Nvidia, A company founded by Jensen Huang in the year 1993. It was the birth of a legend, only to be known years later while having a market cap of 4 trillion dollars and being the most valued company in the world. But how did all this start? A chip, a chip which would later shape everything digital in the world.
Essentially, a chip is the brain of everything, performing millions of calculations to enable you to use your phone, play a video game, or even zoom in on Google Maps. This is all the doings of one single chip!! But, in 1993, when the company was founded, all of this was not very obvious. Why?
Because in 1993, when the PC revolution had just begun, everybody would use big, bulky and boring PC’s, which were easily processed by any chip. In the midst of this mundane world, where everyone was using computers solely for business spreadsheets, business plans, and pie charts on how to increase sales next quarter. The world was introduced to the new concept of video games.
So, after games were introduced. Games went from flat pixels to 3D worlds. This is exactly where the world fell in love with games like Doom and Quake. In the middle of all this fun, Jensen and his 2 co-founders noticed 2 Major problems in the video gaming industry. Firstly, they observed that to play these amazing games, you needed a separate console like a Super Nintendo or a Sega Genesis. Secondly, the computers just couldn’t handle these games at all!!
They didn’t have specialised chips that could run high-speed, high-quality graphics because these chips needed to do billions of calculations every single second to make you play a simple game. For Example, in 1996, Mario 64 required 100 million calculations every second, and Minecraft, launched in 2011, required 100 billion calculations every second and games like Call of Duty and Cyberpunk 2077 require 36 trillion calculations every second.
But the problem was that back in 1995, Everyone laughed at games, thinking it to be a Kiddish domain. In fact, as I mentioned earlier, the whole world was busy chasing business software, servers and spreadsheets, and because of all this, VCs turned game companies away, Engineers did not want to join gaming companies, and the whole industry saw it as a complete waste of talent. But then, why was Jensen so crazy that he wanted to build specialised chips for the games nobody wanted to play? Well, this reminds me of something very wise that Anupam Mittal said. He said, “A Great Entrepreneur is defined by the most non-obvious insight he has of the industry that nobody else believes in”
In this case, while everyone discarded games. Jensen saw something the world hadn’t realised yet. When others saw fun in games, Jensen realised that if he could build a chip that could run games smoothly, they could build chips for anything. If his chips could survive the chaos in a video game, they could be used for anything in any field, from powering robots to playing movies to even launching rockets into space. This is how, in 1995
With this exact understanding, Nvidia launched the NV1, Sound Blaster-compatible audio systems and 15-pin joystick parts and would be compatible with the Sega Saturn Console that was just launching. This chip was planned to be incredible and could handle graphics and sound with gaming controls combined. The Sega Saturn Console and NV1 were meant to be the best combination possible.
In 1995, just around the time the NV1 was launched. Microsoft launched something called DirectX, which was a combination of a set of rules that were specially developed for game developers. So, basically, Microsoft’s DirectX explained that if you wanted your games to work on Windows PC’s (which were being used by most game developers), you would have to follow these instructions. The first rule of DirectX was NV1’s reason to fail miserably. The first rule said that the game developers had to make everything out of triangles, which fundamentally meant that in Call of Duty, there are weapons and a battlefield and many more things. So, the fundamental shapes to draw the weapon, the battlefield field and all the other components that form the basic structure of the graphics have to be drawn only out of triangles. These are called low-poly models. Nvidia said, Let’s not use low-poly models. Instead, let’s use Quadratic Surface Textures, which means that everything could be created out of curved lines and could be made smoother. Microsoft denied the Quadratic surface Textures and said that if you wanted your chips to work on a Windows PC, you would only have to use low-poly models. Out of 250,000 units sold of the NV1, 249,000 came back from the distributors. After that, Nvidia had to beg Sega to break the contract, and Nvidia was left with 30 days before going bankrupt.
Jensen gathered his team, and they decided to make a chip-compliant policy, but they would not test it on hardware; they would only test it virtually. This was like suicide for Nvidia. In the chip industry, it is mandatory to test the hardware chip. It’s basically saying that you make a rocket and fly it virtually, and then say the rocket will work in real life. If you want to manufacture a chip, you cannot ask for 1 from the manufacturer, you will have to create a minimum of 100,000 chips without testing, which meant that if the software lagged by a nano-second then nvidia just mass produced 100,000 bugged chips and they would be finished, but having the whole industry against him, waiting for 8 long weeks for the product to come. Jensen did the miracle; he made a chip compliant with DirectX’s policies in record time without testing. This was the miracle of Riva 128. The Riva 128 worked seamlessly, and developers loved it; it sold a record 1 million units in just 4 months.
Jensen did not stop there. In 1999, Nvidia launched a chip called Nvidia GeForce 256 DDR, the world’s first GPU or Graphics Processing Unit. This may sound complex, but I’ll explain it simply. So, until the GeForce 256 DD, all the chips in the world were CPUs or Central Processing Units, which could only handle basic tasks and could not perform complicated tasks, whereas a GPU or Graphics Processing Units could handle tens of thousands of calculations per second, which allows them to handle extremely complicated tasks. This is where NVIDIA started the parallel processing revolution. But, this is not the end, Jensen realised that if this GPU could make games look better, it could also calculate a landing on Mars, imitate how cells behave in the human body, or even make a robot think like a human. Jensen could use this chip as a brain for AI, Medicine, Rockets and Games. This is when Nvidia stopped being a chip company and started becoming the future of computing.
So, in 2006, Jensen put down all his profits to make something that did not even have a market yet!! Nvidia built CUDA, or Compute Unified Device Architecture, which is a software which unlocked the full power of NVIDIA GPUs. So before CUDA, NVIDIA’s chips could only be used for graphics, but with CUDA, a scientist can simulate disease outbreaks, a self-driving car can see a child crossing the road and stop. NASA could calculate a rocket’s path to Mars in real time. So, CUDA unleashed the full power of Nvidia chips and changed them from gaming chips to life-changing mechanisms. So, regarding all this, you must be thinking that in a few years after CUDA, Nvidia would have become a trillion-dollar company, NO!! Because CUDA was 10 years ahead of its time, it’s like building a rocket engine before humans wanted to go to space. Even AI was not there, no ChatGPT, no Gemini. CUDA was too early, and AI research was still very primitive
So, for 6 long years, nothing happened, and then again a miracle happened! In 2012, in a lab in Toronto, young researchers named Alex and Ilya, with their mentor Geoffrey Hinton, had a dream: they wanted to build a computer brain that could see like a human. So, if you gave the computer an image, they wanted the computer to tell you if there was a dog or a cat, a truck or a car, a hospital or a battlefield, but they did not use a supercomputer. They used a nvidia chip which was powered by CUDA, So when the computer was faced with million of images of faces, animals, places and countries, The GPU of CUDA did not just understand the images, it started learning and eventually the computer learnt how to see and this AI model was called Alexnet and entered the biggest AI competition in the world on the 30th of September 2012. Alexnet did not just win the competition but crushed its competition. This changed the way people saw AI. Alexnet proved that AI can learn if you give it enough data, and that CUDA was the GPU that made the team win.
After Alexnet, everybody realised what Jensen had realised 6 years ago, and this began the AI revolution in the world. Google started using Generative AI in its search engine, Facebook improved face recognition OpenAI built ChatGPT. But wait, all these companies had 2 things in common: they all used Nvidia GPU’s and they all used Nvidia’s CUDA. So after all these giants started using Nvidia GPUs, Nvidia became a 4 trillion dollar company, and that’s how Jensen Huang turned a company 30 days away from bankruptcy into a 4 trillion dollar miracle and became the most powerful company in the world.
