Huang Renxun's trillion-dollar AI empire: written before Nvidia's market value surpassed Apple

Source: AI NPC Liu Yubai

This Monday, Nvidia's stock price hit a new record high, with a market value of over 3.5 trillion US dollars, approaching the world's number one Apple.

Now that Nvidia is rising step by step, many people think that Huang Renxun is a "shovel-selling upstart" who has swept the world with AI chips. But to a large extent, there would be no AIGC today without Nvidia. Nvidia also laid the foundation and cultivated the soil for the AI ​​Tower of Babel that condenses the wisdom of all mankind.

1. Hinton + Fei-Fei Li + Huang Renxun, the iron triangle opens the spring of artificial intelligence

Let's first analyze from the perspective of time, along the way, why Nvidia has become today's upstart, or what unknown efforts and contributions it has made to become today's upstart.

12 years ago, Hinton, Fei-Fei Li and Nvidia CEO Huang Renxun went against the advice of many people around them and insisted on their own way for many years. They accidentally formed an iron triangle and contributed to the first spring of artificial intelligence.

‌ Jeffrey Hinton is the scientist who won the Nobel Prize in Physics a few days ago for training neural networks with physics principles. Fei-Fei Li is famous for creating the ImageNet database and is known as the "AI Godmother". Huang Renxun is the founder and CEO of the chip company Nvidia.

These three people were originally unrelated, but because they were alone in the world and stood at the top of their respective fields, they were connected by the spark of artificial intelligence.

What does it mean to be alone in the world? These three people are really stubborn.

Hinton never studied physics seriously. He wanted to figure out how the human brain works, so he had to move forward in an environment where neural networks were "dead" in academia. Fei-Fei Li's image database was also considered too complicated and an impossible task at the time. Not to mention Huang Renxun. When he lost money in 2006 to build the CUDA ecosystem, the entire company and investors tried their best to stop him.

Why were they so determined? The ideas of these three great people are as follows:

Hinton thought that if people can think with nerves, machines can also, so he developed a deep convolutional neural network. The algorithm was ready.

Li Feifei thought that if children can see and recognize pictures, machines can also do so, so he made the ImageNet library. The data was ready.

Huang Renxun thought that GPUs can process graphics and general computing, so he built the CUDA ecosystem. The computing power was ready.

We all know what happened later. In 2012, Hinton and his students used NVIDIA's GPU to train neural networks and won first place in Li Feifei's ImageNet image recognition competition, shocking the world.

A few years later, in 2015, OpenAI was established, and Huang Renxun also gave them a supercomputer for free. Musk, as one of the founders, accepted this gift. For the study of complex artificial intelligence technology, this is equivalent to giving a top-level Ferrari to a racer!

If OpenAI's release of GPT-3.5 at the end of 2022 is the iPhone moment of artificial intelligence, Huang Renxun is also an indispensable behind-the-scenes founder.

The torrent of history is mixed with the fate of great people rubbing shoulders with each other, and the story is often foreshadowed at the beginning. Great people are often lone wolves, and so are great companies.

The first thing NVIDIA did right was that it targeted graphics computing, which few people paid attention to at the time, at the beginning of its establishment, and took the lead from the start. When it was founded in 1993, NVIDIA was the pioneer of the 3D gaming PC market. Later, after years of practice in game graphics, NVIDIA's chips were able to provide enough computing power for neural networks and intelligently identify kittens and puppies.

The second thing NVIDIA did right was to expand the graphics chip ecosystem at a loss and attract customers from all over the world. Back then, Huang Renxun insisted on investing heavily in CUDA, but making money was a long way off. Today, the CUDA ecosystem with millions of developers has become NVIDIA's moat. (CUDA is like a "translator" and "accelerator" for GPUs. Programmers can use CUDA to write code in ordinary programming languages ​​such as C, which is then translated into a language that the GPU can understand and execute.)

The third point that Nvidia did right was to maintain a sense of crisis and steadily expand its business empire based on chips.Seeing that customers such as Google bought GPUs to do artificial intelligence, Nvidia itself began to make artificial intelligence platforms and software. Recently, Nvidia is even studying chip foundry to build capabilities that are not dependent on TSMC.

In the past decade, Nvidia has reduced computing costs by a full 100,000 times, and flat graphics computing has also developed into spatial intelligence.As Huang Renxun said recently, Nvidia's goal is to build a computing platform for this new world - the world of machine learning, the world of generative AI, and the world of AI with agent capabilities.

The difference between Huang Renxun and the other two scientists is that he is a businessman. Today, Hinton has joined an AI startup, committed to the safe and stable development of AI, and Fei-Fei Li has started a business to challenge space intelligence. But Huang Renxun is different. Nvidia's chips are equivalent to the brain power of artificial intelligence, and have penetrated into all walks of life around the world through intelligence - he already owns a trillion-dollar AI business empire.

Second, Nvidia's trillion-dollar AI empire, known as the "digital world power company"

The following figure is a summary of Nvidia's core customers and services based on network information.

The shocking point is that Nvidia's customers not only involve more than a dozen fields such as global games, professional visual design, data centers, artificial intelligence, self-driving cars, scientific research, medical imaging, financial services, content creation, cryptocurrency mining, education, manufacturing, retail, telecommunications and media entertainment, but also cover the enterprise side, consumer side and even the government side.

When artificial intelligence became the main battlefield of the global technology war, NVIDIA also received olive branches from many governments around the world. In Singapore, NVIDIA helped build a strong AI system; in Japan, they helped build a super smart AI infrastructure; Canada found NVIDIA to cooperate and wanted to use AI to change the world; the US government also pulled NVIDIA to work together on a global AI cooperation project to help less developed countries use AI technology...

Currently, almost all large models in the world are trained with NVIDIA graphics cards, and they have monopolized the AI ​​training chip market with a market share of more than 95%.

This is far from over.

NVIDIA's GPU and AI technology can also help smart home devices become smarter, make drones fly more steadily, and make robots more flexible. From health monitoring to retail management, from logistics optimization to environmental monitoring, Nvidia's technology is almost everywhere, and it can even help fashion designers try on clothes virtually and let sports teams analyze game strategies... In short, anywhere that requires powerful graphics processing or intelligent analysis may be Nvidia's next big stage. Nvidia's status as a "digital world power company" makes Apple and even global technology companies daunted. Nvidia's blessing is still to come. Looking closer, Wall Street analysts expect Nvidia's third-quarter earnings per share to reach $0.74, an increase of 84% year-on-year, and revenue to increase 83% to $33.1 billion. Looking further, according to International Business Strategies, a consulting firm that tracks industry data, the AI ​​chip market is expected to grow 99% in 2024 and another 74% in 2025. This data may be watered, but it is enough to make people stunned. Not only does Nvidia have unlimited potential, but the entire chip industry chain will take off.

Just last week, the market value of Taiwan's semiconductor foundry TSMC exceeded $1 trillion, becoming the seventh trillion-dollar technology company in history and the first non-US technology company. TSMC's market value on this day exceeded the sum of Tencent, Alibaba, and Pinduoduo. You taste it.

Three, written on the eve of the AI ​​productivity revolution

Last June, I did a vote on the official account. Which company would be the first in market value in five years among Microsoft, Apple, and Nvidia? I chose Nvidia (and I still choose it now).

Immediately afterwards, I saw in my circle of friends that the publisher was recruiting authors for "Huang Renxun". At that time, I was thinking that this book was worth writing even if I didn't pay. Because of the importance of computing power to the development of the Internet, it can be described as "stuck in the neck"! At present, it is much easier to deal with data and algorithms than computing power.

In the process of Internet development, we have successfully moved what happened in the real world to the digital world through text, pictures, and videos. The next step will be to enter the spatial intelligence that Fei-Fei Li, Apple VR, and others are building, which is commonly known as the metaverse. In this process, only by processing pictures can we process videos, and only by processing videos can we have the ability to build a three-dimensional digital world that can interact face to face and even run economic systems.

NVIDIA is in control of the main artery of AI development to a large extent, and in turn it is also very dependent on the development process of AI. At present, in addition to poor implementation and low profits, AI development is suspected to be another wave of bubbles, and it also faces many side effects of AIGC entering society for the first time.

For example, from short videos to AIGC, VR\AR, people will increasingly be unable to distinguish between reality and illusion; the personal abilities of many artists, white-collar workers, and entertainers are used as data to train AI, which in turn makes them unemployed; many people use AI generation technology to create gimmick content on a large scale to gain traffic...

Since September, I have been monitoring global AI trends every day. The good thing is that AI intelligence is being integrated into and supported in various fields such as medical care, scientific research, finance, aerospace, and marketing. The number of C-end users has also increased significantly, and the AI ​​smart assistant Doubao has been downloaded more than 100 million times; the bad thing is that there have been various AI scams, AI pornographic photos, large model infringement artists, and AI lawyer misjudgments...

These problems will affect the sustainable and healthy development of AI, and will further affect Nvidia's business. In addition, although Google, Amazon, OpenAI, and Huawei are all developing their own chips, the impact on the Nvidia empire is not that great.

Not only is the economy turbulent, but the Internet, a grass-roots team, is also more messy than before. We are now on the eve of the AI ​​productivity revolution. The economy, culture, politics and even values ​​will be slowly reshaped. There will be blood and gore, but there will also be opportunities.

What can we say about this era? Looking forward, it was the feudal dynasty and the arduous war of resistance. Looking back, it is the AI ​​productivity revolution. At least we will not starve or freeze. The country is relatively stable and there is no war. It is really fortunate. NVIDIA has also experienced many times of dying. Let's take a long-term view of history and magnify the future of mankind. Let's be a passer-by in the world and enjoy life.

Heroes emerge in troubled times. In addition to the rising NVIDIA, there are countless individuals shining.