Editor's Note: As the tide of the AI revolution reshapes the global technology landscape, a chip known as "NVIDIA" is quietly driving this change.

In a recent in-depth dialogue with Mind Observer, Stephen Witt, author of "Jensen Huang: The Heart of NVIDIA," shared his insights into Huang Renxun's legendary rise from being an immigrant on the edge of Silicon Valley to becoming the "AI Guru."

[Interview/Observation Network Mind Observer]

Mind Observer: Let’s start with the preface of your book. There is a sentence in the preface that may provoke readers' thoughts. Huang Renxun said, "He will always consider himself part of the American immigrant population, and deep down he will always be Chinese." How do you interpret this self-identification?

Stephen Witt: I was surprised when he told me these words. The conversation took place at Danny's Restaurant (Note: Huang Renxun worked there as a dishwasher during his teenage years. In 1993, Huang Renxun and two other founders established NVIDIA at Danny's Restaurant). Huang Renxun is of Chinese descent, but he is not entirely a native Chinese person. I think if he were to move to mainland China or Taipei now, he would not adapt so well. To be honest, in some ways he has already become very Americanized; for example, he speaks fluent English, has lived in America for 50 years, and possesses the ability to view issues from an outsider's perspective.

Huang Renxun at Danny's Restaurant, where his career started as a "dishwasher"@SVBJ

In his life, the formation of personal character began at the age of 10 when he entered a boarding school in Kentucky, and he had a "hyper-realistic" experience, as I described in my book — suddenly arriving at a foreign boarding school. This indeed made him believe that he could always look at issues from the outside. This is how I understand his self-identification.

NVIDIA, of course, has very American characteristics, which is a company that allows employees to act swiftly, they like to break rules, and they are willing to face risks. It also has very Chinese-oriented aspects, such as employee work ethics, expectations for work methods, and organizational structure, etc.

Huang Renxun has always been focused on fulfilling personal responsibilities. He used to get perfect scores in various exams, but he is also not afraid to break rules.

The characteristic of Chinese people might be that they are excellent students in school, achieving high grades and working hard. The American trait is to act quickly and break conventions. Now Huang Renxun possesses both qualities of Chinese and American cultures, and I think this is precious.

Mind Observer: You also mentioned that sometimes it feels difficult to communicate with Huang Renxun. Is this due to his personality or related to NVIDIA's business?

Stephen Witt: Mainly because of his personality. Huang Renxun is straightforward and open, and he is also eager to play entertaining and fun things. So he makes jokes or says some humorous things just to keep the atmosphere light and enjoyable, allowing the spotlight to continue shining on him. Therefore, he is an interesting interviewee.

He is constantly thinking deeply, and often does not particularly attempt to convey any implied meaning, so you don't know what he will say next. I think if you are an NVIDIA executive, you often cannot predict what he will say next. This is why those around him may find him unpredictable in mood.

Even when he is in a bad mood, including during low points, he is still on point in managing the company.

Mind Observer: Do you think his way of thinking is more like that of a front-line engineer rather than a businessman.

Stephen Witt: Yes. A close friend of his once told me that Huang Renxun is an excellent engineer, a world-class computer scientist, yet he actually has no background in commercial disciplines. I'm not sure if he has ever taken even one commercial course, but he learns quickly. I think if you are an engineer, learning commerce is relatively easy; but conversely, if you are a businessman, learning engineering knowledge is relatively much harder. Therefore, I think Huang Renxun's engineering background is very helpful to him.

I think he is also an architect, an inventor, and a logic chip engineer. As the helm of NVIDIA, he must have such qualities and DNA. His traits can even be traced back before he founded NVIDIA.

He has always been good at providing and selling products to the most demanding and advanced customers. For example, he would pitch to a small tech team by saying, "How can I build a computer for you? This will greatly increase your productivity and make your work efficiency 1000 times better." He would even take the risk of losing money.

So it seems he is more like an engineer than a businessman because he is very interested in these extremely complex and difficult engineering problems. These engineering problems he pursues may not bring immediate profits or returns in the short term. From a long-term perspective, this has proven to be a brilliant and great business strategy.

For a long time, many people outside questioned whether Huang Renxun was only focused on engineering and ignored customer business needs.

Mind Observer: Over the past few years, NVIDIA has implemented several quite successful mergers and acquisitions. Looking back, these mergers have had a huge driving force for the company's development. Mergers involve a lot of financial knowledge, so he needs some assistants who understand financial operations around him.

Stephen Witt: Colette Kress, NVIDIA's CFO, has strong capabilities, and Huang Renxun communicates very well with her.

Huang's "secret technique" is that he can use the professional language of the best computer scientists and astronomers. All of NVIDIA's achievements today are the result of his dialogues with world-leading technical experts.

Mind Observer: You also mentioned some of Huang's key lieutenants in your book, such as John Nickolls and Curtis Priem, engineers who have achieved remarkable feats in NVIDIA's development history. Some of them have left. However, how does Huang identify and retain top talent within his team? How does he build a high-quality R&D team?

Stephen Witt: That's a good question. Huang Renxun always makes a list of the most talented engineers. When NVIDIA first developed 3D graphics rendering technology, they listed all the people they wanted to train and recruit on the whiteboard in his office – these were all the people they wanted to poach from other 3D graphics companies. They would persuade these people to join them.

Part of the reason is that Huang Renxun himself is very charismatic, and excellent engineers want to follow him. For example, Bill Dally, who was once the director of the Computer Science Department at Stanford University, Stanford University is one of the leading research institutions in the world, and he may be one of the most prestigious figures in the scientific circles.

Bill Dally (left) at NVIDIA's GTC conference this year, talking with Meta Chief Scientist Yann LeCun

In 2009, he resigned from that position to work at NVIDIA, when NVIDIA's stock price had plummeted, and their newly developed graphics card looked like it was about to fail. But Bill Dally still chose to leave Stanford to join NVIDIA. At that time, I wondered, why did he do that? What motivated him to leave academia and go to a company that treated him like family? And at that time, he just thought Huang Renxun was a natural leader, and wherever he went, you wanted to follow him.

Bill Dally also received job offers from other companies, including Intel. Despite Intel being 100 times bigger than NVIDIA, he refused Intel just to work with Huang Renxun.

Huang Renxun's ambition is to find the hardest problems in the world and solve them. Intel may have done that in the past, but now it doesn't. Huang Renxun seeks out the top academic computer scientists in the world, solves the hardest problems, and builds a system for them.

If you are an inventor, you need to become a cutting-edge technical expert, just like the greatest car mechanic wanting to work on Formula One cars, always wanting to stay at the forefront globally. This is why these people work at NVIDIA, and this is why NVIDIA can attract the best engineers in the world.

Mind Observer: Huang Renxun has a unique management method, such as listening to direct reports from thirty to forty people every day, and asking employees to write three important things via email, etc. Why is this management method difficult to replicate in other high-tech companies?

Stephen Witt: Very few people have the ability to manage information flow like Huang Renxun. Managing fifty or sixty people directly requires a very clear mind and composure, and most people would feel confused and unable to handle it. I think I couldn't do it, and I want to say that most people can't, they don't have enough brainpower.

NVIDIA's executives are all excellent. Many of them could be outstanding CEOs in other companies. But because they work with Huang Renxun, they earn more at NVIDIA than they would as CEOs elsewhere. Moreover, Huang Renxun gives them a lot of trust. He would give these executives quite relaxed work instructions. He would say, "Go ahead and do this, report back in six months." He believes that person will complete the task without any supervision or oversight.

All the people around him are excellent, and he fully trusts that they will work hard to provide critical products and decisions without frequent supervision. I think this management style will become more common with the advancement of AI technology.

For the CEO themselves, the most difficult part is how to anticipate the visibility of the company, assess whether goals and instructions are effectively implemented, etc., such as how do I know if they are being effectively executed?

I think AI will make CEOs clearer about whether the instructions are happening and how fast, which will make managing teams easier. CEOs will directly lead more and more people.

Sam Altman, president of OpenAI, once predicted that in the future you will have a company with only 20 employees but revenue reaching $1 trillion. So the future form of companies will be the CEO plus direct employees, managing millions of AI agents. In that future AI-driven world, this management style may become more common.

Mind Observer: NVIDIA's rise benefits from the globalization system. Globalization has made the flow of talent smoother. Huang Renxun and Lisa Su, president of AMD, are representatives of American immigrants, but now the Trump administration is preparing to move against globalization and may modify the H-1B visa policy. What impact will this have on NVIDIA's global talent recruitment?

Stephen Witt: This is bad news for the U.S. economy and the entire Silicon Valley. America's competitive advantage lies in recruiting top engineers from all over the world. They come from India, China, Japan, and the Middle East, usually the top graduates from their respective schools, which is the source of Silicon Valley's development. If you walk into Silicon Valley, you will find that most of them are from overseas.

So if the visa policies are modified or stopped, this will be very bad for Silicon Valley. Similarly, I think tariff policies are detrimental to the U.S. economy. Global trade benefits all participants, while the tariffs imposed by Trump are, in my opinion, foolish.

Mind Observer: Have you bought NVIDIA stocks?

Stephen Witt: That's interesting, I should have written this experience into the book. Back in 2002, I bought some NVIDIA stocks when they were still a gaming GPU company. The stock price dropped for a while, and then I was very frustrated. Around 2005, it rose again, so I sold it at some point and never bought it again. For a long time afterward, I thought my decision was wise because they didn't rise much after 2005 for a long time. Holding them for another five or six years wouldn't have made much money either. The big rise in NVIDIA stocks happened later.

But I no longer hold NVIDIA stocks. I think as a journalist, I must maintain sufficient objectivity. If I heavily invested in NVIDIA and tied my economic fate to NVIDIA, I might sell more books, but now I really don't need to own stocks to benefit from them. I personally don't know much about stock knowledge, but I understood some basics earlier.

Mind Observer: But if you bought some stocks a few years ago, you would have made a fortune now. Many NVIDIA employees have become millionaires. Does Huang Renxun need extra incentive measures to motivate these wealthy employees to keep striving and maintain their hunger?

Stephen Witt: I don't think he has to do that. I talked to Jay Puri, NVIDIA's Executive Vice President of Global Business Operations, who holds $700 million worth of NVIDIA stocks and is very wealthy. He looks like he doesn't need to work, but he told me, "I can't quit, I'm working on something that is the most important technology in history." He sees his career as a matter of life value, believing that what he is doing is changing the whole human society, the most important thing in the world. If he stops working, he will miss many opportunities.

So even though they have so much money, their purpose is still to push forward, continue building such systems, and stay at the forefront of technology. I don't think some people become financially independent and prepare to retire — in most cases people do that. NVIDIA offers a once-in-a-lifetime opportunity to work: producing robots, automotive software, and various cool AI products.

Mind Observer: NVIDIA is no longer just a gaming GPU company. It has begun to enter some biomedical fields, such as developing simulation software for pharmaceuticals. What is the upper limit for NVIDIA? How much potential does it still have?

Stephen Witt: This is also a good question. NVIDIA has developed approximately 300 of the most cutting-edge software packages, which Huang Renxun calls his "treasures." These cover drug discovery, medical diagnosis, medical imaging, animation, gaming, physics, quantum computing, and more.

Their strategy seems to be not to miss any frontier technology, just like they didn't miss the AI wave. Therefore, they must always stay at the forefront of the most demanding technical and scientific computing systems.

If Huang Renxun doesn't build the most advanced computer for the most advanced research groups, he will feel uneasy and anxious. He worries that others will build it and surpass him, so he must stay at the forefront, such as drug discovery and quantum physics. He also seriously believes that the next trillion-dollar market may come from robotics.

Huang Renxun is very talented in robotics, investing heavily, and drug discovery and medical science may be another major event for him. He basically thinks of some fields.

NVIDIA and Huang Renxun's "secret technique" is to be able to talk to scientists and world-renowned researchers in their language.

Mind Observer: Talking about quantum computing and robotics, did you attend NVIDIA's GTC conference this year? Or did you discover anything interesting?

Stephen Witt: I heard he was a bit unhappy at the GTC conference. Last year, I saw his GTC live show, which was quite exciting, one of the best live shows I've ever seen.

But let's talk about the two things you mentioned, robotics and quantum computing. NVIDIA is working with Tesla to create an Omniverse platform for robotics, and the company has invested a lot of money in it. Essentially, this is a digital simulation system for robotic simulations. Huang Renxun collaborated with several scholars to compile a list of questions, including what you think robots will do in the future and what benefits we hope to gain, etc.

NVIDIA's Omniverse platform

The primary task people hope robots will perform is washing dishes, cleaning bathrooms, and cleaning toilets, among other things. We can let robots wash a million bowls in the real world, but we need to train them how to operate dishwashing in virtual environments.

What Huang Renxun wants to do is create a very realistic digital simulation that can mimic reality. In the digital world, if robots break a million or even a billion plates, who cares? They are all virtual.

Then slowly, if the simulation is realistic enough, robots will learn how to wash dishes.

When you're done, you can download it from the computer and deploy it to real-world robots, through which Huang Renxun believes he can position himself at the center of the robot revolution, a $1 trillion market, and this is NVIDIA's major initiative: building a robot training platform.

Of course, scientists have encountered many problems in the large-scale production of robots, which is a relatively disappointing field.

Earlier this year, Huang Renxun actually made some negative comments about quantum computing. He said quantum computing is a technology that will only become "very useful" in 15 to 30 years. Then came strong protests from customers. So I think he was trying to apologize at the GTC conference. He said, "Sorry, I didn't mean it. Maybe quantum computing is closer than I imagined. I'll set up scenarios for you on NVIDIA's quantum computing platform, please use it."

Huang Renxun doesn't want even one frontier patent in any scientific field to slip away from him. If there is a scientific field not connected to the NVIDIA platform, he becomes very anxious.

Mind Observer: NVIDIA has indeed made many revolutionary innovations in the data center sector. But as GPU clusters grow larger, the energy consumption of data centers is becoming an increasingly serious problem. People are particularly concerned about how NVIDIA plans to solve the energy dilemma.

Stephen Witt: NVIDIA seems not to care too much about this issue. I mean, in a sense, this is mainly a client issue, not theirs. NVIDIA actually wants chips to be more energy-efficient. And to some extent, chips are more focused on power consumption than ever before. More importantly, the demand for chips is higher than ever before.

People frequently use AI large language models, such as ChatGPT or DeepSeek, and they make many complex requests, requiring the generation of images, writing term papers, and planning. Then scientists and businesses also demand very complex things from artificial intelligence. Therefore, when computational power pipelines bring these demands to data centers for training and inference, energy consumption is significant.

The energy consumption is insane. When ChatGPT launched its new version, it added one million new users in a single day, so this creates an astonishing demand for electricity. Now NVIDIA is continuously improving energy efficiency, but it's not that the chips are inefficient; it's just that the demands placed on them are much higher than ever before.

I don't know what the answer to solving this problem will be in the future. Is it building more nuclear power plants? I think that might be the case. Burning coal or oil is not environmentally friendly. Nuclear energy is also expensive and has some drawbacks, but I think this is currently the best solution. We are building some nuclear power plants in the United States. I think the same is true in China. No one knows what the answer is yet regarding how to meet the growing demand for energy consumption in the future.

Mind Observer: NVIDIA's strong rise is closely related to the global AI craze. In your book, you mentioned that when discussing AI and human destiny, you had a fierce and unpleasant conversation with Huang Renxun, and he said, "You found the wrong person to interview, maybe you should ask Musk," what are the details behind this?

Stephen Witt: I really wanted to discuss this topic with him at that time. Birth rates are dropping significantly worldwide. Whether in China or the U.S., we are having fewer and fewer children. Meanwhile, we have increasingly powerful computer systems that seem capable of doing anything we can do. What will the future of humanity be like? If in the future we humans become the minority, what does that mean? If robots and computers can truly replace us and do anything we want, what is our role?

But Huang Renxun absolutely doesn't want to talk about this. He is a very pragmatic person and dislikes speculating about the future. He is also an ambitious person who likes to ask open-ended questions about what will happen in the future. But he really doesn't like these AI and human destiny issues, and I still think this is an important topic. Honestly, I don't know the answer either.

I am very interested in chess, which is an interesting example because artificial intelligence has long surpassed humans in playing chess, and all related research is now done by computers. But despite this, chess remains popular. In fact, since it became somewhat outdated, its popularity has increased. My phone plays chess better than almost any living chess player.

But people still enjoy watching humans play chess. Perhaps it's the same for writers. I think soon artificial intelligence will be able to deliver or write manuscripts better than I can, but perhaps future readers will be willing to read my works, even if my works may be worse than those generated by computers, just like I enjoy watching chess players play online, whose skills are not as good as those of computers. But in a way, I prefer watching humans play chess rather than watching two computers play each other.

In the AI era, the value of presence in the real world is actually increasing, or the value of having a distinctive, iconic friend is increasing. Computers can mimic becoming our friends, and the ability to maintain originality or present a human-like face is important.

Mind Observer: Do you think Musk is a suitable person to talk about this topic?

Stephen Witt: Musk has talked a lot about human landing on Mars. We need to remember that when Musk founded OpenAI, it was with a mindset of "red alert" — the original intention was to believe that artificial intelligence systems would take over humans and possibly harm them.

Times have changed, and I notice that now in Silicon Valley, this topic is not as fashionable as it used to be, perhaps because it involves too much money-making, but this remains a huge risk. I think no one has answered this question sufficiently, and this issue will not disappear. I don't know how Musk will respond to me, but he seems very interested in this topic.

Mind Observer: Musk is also very concerned about birth rates.

Stephen Witt: The global decline in birth rates is quite alarming, and it is happening in both the U.S. and Asia. We are currently on a trajectory where we may lose more than half of our population in the future. Historically speaking, the U.S. has been able to maintain a relatively high birth rate because we have accepted a large number of immigrants, but this situation is starting to change.

Mind Observer: This issue is related to export controls. Although NVIDIA has developed some compliant GPU types for the Chinese market, local GPU manufacturers in China are also making progress, especially in AI training and inference. How much impact will this have on NVIDIA's future performance?

Stephen Witt: I think this is Huang Renxun's biggest concern. In other parts of the world, most chip companies cannot compete with NVIDIA's latest products. But NVIDIA is prohibited from selling its latest products in China, creating an alternative competitive rule that allows Chinese companies to compete with NVIDIA and reach NVIDIA's level. For China, the difficulty lies in breaking through lithography technology, which may be the current biggest bottleneck.

NVIDIA does a lot of things by communicating with the world's greatest scientists. However, I think China can do this too. So I think Huang Renxun's biggest risk, and the threat he considers most dangerous to NVIDIA, comes from competitors in China. In my opinion, this is quite possible.

I have been to some relatively affluent areas in China and seen firsthand how advanced the technology is there, and how well the systems operate. Robots, subways, high-speed trains... all these incredible technologies, some of which we Americans haven't even researched yet. So the speed at which Chinese society is developing and advancing is simply unbelievable to me.

Mind Observer: NVIDIA has to deal with some geopolitical issues, and their public statements have always said they will try their best to comply with export control regulations. Did you feel during your conversation with Huang Renxun that he has concerns about deeper political issues?

Stephen Witt: He wants to sell chips to China not only because it makes money, but more importantly, it makes it harder for potential competitors to emerge in China. He worries that someone in China will develop faster and more cost-effective chip systems. Huang Renxun has always said that the rise of DeepSeek is good news for NVIDIA, and he also mentioned this at the recent GTC conference.

In fact, DeepSeek has created enormous demand for his products. If DeepSeek were a hardware competitor, it would not be good news for him. And if you can't sell the most advanced chips to China, it creates space for competitors to emerge.

Huang Renxun at NVIDIA's New Year party in Shanghai office in January 2024

You know, Huang Renxun is not a political figure. He has never made political donations, nor has he visited Trump's Mar-a-Lago estate (Note: Huang Renxun visited Mar-a-Lago after this dialogue to discuss H20 sales to China), and he did not attend his inauguration. To my knowledge, he has never expressed any political views.

But his company has grown so large and is so important in the industry that he can now force others to accept and collaborate on certain things. Huang Renxun says he will always comply with legal regulations. Of course, I know he thinks these export restrictions are foolish.

Mind Observer: DeepSeek also got some inspiration from NVIDIA's product architecture.

Stephen Witt: DeepSeek started with quantitative trading, and their inspiration may have come from American hedge fund tycoon James Simons, which I find very interesting. They have indeed achieved remarkable success.

Huang Renxun's view is that DeepSeek's popularity will only raise higher demands for AI. So when we lower the cost of AI training, demand will actually increase. This is NVIDIA's core philosophy, looking back, they have always been searching for those "limitless demand" applications. For example, 3D graphics rendering, no matter how good we render, customers always want better rendering effects. Demand is infinite, no matter how good we do, players always want better ones.

Huang Renxun believes that no matter how intelligent we make computers, people will always keep posing increasingly complex demands. This will create huge demand and pressure for the reasoning systems serving them. So they will input increasingly complex queries into ChatGPT or DeepSeek, creating ever-growing demand for computational power, and selling computational power is NVIDIA's job.

If Huang Renxun thinks DeepSeek poses a threat to NVIDIA, this is a complete misunderstanding of him. What is actually harmful to him are export controls and tariffs.

Mind Observer: Huang Renxun has always said that NVIDIA is essentially a software company, and CUDA's moat appears very secure.

Stephen Witt: They have world-class hardware designers. NVIDIA people told me, "AMD can also manufacture chips that are just as good as ours."

But beyond chip hardware, there are compilers and software development kits. NVIDIA has achieved a lot of fundamental computer science innovations in these areas, and Huang Renxun calls them "my treasures." Those things面向 scientists and developers, and the parts of the computing stack that interact with actual power engineers are what he considers the hardest things to replicate — this is where he believes he has a competitive advantage over others.

Mind Observer: To complete this book, you interviewed hundreds of people. What special methods did you use to verify facts, and what interesting feedback did you receive?

Stephen Witt: Why hire fact-checkers to verify the facts in the book? I have strong independence in this regard, I am bold, and I will call relevant people directly: "Hey, what do you want to talk about with me?" I don't have any other complex strategies.

I mean, usually I think of reasons to make them want to talk to me, or find ways to contact them early. This method is not something pioneering, and many people also refuse to talk to me. So, if I have talked to 200 people, it means I tried to contact 400 people, but half of them didn't reply, but I like casting a wide net. Through chatting with others, I have already mastered everything I need.

In a way, I wish I could spend more time on this book. I wrote this book quickly, and it was published quickly because of market and reader considerations. But honestly, I wish I could spend 10 years writing it. Now I feel like I have only scratched the surface of the issue, and I really want to go back and rewrite the book from scratch in a completely different way. The story is so important and captivating.

Mind Observer: From the small Danny's Restaurant to the world's highest-valued technology company, there are so many highlights along the way. If you encounter Huang Renxun again, what other questions would you ask him?

Stephen Witt: Ask him about Trump and everything else. For example, more about his family dynamics, about his time in Taiwan, and his experiences returning to Taiwan. For instance, I really want to know what it means for him to have "dual nationality" and what it feels like to have Chinese and American "bloodlines." I also want to interview his wife, daughter, and son. So if I have the chance, I will try my best to do so.

Mind Observer: You also mentioned in your book that, in terms of age, they got married quite young.

Stephen Witt: Yes, he loves his wife. Generally speaking, this is particularly uncommon for Silicon Valley business elite CEOs. For example, Jeff Bezos got divorced, and Musk is even more involved with multiple women. But this is not Huang Renxun's style; he is still with his wife.

Mind Observer: Final question, if you could give readers a very convincing reason to buy your book, what would that reason be?

Stephen Witt: This is a groundbreaking story about technology, with the paradigm of artificial intelligence running throughout. AI is a revolution, and Huang Renxun made it a reality. Without Huang Renxun, our AI technology would be 10 years behind. He is the one who brought us into the future.

This article is an exclusive piece from Observer Network, and the content purely represents the author's personal views, which do not necessarily reflect the platform's views. Unauthorized reproduction is prohibited, and legal action will be taken if violated. Follow Observer Network's WeChat account guanchacn for daily interesting articles.

Original link: https://www.toutiao.com/article/7493820137372434963/

Disclaimer: The article solely reflects the views of the author and welcomes your opinions by clicking the "Like/Dislike" button below.