’t exist today. Back then, engineers were happy to get a free fountain drink that costs a few cents (my favorite at Nvidia was Mountain Dew). Today, software engineers will complain if their Michelin star made free pasta is al dente enough. Back then, you had to know everything because there was very little infrastructure out in the wild. Today, anyone with a few months of experience calls themselves a senior engineer. 

Nvidia was very much a hero culture. Because Nvidia was borne out of an incredibly intense and bloody era, breakthroughs were of paramount importance. Some of this was common to all of Silicon Valley, but Nvidia, in particular, gave roles/responsibilities according to engineering prowess. What this meant was, brilliant engineers had the influence. Brilliant engineers made sure the best designs and thoughts were expressed through business strategy. But brilliant engineers don’t always make for great managers. The culture was brash. Feelings got hurt. Stupidity wasn’t tolerated. Wins were celebrated. It was the first time in my personal life that I realized there were other people who obsessed about the semantics of programming languages and the Linux kernel as much as I did. It was a meritocratic culture of passionate people who cared about things that most people will never know about. I loved and miss that culture today. 

The graphics war isn’t the only war in semiconductor history. The industry is littered with dead bodies. The memory war had Elpida, NEC, Hitachi, Qimonda, Hynix, Micron, and so on. Three winners emerged in Samsung, SK Hynix, and Micron. The x86 CPU war had Transmeta, Cyrix, National Semiconductor, IBM, etc. Today we have just Intel and AMD. Intel’s had some misexecution where if continued could one day result in AMD being the last man standing. There is a question today of if the idea of an x86 processor is even relevant since the likes of Google, Amazon, Apple and so on are producing ARM based alternatives.

Every day in the semiconductor industry, history is made. A new architecture. New tools. New processes. And so on. They involve a disparate number of fields including semiconductor physics, optics, ASIC engineering, digital verification, and so on. And yet, the broader world will probably never know or understand what those breakthroughs are. A company without breakthroughs and innovation is a company with a death sentence. So in many ways, semiconductors are a perpetual race that’s run at breakneck speed by some of the smartest people humanity has to offer. 

Back in the day, the Nvidia Tegra was important to the company. Jensen, himself, is not only an engineer’s engineer, but a general who leads from the front. So while the Tegra team was in the trenches, he was there as well. He rolled up his sleeves, sat where the QA team sat and did QA. You can imagine, if the CEO who defined an industry is sitting down beside you doing QA, you had best believe that you had a motivated team. Who wants to be the engineer seen as introducing a bug that Jensen found or to be seen moving at a pace unbefitting of Nvidia. Furthermore, you are being led into battle by a general who understands the very nature of what you’re working on. Jensen led by example, and as a result, leaders at Nvidia rolled up their sleeves. It was an all in effort. Contrast that against Intel, in 2018. Intel was led by Bob Swan, an MBA who has probably never touched an oscilloscope, probably has no idea what the linear region of a MOSFET is, and likely cannot comprehend what it is he is allocating capital towards. And that’s why I think Nvidia will continue to win. In an industry where the players, products, business war, and world moves so quickly, the only durable thing really is the culture and Nvidia was an intense one.

During the run on Silicon Valley Bank, I got curious and did some due diligence on banks as fear tends to create good investment opportunities. As an engineer, you can imagine, I quickly realize I am quickly out of my depth. I did, however, come across a portfolio manager who was a former bank analyst. When asked why they didn’t invest in banks, they responded “if you knew the nature of banks, you wouldn’t invest in them”. I feel the same way about semiconductor design companies. There probably never will be deep, impenetrable moats in the category like there are in credit ratings, scale advantaged business, and so on. This isn’t true of the entire industry. The electronic design automation (EDA) industry (Synopsys, Mentor Graphics, Cadence) is an oligopoly that is unlikely to be broken. So when I see the rush today of investors bidding Nvidia to unseen multiples, I have two thoughts: you can’t bet against Jensen/Nvidia but you also can’t be sure someone else won’t one day eat Nvidia’s lunch. Today, the war isn’t about graphics, it’s about AI. Google launched their Tensor Processing Unit for public use in 2018. Jensen’s familial relative, Lisa Su (CEO of AMD), is a formidable competitor as well. As a result, if I did have to invest in the industry, it would be in multiple companies where there is conviction that one of the players will win, rather than just one. But I don’t invest in any of them.

When I think about what goes into other industries, I think semiconductor design companies are awful businesses. Like energy or mining, the industry is cyclical. However, unlike energy or mining, the industry will likely forever be R&D intensive and cutthroat. An oil exploration & production (E&P) company can halt capex investment if they cannot justify the spend. A semiconductor design business is always at war and cannot stand still. A semiconductor business has the laws of physics to contend with. Moore’s Law could very well end, eventually eroding technical advantages over time. An E&P company just needs to offer a product below a price point and has full discretion of when it does or doesn’t make sense. Louis Vuitton can sell a piece of leather with a logo for 70%+ margins. For a semiconductor design company to achieve the same margins, either the product must be niche so others cannot justify the capex or the sun/stars must align. That is, there must be a correct prediction of market demand, on-cycle timing of product, competitive IP, brilliant engineers work day and night to deliver a differentiated product, and so on. 

Capitalism is a miracle in this respect. We, as consumers, get to buy a GPU/CPU/ASIC/etc. that is the culmination of enormous effort, brilliance, and sleepless nights. Somehow, akin to magic, sand is turned into silicon and that silicon has enabled a step change in human productivity and arguably the human experience. We get to buy that product for less than the monthly cost of groceries for a family of 4, way less than a French craftsman’s leather purse, less than a few months of a Starbucks addiction. So I would contend that semiconductor design companies are miracle makers, but hardly durable, high quality businesses. However, of these businesses, it is still hard to bet against Jensen who has crafted a brilliant company that has thrived in decades of cutthroat war.