How Nvidia Became a $1 Trillion Company

This video is brought to you by Aura Hi. Welcome to another episode of Cold Fusion 220%. That’s the astonishing increase in the share price of the california-based chip making giant. They’ve secured the position as the highest performing stock in the S&P 500 for 2023. Traditionally, NVIDIA is the company behind companies. It powers Netflix, Adobe, Airbnb, NASA, and even Kellogg’s cloud services. For those outside the tech space, NVIDIA is probably the largest company that you’ve never heard of. According to Amazon, more than 90% of Fortune 100 companies use AWS or Amazon Web Services, which is powered by Nvidia’s hardware today. NVIDIA helps power AI systems like ChatGPT, and partners with Google, too. It’s also used in Amazon’s robot warehouses. In summary, NVIDIA is a much bigger deal than most people realise. They’re not just a company that makes graphics cards for gaming, even though that’s what they’ve become known for. They’re selling shovels in a gold rush, providing the infrastructure for the digital age. And for that, today NVIDIA is worth over $1 trillion. While there are questions of whether a company with a $1.2 trillion market cap is part of a bubble, there is no doubt the influence NVIDIA has had on the wider industry. From very nearly going bankrupt in 1995, it’s been a remarkable turn around. In this episode we explore Nvidia’s extraordinary journey from technical feats to price controversies, unveiling the company’s true size and history, the 1990s. A decade of fuzzy TV, dial up Internet and beige boxes. Those of us who grew up in this decade also witnessed the evolution of the personal computer. In the early 90s, rendering a file browser window or chunks of text was about all the graphics that you’d need. CPUs were powerful enough for that. Unless you were in the animation or engineering industry, there was no need for complicated graphics. If you wanted that, video game consoles were the answer. But what if there was a separate hardware processor that could easily bring graphics to the personal computing world? Tucked away in sunny California, a trio of engineers, Jensen Huang, Chris Malachowski and Curtis Premium, were thinking the same thing. They would meet at a local Denny’s, where endless cups of coffee fuelled their conversations. At one of these meetings, the trio noticed something ACPU, which is the heart of the computer, is designed as a general purpose processor that can only tackle 1 task at a time, a method known as serial processing. At the time, generating 3D graphics for video games placed tremendous repetitive, math intensive demands on CPUs. If there was a separate chip that could do parallel processing, it would be a game changer. In other words, this chip would divide complex tasks into smaller ones and carry them out at the same time. This approach is far more efficient for tasks like rendering 3D graphics where many calculations can be done at the same time. This new chip wouldn’t replace the CPU, but would compliment it for graphical tasks. This would later be known as AGPU, and this idea was the beginning of NVIDIA. It was a simple idea with modern hindsight, but it would revolutionise not only computing but computer science over the next 20 years. But right now, Nvidia’s focus was on the gaming experience on PCs in the 90s, and that market had potential big potential. So in 1993, in a condo in Fremont, CA, NVIDIA was founded with the idea of applying the use of parallel processing GPUs in home computers. Fun fact, the name NVIDIA is actually the port mentor of envy, short for next version and NVIDIA, which means ENVY in Latin. You’ll notice all the green on their logos. This represents the envy which they hoped everyone would feel due to the computing power their chip would generate. Before we continue, let’s hear a quick word from today’s sponsor Aura. Every day, countless people are affected by data brokers. They’re making a fortune by selling your information to robo callers and scammers. They can know a lot about you, including your name, phone number, and address. But how could this be stopped? This is where today’s sponsor Aura comes in. They can identify data brokers and submit opt out requests on your behalf. Brokers are legally required to remove your information if you ask them, but they make it notoriously hard to do. You can let Aura handle that for you. Aura does so much more to protect you and your family from online threats. It’s easy to set up so you don’t have to download several different apps to get things like parental controls, antivirus, AVPN, password management, identity theft, insurance, and more. You get everything all at once at an affordable price. Visit aura.com/coldfusion to get started on a two week free trial. Thank you. All right, back to the video. Nvidia’s founders were top engineers. Jensen Huang, A Taiwanese American electrical engineer, had a track record as the director of core Ware at LSI Logic and a microprocessor designer at AMD. Remember the name LSI Logic? This company would prove pivotal to NVIDIA making it off the ground at all. The two other founders were also capable. Chris Malachowski brought engineering expertise from his time at HP and Sun Microsystems, and Curtis Premium was a former graphics chip designer at IBM and Sun Microsystems. Though being a qualified engineer is one thing, but launching a startup is another, and this is something that Nvidia’s founders learnt first hand. In fact, at first they weren’t even sure what to do, so they hired a lawyer to come to one of their meetings in 1993. The first step was incorporating the company. To get the ball rolling, they needed some initial capital to value the company’s shares. Jenson Huang had $200 in his pocket, which he handed to the lawyer. This modest sum not only got the company incorporated, but also earned Huang a 20% stake in NVIDIA. The next step was to get the company funded. This process, however, was not easy. As Nvidia’s founders soon learned, venture capitalists weren’t interested in polished business plans. Instead, they banked on their founders reputation, past success, and a grand vision big enough to be worth the risk. NVIDIA would eventually secure $20 million in funding from Sequoia Capital and Sutter Hill Ventures. To an outsider, this seems absurd. How could a brand new company secure $20 million in funding just like that? In reality, it was Jensen Huang’s previous connection to LSI Logic, the company that I mentioned earlier, that made the funding possible. Jenson would reach out to LSI CEO who would arrange a meeting with Sequoia Capital. Sequoia had previously invested in LSI Logic, which delivered a record-breaking $150 million return through its IPO. With Huang’s reputation with an LSI, Sequoia was willing to listen when he came knocking. At first Sequoia’s founder was sceptical about investing in a graphics card company, but they eventually took the gamble and invested an initial sum of $2,000,000 followed by a further 18 million. What I’m going to say next is some key context here. This was a very risky decision at the time. There were 89 other companies with similar ambitions to NVIDIA, but the fact of the matter is, except for AMD and NVIDIA, none of these other companies would survive. When NVIDIA went public in 1999, it was valued at 600 million, so the risk of the initial investment paid off handsomely, multiplied by 100 times handsomely. But we’re getting ahead of ourselves. At this stage in the story, NVIDIA was just a small group of engineers. After receiving funding, it took NVIDIA 2 years to build a team and create their first product, the NV One, which was officially launched in 1995. During the development of the NV One chip, NVIDIA secured a deal with Sega. This was to manufacture NV One chips for their gaming consoles. These chips were intended for games such as Virtua Fighter and Daytona. The thing was, NV One chips were also available for computers, so that means you could play Sega Saturn games natively on your PC in 1995. For context, the first PlayStation only came out a year earlier. Till this day, I don’t think I’ve seen anything similar done since. It’s like if Sony was to build a computer graphics card and you could slot that into your desktop and play PS5 games natively. But there was a problem though. NVIDIA chose to use a quadrangle based image rendering architecture for their NV One chips and Sega was not pleased with this. Regardless, Nvidia’s leadership refused to change their decision. So the question is, why did they choose to render polygons and quadratics instead of triangles? Well, that choice is explained in this video by Lazy Game Reviews. This uses quads or squares instead of triangles to speed up rendering by reducing the CPU workload. At least in theory, this results in fewer polygons and renders better rounded objects. So you could actually make things with squares instead of just a bunch of triangles everywhere. And at the time, with a lot lower speeds of CPUs and memory clocks and all that stuff, it made sense to use squares instead of triangles, at least in some situations, to get better rounded objects or models. In your game, because triangles you had more processing power required. Now, as memory got better and better, this made little sense. And it also kind of made little sense at the time because this meant that there was no support for Open GL. They didn’t even want to include it. I read a statement by NVIDIA and STS Thompson saying that they didn’t think consumers would need Open GL at all, so they just didn’t bother with it. And then by the time Direct 3D came along and DirectX, this couldn’t do that either, because it didn’t support rendering through quads, only through traditional triangles and those types of polygons. They did release a driver or a patch for it to update it to run DirectX, but only through software. So you’re effectively running a hardware accelerator in software mode to run hardware accelerator DirectX Direct 3D stuff. It made no sense. Despite the innovation, the project was an abject failure. It was designed to be a Jack of all trades, combining 3D graphics, video and audio processing, and a host of other functions into a single chip. As Jensen Huang would later put it, it was an octopus, Complicated, multifaceted, and perhaps two ahead of its time. The end result was a failure to capture the market’s attention. What the market did want was something much simpler. They wanted a dedicated 3D graphics chip. They didn’t want an all in one solution like the NV One chip. Aside from that, it was overpriced compared to the competition. To make matters worse, the wider industry was moving in a different direction. Microsoft had recently introduced Directxa new API that favoured a triangle based standard for 3D graphics rendering. It made coding for graphics easier for developers, so they naturally flocked to it. This shift meant that the NV One chip was incompatible with DirectX. As a result, many existing games suffered from performance issues or were outright unplayable on Nvidia’s new chip. The incompatibility was a significant blow to NVIDIA. Their customer partner, Diamond Multimedia returned the vast majority of the 250,000 MV 1 units they had purchased. They cited poor sales and a lack of game support. This return nearly bankrupted NVIDIA, forcing them to lay off 40 of their eighty employees and leaving them with a stockpile of unsellable chips. In essence, it was a $10 million disaster. Amidst the turmoil, NVIDIA was in the process of developing the NV Two chip for Sega, but it too was based on the same outdated quadrangle architecture. NVIDIA now knew that they had to pivot to survive. Jensen Huang had a major problem here. In a Hail Mary move, he approached Sega’s CEO with a request. Let us be released from the contract, but with full payment. NVIDIA desperately needed this money to keep the company afloat. In an act of extraordinary kindness, Sega’s CEO agreed to Huang’s request. This act of generosity allowed NVIDIA to redirect its focus and resources towards a path that aligned with the industry’s direction. If Sega pulled out their investment at that time, NVIDIA would have ceased to exist. Overall, the NV One’s failure became a critical learning experience for NVIDIA. They realised the need to understand customer needs and market trends. This shaped NVIDIA strategy moving forward. I think it’s pretty wild how close NVIDIA came to failing and now being worth over a trillion dollars. Making the decision to listen to the market and focus on PCs was the start of their success. After their initial failure, NVIDIA chose to launch into the PC industry, this time with a straightforward product. The PC market was booming and all they had to do was lie in weight and pounce at the right time. In 1999, NVIDIA entered the PC graphics market with the G Force 256 graphics card. This card wasn’t just a hit, it transformed the landscape as the first programmable graphics card. It also popularised the term GPU, or graphics processing unit. It allowed both regular gamers and developers to explore new possibilities with shading and lighting effects. This significantly enhanced the gaming experience. As NVIDIA rode this wave of success, the company went public in the same year. By 2000, NVIDIA had secured a major deal to develop graphics hardware for Microsoft’s Xbox game console. This deal gave NVIDIA a substantial $200 million advance, a far cry from the previous game console effort. With Sega. The Xbox would launch in 2001 with a custom console ready chip called the NV 2A. With this success as a launchpad, NVIDIA thrived throughout the 2000s. They played a key role in crafting Sony’s PlayStation 3 processors. Massive brands in the computer industry like Apple, Dell, and HP started using invidious GPUs. The company continued to grow from strength to strength. Let’s take a look at the strategies and technologies that got them there. One of Nvidia’s most important strategic decisions was to operate as a fabulous chip company. So what is a fabulous chip company? Well, fabulous chip companies design the chips, but they don’t physically manufacture them. Jensen Huang, NVIDIA CEO, was very particular about this decision, a decision to drastically keep costs down In invidious case, TSMC or Taiwan Semiconductor Manufacturing Company has been their manufacturing partner for the last three decades. TSMC is the world’s largest and most advanced semiconductor foundry. It provides manufacturing services for other big chip design companies including Apple, AMD, Qualcomm and others. TSMC played a major role in Invidia’s continuous success as they have the best manufacturing capabilities on Earth. Jensen Huang knows this well. In his own words quote, NVIDIA today would not be here if not for the pioneering work TSMC did. While Nvidia’s manufacturing was outsourced, the company could direct resources to innovation. They began to shift their focus beyond gaming chips and graphics. They realised GPUs and the parallel processing power they possessed could be used for so much more. NVIDIA now sought to give developers the tools and software to harness the full potential of the GPU. In 2006, NVIDIA launched a software toolkit called CUDA. Prior to CUDA, programming a GPU required a lot of tinkering and was a real headache for coders. It involved complex low level machine language or APIs. But with CUDA, it became much easier to write programs that could run on GPU’s using standard languages like C, C++ and Java. The impact many more applications, rather than just graphics, could use parallel processing power. It opened up possibilities for all sectors that required dealing with huge amounts of data. The trend started in 2008 when the Tokyo Institute of Technology used CUDA accelerated GPUs for their supercomputer to Bombay 1.01. Since then, many other data centres such as Amazon Web Services, Google Cloud Platform and Netflix have adopted GPUs for various applications. These include cloud computing, machine learning, video streaming and web search. But the real acceleration came about with the AI boom. 2012 is the year that many consider The Big Bang moment for AI. This was also the year that Nvidia’s innovations would be involved in a significant AI breakthrough. This breakthrough happened at Imagenet, an annual global competition where teams from around the world would compete to see how well their software could recognize objects and scenes and images. Alex Krzyzewski, APHD student at the University of Toronto at the time, participated in the competition with a novel approach. He leveraged the power of GPUs to enhance the computing efficiency of his deep learning neural network, Alexnet. Deep learning is a process where computers can learn by themselves without the help of programmers. By connecting to NVIDIA GeForce GTX5080 gaming cards to his computer, Krozewski was able to process Alex Net’s neural network with 1.2 million images. The result was remarkable. At the time, the image recognition accuracy showed a 15% error rate, down from 25% the year before. It’s laughable today. Tools like ChatGPT, Google Gemini and even Google Lens can do that now with much more accuracy. But back then it was revolutionary. It was this event that thawed the AI winter After decades of experts accepting that neural networks would never be practical. This proved them wrong. The NVIDIA GeForce GTX 580 cards were significant because they were optimised for CUDA, which of course allows for the implementation of parallel algorithms. They were crucial for deep learning tasks performed by Alexnet AI. Now the door had been swung wide open for many possibilities in deep learning applications and NVIDIA was ready for it. The strategic time and investment into CUDA and the software ecosystem around it was paying off. Nvidia’s GPUs are now the de facto standard for training and deploying generative AI and large language models. For example, Open AI, renowned for ChatGPT uses a 100 series GPUs, and this is for the rapid and widespread deployment of their AI models. After ChatGPT burst into the scene in November 2022. All of the major tech companies hoping to jump on the AI hype scrambled to get their hands on invidious GPUs. The A-100 GPUs were especially flying off the shelf. Microsoft alone was using a whopping 10,000 of them. One single, a 100 unit, would cost $10,000, and NVIDIA was selling hundreds of thousands of units. As a result of this frenzy, Nvidia’s sales soared to $13.5 billion in the quarter ending July 2023. That’s a staggering 101 percent increase from the year before. On May 25th, their stock obtained $184 billion in market value in just one day. And finally, on May 30th, 2023, NVIDIA became the 6th company in the world to reach that magic $1 trillion market cap, joining the likes of Apple, Microsoft, Amazon, Alphabet, and Tesla. At the time of writing in December 2023, the company is worth $1.15 trillion. Nvidia’s impact on the healthcare industry is also notable. IN2022A genomic sequencing technique by NVIDIA set a Guinness World Record for the fastest DNA sequencing, completing the feat in five hours and 2 minutes. Invidious chip technology has impacted further fields. The Tegra chips once designed for mobile phones, drive major e-commerce operations and self driving cars. Tesla used those chips to power their model threes from 2016 to 2019. Tesla now uses its own chips, but NVIDIA provides semi autonomous driving technologies for other automakers like Mercedes-Benz. So how was such a company run? How were pivots made so quickly to serve so many sectors with 110th the employees of Microsoft? Let’s take a quick detour into Nvidia’s management style, as I think it’s quite interesting. Jenson Huang, CEO, manages a team of 40 directly, and he believes in the power of a flat organizational structure. There’s no exclusive meetings with just VPS or directors anyone can join and pitch in. No career advice sessions either because as Huang puts it, his management team has already aced it. He also opts for a more dynamic approach by encouraging anyone in the company to shoot him their quote top five things on the fly. When it comes to information flow, everyone is in the know at all times. This ensures that information zips around quickly. There’s no planning cycle such as a grand five year or one year plan. It’s all about being agile and adopting to the ever evolving business and market conditions, especially in the rapidly changing field of AI. Jensen also has an eye for top notch talent and keeping the team lean, Though amidst the post COVID economic slump, major tech firms like Facebook, Amazon and Google dismissed thousands of workers. In contrast, NVIDIA pledged to avoid layoffs or salary reductions, promising raises for staff instead. Since their layoff disaster in 1995 with the Sega chip, Nvidia’s internal operations in regards to this has changed drastically. An example of this was in 2008 when invidious Tegra chip failed. NVIDIA was hoping to enter the mobile industry and created the Tegra smartphone chip. As a side there was actually a Tegra powered Android tablet called the Asus Transformer. I had one of these back in the day and I found the performance to be pretty underwhelming. Many others must have thought so too because the chip failed to gain traction and flopped. A team of more than 1000 engineers were responsible for the development of the Tegra series. Instead of getting rid of them after the chips failure, Jensen retained the team and encouraged them to explore new avenues. That team later collaborated with Nvidia’s existing automotive unit, and together they would create chips for self driving cars. So we’ve touched on a lot of what Nvidia’s involved in and what they’ve been doing, but there’s still one key sector that we need to talk about. Their original Focus, gaming. Although NVIDIA has diversified, they have not totally abandoned the gaming market. GPUs are still at the core of their business. In 2018, NVIDIA launched the RTX series of graphics cards, popularising a technology called ray tracing. Ray tracing is a demanding technique that simulates the paths of light. It creates more realistic shadows, reflections, and effects. Previously, this usually wasn’t a real time effect, It was only mainstream within the 3D arts and animation sectors. In Jensen Huang’s view, ray tracing was only possible because of invidious research in physics simulation and artificial intelligence. He explains that NVIDIA uses AI to generate 7 out of eight pixels in a scene based on one ray traced pixel. It’s like a jigsaw where the AI fills in the missing pieces. NVIDIA also popularised the AI upscaling of video games in real time, a process called DLSS. In theory, it allows for sharper, higher resolution games without the computational load, but in all of this there’s been an issue with Nvidia’s GPUs, and now we get into the controversies. The year is 2018 and the crypto mining boom is at its height. Miners rushed out to buy as many GPUs as they could to make some cash. As a result, demand for GPUs went through the roof and the supply dried up for those who normally use them. That is gamers. Prices soared by upwards of 71% in some cases. Amidst this surge, Nvidia’s gaming revenue increased by 26% to 1.65 billion. NVIDIA would tout that the initial 2018 growth in revenue was caused by the gaming market and they downplayed the effects of the crypto boom, the real cause of the revenue boost. With this wording, investors ability to accurately evaluate the company’s performance was compromised. Their unknowingly risky investments eventually crashed along with the crypto crash. This incident attracted the attention of the US Securities and Exchange Commission and NVIDIA was charged with unlawfully shrouding information. NVIDIA would settle in 2022 by paying a $5.5 million penalty, and they didn’t admit any guilt or wrongdoing, but they did agree to improve their revenue source disclosures. In response to this, in 2021 NVIDIA decided to introduce a dedicated cryptocurrency mining processor, or CMP line. This was supposedly to address the shortage in gaming cards, but at $4300 per CMP, crypto miners would just rather buy regular gaming graphics card. Jenson Huang is the only CEO of a tech giant who has been at the helm since the company’s inception. It’s been 30 years, and his long and successful leadership has also attracted some criticism over the years. One of the most controversial aspects of his management style is his relentless pursuit of profit, which sometimes borders on greed. After being fined by the SEC, the lack of transparency left many investors feeling misled. And aside from this, the gaming community in particular is not thrilled for shelving out so much for their graphics cards. There’s an unmissable chatter online conveying that NVIDIA cards are just too overpriced, especially in the mid range sector where AMD are more competitive. A notable incident that brought Huang’s business ethics into question was the 2022 fallout would EVGA. EVGA was a key partner in producing Nvidia’s GPUs. They chose to sever ties with NVIDIA and withdraw from the GPU market. This decision, according to EVGACEO Andrew Hahn, was rooted in principle, stemming from Nvidia’s unsatisfactory communication about contentious pricing strategies. Such controversies have cast a shadow over Huang’s leadership, challenging the integrity of his profit driven approach. Maybe to emphasise the challenges, Huang recently stated in an interview that if he was allowed to go back in time he would choose to not relive this career path. Starting and running a company the size of NVIDIA was a headache. He says that he faced many struggles and hardships as an entrepreneur, all the way from laying off those employees with the NV one chip back in 1995 to competing with bigger rivals and dealing with the constant pressure of uncertainty. In his view, nobody in their right mind would want to endure that pressure. If you were magically 30 years old again today, in 2023, and you were going to Denny’s with your two best friends, who are the two smartest people you know, and you’re talking about starting a company, what are you talking about starting? I wouldn’t do it. I know. And the reason for that is really quite simple. Ignoring the company that we would start. First of all, I’m not exactly sure the reason why I wouldn’t do it. And it goes back to why it’s so hard is building a company and building a video turned out to have been a million times harder than I expected it to be, any of us expected it to be. And at that time, if we realized the pain and suffering and just how vulnerable you you’re going to feel and the challenges that you’re going to endure, the embarrassment and the shame and you know the list of all the things that that go wrong. I don’t think anybody would start a company. Nobody in their right mind would do it, and I think that that’s NVIDIA has seen massive growth over the past two years. They were just in the right place at the right time. The share price has increased by 220% in the last year and revenue growth 206%. But will it last? Opinions are split. There are two main viewpoints. One is that AI will continue to be a massive industry and it’s only just getting started. This viewpoint believes that NVIDIA will keep riding this wave. They’re currently going all in on AI research, publishing paper after paper that are complete bangers. The other view is that this growth is unsustainable. As it turns out, NVIDIA isn’t going to be the only player in the GPU slash AI hardware space. Both Amazon and Microsoft are looking to make their own custom based chips on the desktop computing space. Microsoft has entered the arena with their Arc series. AMD is also gaining ground. Furthermore, Nvidia’s heavy reliance on Taiwan based TSMC for it’s semiconductor chip production may prove problematic. This is due to China’s increased military pressure on Taiwan. But this isn’t just for NVIDIA, but for the entire global semiconductor supply chain. There are now more threats in this chip space to NVIDIA today than there have been in many years, so we’ll need to wait and see what happens. So there you have it, the NVIDIA story. So far it’s been quite a journey, so I hope that you found it enjoyable and picked up some new insights. Currently NVIDIA is riding on an all time high, but whether this is a pivotal trend or a bubble remains to be seen. I’m curious to hear your thoughts. Do you have an NVIDIA card and what’s your opinion on the company, positive, neutral or negative? And were you surprised by the amount of things they have their hands in? Feel free to discuss in the comments section below. Alright, so that’s about it from me. Thanks for watching. My name is Digogo and you’ve been watching Cold Fusion, and I’ll see you again soon for the next episode. Cheers, guys. Have a good one, Cold Fusion. It’s new thinking.