Source: Sequoia Capital

Compiled by: Yuliya, PANews

Editor's Note: In the past, our data centers merely stored files for human retrieval; now, computing is moving towards generation, with every word, every image, and every video being generated in real time and highly customized based on the requester's context. In this global wave, Konstantine Buhler, a partner at Sequoia Capital, engaged in an in-depth conversation with Jensen Huang, founder and CEO of NVIDIA, to discuss the significant changes in computing technology. Huang believes that automation will not lead to unemployment, but rather a comprehensive increase in the demand for labor and an elevation of professions themselves; people will not lose their jobs because of AI, but may be replaced by those who skillfully utilize AI.

AI Factories and Generational Leap in Computing Models: From Retrieval to Generation

Constantine: Thank you very much for coming, Jensen Huang. We are in the midst of a massive AI revolution, one that may even surpass the Industrial Revolution in scale and speed. You've stated that what's happening now is the largest infrastructure construction in human history. At the heart of this construction is the AI factory, and the company enabling all of this is Nvidia. Can you tell us what an AI factory is? And why it's the most worthwhile investment for all businesses in the next decade?

Jensen Huang: You can understand AI in many ways. The most familiar way for the general public is probably through interacting with chatbots via web browsers: you give it a prompt, and it replies with a message. Even if you've been using AI for a while, you'll find that its capabilities have evolved remarkably over the past two or three years.

Two years ago, people heard about ChatGPT. Essentially, it's computer software that can understand the information you input. It can sense and understand information, and transform and generate other content. For example, you can give it a PDF file and ask it to summarize it—that's text-to-text; you can also ask it to generate an image based on a story—that's text-to-image; or give it a photo and ask it to describe the scene—that's image-to-text. This capability was called generative AI two years ago.

But beyond understanding and generation, the ability to think is even more valuable. The foundation of generative AI endows it with the ability to think internally, reason step by step, and solve problems. Moreover, it can now generate control commands to use tools, whether it's using digital tools such as browsers, spreadsheets, Photoshop, and AutoCAD, or in the future, to control mechanical systems (that is, robotics and self-driving cars).

Two years ago, people found ChatGPT amusing—it could write poetry and songs, but occasionally it would ramble on. Today, two years later, we have agentic systems. AI is no longer just about understanding information; it can now reason and do useful work. Because it can do useful work, AI generates real business value. We won't pay friends who only know how to talk big, but we will pay those who can actually do the work. People are now hiring AI by the hour, paying it $20 to $30 per hour. This is why it has become the fastest-growing software business in human history.

From the perspective of upstream industry logic, we need to return to first principles. The basic concepts of the computer industry as we know it today were established approximately 64 years ago . At that time, IBM launched the System/360, which is why IBM became the world's most valuable company at the time.

For the past 60 years, computing has essentially been about pre-recording and retrieval: you write a story, take a photo, record a video, and save it as a file on your hard drive; when you want to use it, you retrieve it from the hard drive. That's why those buildings are called data centers. They simply store data and don't do much computing.

But things have changed. In the AI era, every time you provide new background information (context) and a new request, AI will perform real-time understanding, reasoning, and generate entirely new results. For example, my current speech is generated in real-time based on the different backgrounds of everyone in the audience, rather than being a verbatim reading from a script. That's what we call intelligence.

In the future, every pixel, every sound, every video, and even every advertisement and news item will be completely generated and tailored to you, rather than being pre-recorded and then retrieved. This means that in the future we will need a large number of generators , which are the large computers we are building—this is the AI factory.

The intelligent network enveloping the earth and the generator of the digital age

Constantine: How large will this generator be?

Jensen Huang: Currently, we provide information and intelligent generation to approximately 1 billion people worldwide. But because AI has become agents, they can work independently, and one agent can even communicate and collaborate with another. Within NVIDIA, there are likely hundreds or even thousands of agents talking to each other and solving problems (of course, they all operate within secure sandboxes and guardrails).

This means that in the future, not only will humans be using the internet, but hundreds of billions of intelligent agents may be working tirelessly day and night on it. Enterprise agents, self-driving cars, robots, and even the systems within every building will all be communicating with each other. All instructions and all thoughts will be generated in real time.

It's like there's a thick computing network, like a cocoon enveloping the entire Earth. This sounds exaggerated, but it has actually happened twice in history:

The first instance was 300 years ago when Siemens in Germany created a machine. When you ignite it, it outputs a hidden, powerful force—electricity. Today, power grids (electric power grids) cover the entire Earth.
The second is the internet, which originated in the United States 35 years ago and now also encompasses global communications.

Now, we've entered the third network after energy and communications: the intelligent network. Nvidia's current core business is building this new-era generator (Dynamo). 300 years ago, generators input the physical movement (atoms) of water, wind, or coal and output electrons; Nvidia's machines, on the other hand, input electrons (electrical energy) and output digital data. These digital data, through different combinations, become language, mathematics, or the language of proteins and human biology, the language of physical laws and climate prediction, and even the language of the 3D world, robots, and autonomous driving.

These two machines, separated by 300 years, are remarkably similar: atoms go in, electrons go out; electrons go in, digital data comes out. These digital data are what we call tokens, or intelligence. We mass-produce these intelligent tokens in factories, and that's the significance of AI factories.

Constantine: We are in the midst of a confluence of multiple revolutions. From the energy transition and routers in global telecommunications networks to the GPUs and AI factories at the heart of the smart revolution, such as the H100 or the latest Vera Rubin architecture. It's about integrating everything that's needed.

Jensen Huang: Yes, our computing unit is called a "rack." There are 72 chips in one rack. This year we plan to manufacture approximately 8 million of these components. One rack weighs 2 tons, costs $4 million, and contains 1.5 million parts. It's the most expensive piece of equipment in the world, but we mass-produce them like we manufacture mobile phones and ship them to data centers around the world. These things are huge; moving them is definitely a physically demanding task.

The Five-Layer Cake Investment Logic for Participating in the AI Era

Constantine: This is a very exciting picture. How can we, whether large corporations or individuals, participate in this revolution?

Jensen Huang: Investing in the AI industry, you can think of its industrial layout as a five-layer cake. You know, a $50 billion AI factory can generate $300 to $400 billion worth of intelligence, and its return on investment is astonishing. So what are these five layers of cake?

The first layer is energy: the most basic element, the generator. This represents the biggest growth opportunity for the energy industry in generations. Sustainable energy (nuclear, wind, solar, hydrogen, etc.) will receive massive investment to support computing. Investment will be made wherever energy can be generated. This is why companies like Siemens, Mitsubishi, and GE Vernova are performing so well right now.

The second layer is chips and computing facilities (Chips/Computers): including chips, computers, network equipment, switches, and silicon photonics technology, etc.

The third layer is infrastructure: this includes land, electricity, building shells, funding, and the daily operations of data centers. Currently, these resources are in a state of extreme shortage.

The fourth layer is the Models layer: these are large models built on cloud infrastructure. This is the most market-driven and investment-intensive area in human history. Well-known examples include OpenAI and Anthropic. But remember, AI can learn more than just natural language; it can learn anything structured. We are learning the laws of the physical world—for example, I sat down with great confidence not because I had a 47% chance of falling through the chair, but because I had 100% faith in the laws of physics. AI can similarly learn the meaning of proteins, the significance of genes, and the function of cells. The physical world's industries are worth a staggering $80 trillion, a currently less discussed but extremely important frontier area.

The fifth layer is the Applications layer: based on the underlying technologies, countless startups are reshaping industries such as financial services, law, accounting, transportation, and logistics. Last year, venture capital invested $100 billion in this top layer, the highest amount of VC investment in human history.

This future will be incredibly vast. We're only at the beginning; this year alone, an estimated $1 trillion is being invested in this ecosystem. But I predict that AI will become a massive ecosystem worth $20 trillion annually. How important is intelligence? Who needs intelligence? How much do you need? Figure these things out, and you'll know where to invest.

AI isn't here to steal your jobs; it's here to help you advance your life.

Constantine: This is not only a multi-trillion dollar market opportunity, but also a massive explosion in hardware and applications, which will create a large number of real jobs for humanity.

Jensen Huang: Absolutely right, and we must emphasize this point. Every country and culture has a different attitude towards AI. But I sincerely advise everyone: be wary of the plots in those Hollywood science fiction movies. Don't keep listening to people saying things like "The Terminator is coming," "The technological singularity has arrived," or "There's a 20% chance AI will destroy humanity." That's complete nonsense.

Some have even threatened that "we have no idea how AI works; it's too mysterious, and maybe it will just disappear on its own tomorrow." This is utter nonsense. AI is just computers and software, and engineers certainly know how it works. Otherwise, how could they make it safer and smarter every year?

Modern AI has significantly reduced illusions; the knowledge it generates is accurate and context-aware, and it will even research information when it encounters something it doesn't understand. Before answering you, it might even reflect on its own actions, comparing several options before giving you the answer. Just as cars today are far safer than they were 100 years ago, the tech industry is making every effort to make AI extremely safe.

So focus your attention on what's certain. I'm pretty sure of one thing: you might not lose your job because of AI, but you will definitely lose your job to the person using AI.

Since this is a technology that can give humans superpowers, then go and use it! Tell your loved ones, your children, your company, or your country: embrace AI.

Constantine: But when it comes to work, everyone is indeed very anxious.

Jensen Huang: I get angry whenever I hear someone creating panic about jobs. This year we've invested $1 trillion into this ecosystem, including energy, chips, infrastructure, the model layer, and the application layer, all of which are creating far more jobs than ever before.

Some might ask, what about traditional positions? Here's a common misconception: people confuse "job" with "task."

For example, I'm a CEO. My daily "tasks" mainly involve typing and speaking. Now, AI is much better at typing and speaking than I am, it's superhuman, but as a CEO, I'm actually busier than before.

Let me give you a more profound example. About 12 years ago, a top computer scientist stepped forward to warn everyone that computer vision could tirelessly analyze medical images, missing not a single detail—it was already at a superhuman level. He asserted that the first profession to be eliminated by AI would be "radiologist," and advised everyone not to study that major.

His technical judgment was absolutely correct. Computer vision is now integrated into all radiology systems, and radiologists are using AI to assist them. But what's the result? The global demand for radiologists has actually increased!

Why? Because radiologists' purpose isn't just to interpret images, but to diagnose diseases alongside clinicians. Automation has greatly increased their efficiency, allowing hospitals to handle more patients on the waiting list, making radiology departments more profitable. Hospitals, seeing increased profits and more patients, end up hiring even more radiologists! Those who heeded the warnings and didn't study radiology have missed out on a crucial opportunity.

Similarly, some people have recently claimed that since AI can write code, 90% of software programming is obsolete, and we no longer need software engineers. However, the truth is, we are hiring more software engineers than ever before! This is because the purpose of software engineers is to solve problems and innovate, not to compete in typing speed. Writing code is merely a task; problem-solving is the core objective.

AI won't eliminate jobs; instead, it will enhance their value. If I'm a plumber today, I might just be following blueprints; but with AI tomorrow, I could also be a kitchen designer. If I'm a furniture seller or carpenter, in the past you only expected me to nail wood together, but with AI, I can provide you with a complete interior design plan, making your home incredibly beautiful. My professional skills have been elevated!

Therefore, I believe the current narrative that AI will lead to human unemployment is completely false; it's merely a ploy to scare people away so that the authorities can profit from it . Throughout my career, computer technology has become increasingly complex. In the past, only 2% of the population could program in C++ (perhaps those of you in Silicon Valley's venture capital circles know more about this). Now, because of AI, as long as you understand human language, you can program. For the first time, we have truly closed the technological gap, and we must lead everyone into this new era.

AI Factories and Generational Leap in Computing Models: From Retrieval to Generation

The intelligent network enveloping the earth and the generator of the digital age

The Five-Layer Cake Investment Logic for Participating in the AI ​​Era

AI isn't here to steal your jobs; it's here to help you advance your life.

The Five-Layer Cake Investment Logic for Participating in the AI Era

A dormant whale address that had been dormant for nine months withdrew over 404 BTC from a CEX.

Popular Articles