Jensen Huang on China, GPUs, and export controls | Lex Fridman in Taiwan

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: Jensen Huang on China, GPUs, and export controls | Lex Fridman in Taiwan · Channel: Lex Clips · 11:57 · Watch on YouTube

Topics: AI governance and regulation, Markets and economy, Open source and software ecosystems

Summary

Jensen Huang argues that export controls can create a short-term competitive edge by restricting access to valuable technologies, but that long-term security and leadership depend on a prosperous technology-and-economic flywheel. He favors balancing selective restrictions with broad platform reach, which supports customer adoption, wealth, and continued research and development. Huang and Lex Fridman also discuss how interdependence and collaboration can reduce the risk of conflict between competing countries. They conclude by emphasizing shared human interests and the value of competing as friends rather than making enemies.

Key points

Huang says goodwill is not zero-sum and describes wanting colleagues and people at other companies to succeed.
Restricting access to valuable resources can improve one country’s position, but other parties may create their own alternatives.
Huang argues that prosperity is important for long-term national security and domestic harmony.
He describes a flywheel in which technology leadership and economic leadership reinforce one another.
His proposed export-control approach is to be highly selective while preserving the technology and economic flywheel.
Huang and Fridman say competition should continue, while countries should do what they can to avoid conflict.
Interdependence across AI technologies, universities, and scientific collaboration can make conflict harder.
Huang says the United States benefits from a global platform reach, including because much of the world uses English.
He calls the flywheel delicate: platforms can reach a tipping point, but the flywheel can also slow down.
They close by advocating understanding, collaboration, friendship, and competition without turning one another into enemies.

Chapters

Start	Chapter	Summary
00:07	Humility, goodwill, and helping others succeed	Huang reflects on treating people well and wanting others, including colleagues at other companies, to succeed.
01:07	The case for and against export controls	Huang explains how restricting valuable technologies can create an advantage, then contrasts that with the benefits of long-term prosperity.
03:12	The technology and economic flywheel	Huang describes how wealth, customers, and technology leadership can reinforce one another and shape a country’s long-term position.
05:16	Competition, interdependence, and conflict	Huang and Fridman discuss avoiding conflict and how shared technologies, education, and collaboration can connect countries.
08:22	Global reach and balancing export controls	They discuss America’s innovation and platform advantages, the value of global reach, and the balance between national rights and sustaining the flywheel.

Quotes

The case against it is simple. — 01:39

a more sophisticated long-term strategy is to optimize for a flywheel. — 03:43

The harder it is to decouple the harder it is to have conflict — 06:49

Technology without customers is useless. — 08:52

compete as friends versus dividing each other and making enemies of each other. — 11:27

Concepts and entities

Jensen Huang, Lex Fridman, NVIDIA, controles de exportação, ciclo de inovação e prosperidade, AI stack, interdependência

Full transcript (original language)

00:07 [cheering] >> Beatles in the 60s. Through all of that, maintaining humility. >> What a life, Jensen. >> Does that weight ever get heavy, that you have this responsibility, millions of people looking up to you? >> I It doesn’t, you know, because I I don’t take it I don’t take it like that. You know, people have always been nice to me. >> Even when you had nothing? >> Yeah. And I was always nice to other people. >> Yeah. >> You know? And I was always polite. And I always wanted people to succeed.

00:37 I wanted them to be happy. When I was going through school, I was always hoping that all the other kids would do well in school. And I was I was happy to share my homework with everybody. I like seeing other people succeed, even my colleagues in other companies. >> It’s a funny thing how the world works. If you just celebrate others and you want them to succeed, somehow you succeed. >> Well, for example, 43,000 employees at Nvidia want me to succeed. Literally.

01:07 You know what I’m saying? >> And you want you want each of them to succeed. >> Exactly. >> And you want I mean, I’ve seen you. You want people at other [snorts] companies to succeed. >> And they want me to succeed. >> And that There’s a ripple effect to the whole thing. >> Yeah. Exactly. There’s that You know, goodwill is not a zero-sum game. It’s not limited by physics. >> [clears throat] >> Can you think through talk through your export controls view? You’ve been had strong opinions that um export controls

01:39 is not the way America should win. >> Mhm. >> Can you still me on the case for that and against it? >> The The case against it is simple. If there’s something that that is valuable, something that’s useful, something that is a force multiplier. So, for example, uh rare earth is a force multiplier, computer technology is a force multiplier. Um you know, technology in general is a force multiplier. You could argue that knowledge is a force multiplier. >> Mhm. >> And intellectual property is a force

02:09 multiplier. And so so if um we have assets like that then depriving other people of it um enhances our position. Deprives them of theirs. And >> It’s a way to gain a competitive edge. >> Yeah. >> It’s by hoarding that particular resource. >> Right. Keep it out of the hands of others. And that’s that’s not completely untrue uh with respect to uh the physical things. Like for example

02:40 uh it is not untrue with respect to to uh uh refined rare earth. It’s not untrue uh when it’s relates to uh bullets, bombs. Um I think that’s generally not untrue. To the extent that the the opposition can and other parties can therefore create their own um it is generally not untrue. Uh the other the other side of the argument is that if our goal is to optimize for long-term

03:12 national security, long-term um long-term domestic tranquility and harmony then prosperity is the best way to do that. The the safest countries in the world are the richest countries. The reason why the United States is so powerful militarily is because we have extraordinary wealth. And that extraordinary wealth came along with extraordinary technology, and that technology and wealth becomes a flywheel. And um that flywheel is ultimately what

03:43 you’re optimizing for. So the question is are you optimizing for someone else not having access to this technology, or you optimizing for a flywheel. And I I suggest that a more sophisticated long-term strategy is to optimize for a flywheel. And in order to optimize for that flywheel you have to have both simultaneously technology leadership and wealth. That flywheel is really important. This is no different than, you know, a lot of

04:14 computing technologies and platform technologies. The the concept of wealth is basically customer engagement and installed base and that installed base and customer use usage drives more consumption of your technology, which then drives more wealth, which allows you to scale your technology faster than anyone. And so so this concept of flywheel which comes along with uh technology leadership economic leadership, which fuels technology

04:45 leadership, which drives more economic leadership. And that flywheel’s caught is what what causes companies and countries to stay ahead. That’s what I would I would I would suggest that instead of finite things that are concrete and has no no um positive feedback system that that creates I would advocate that that the better strategy is to be quite extremely selective,

05:16 extremely selective about uh access to uh certain technologies. But aside from that, balance that with optimizing for that technology and economic flywheel. So there’s a balance. >> I think the question is does America flourish if that flywheel includes China or doesn’t include China? That’s the open question for America and the open question for humanity. There’s a line that’s crossed when two nations are competing as

05:47 you can even say geopolitically friends competing versus becoming enemies competing. And there’s a line and once you cross that line you I think have the the potential to destroy everybody’s flywheel. >> That’s right. I would say that that um we should do everything we can to avoid conflict. We should compete. That we can’t do without. And but we should do everything we can to

06:18 avoid conflict. And and so the question is is therefore if you have a relation if you have relationships that are competitive um there are strategies that cause that relationship to leads towards conflict and there are strategies to avoid conflict. >> I think that to the degree there’s inter-dependence is something you and I talked about amongst let’s say AI AI stacks. Everybody’s using the same hardware, everyone’s using the same software.

06:49 Um everybody’s going to each other’s universities, there’s friendships, there’s scientific collaborations. Uh there’s friendly competition amongst different companies but nevertheless it’s it really fundamentally is collaboration. The more of that we have the harder it is to actually divide the two different peoples in order to then have a something like a global conflict. >> The harder it is to decouple the harder it is to have conflict and I agree with that. Now, one of the things

07:20 that that um is helpful and just now the way you described it is absolutely right. In fact, AI is not a model. AI is a whole stack of technologies and they’re all contributing to AI. You can’t have AI without energy. And it and we know that that um certain countries and certain regions have extraordinary leadership and capabilities and energy while others don’t. And so if you keep going up and down

07:51 that stack, you’ll notice you you observe that depending on where you’re talking about um leadership is happening in different places. >> Jensen and I as as uh proud Americans, I I personally believe that uh America has the most innovative people. Like the craziest, wildest, out there, big, big, moonshot ideas. I think America’s on top of the world in doing that. So uh as you said

08:22 >> Well, Nvidia is an American company. >> Yeah. >> Uh Anthropic, OpenAI, um you know, Google. Uh there there’s just so many so much overwhelming evidence that what you’re saying is absolutely right. The The compounded chemistry that leads to this innovation capability over time >> Mhm. >> um is overwhelming. One of the one of the overwhelming advantages that we have is that our flywheel is also the best.

08:52 And the reason for that is because the vast majority of the world um responds to English. And so if you’re an American company and you build on an American tech stack using American standards you’re fairly deployable throughout the world. That’s not true with China. And so we have we have a global flywheel advantage. Technology without customers is useless. >> Yeah, you have to use that advantage. Don’t take for granted your advantage. There is a global power to English as a

09:23 language. It’s just the way that America’s able to create an ecosystem that’s used by everybody. That’s a real real advantage. So closing yourself off from the rest of the world for some sort of uh the benefits you do get from isolationism, there’s a real cost to that. You lose that advantage. >> Exactly. Which is exactly the reason why I believe that on the one hand, we want to keep great technology to ourselves. On the other hand, we need that flywheel. We need exactly the

09:54 sentence that you are applying just now. Every single word of it. And so, we want to have as much platform reach as possible because it generates that flywheel, which generates the wealth, which generates our R&D capacity to continue to innovate. Now, that flywheel is also really delicate. You could see flywheels once it gets going, it becomes, you know, a bit of an exponential. And platform quickly seems like overnight after being

10:26 built for a decade, two decades, in our case, three decades, all of a sudden, it reaches a tipping point and it’s everywhere. On the other hand, that flywheel can stop. And it could, you know, it could slow down very quickly. And so, so I the the question about export controls is just this balance. And and every country deserves Every country has the right for the technologies that the capabilities or the assets, the resources that they’ve created for, you know, the support of their national advantage, they have the right to decide

10:57 how they want to use it. And I And And that’s a That’s a sovereign right. That’s a national right. On the other hand, we also want to make sure that the national treasure of the United States, which the technology industry is part of, that national treasure has the largest possible flywheel, the largest possible platform reach, so that we can sustain that, you know, that economic and technology leadership flywheel. So, that’s the balance. >> And at the end of the day, as we’ve been sharing a lot of food, I think we have to remember

11:27 that uh all the humans all across the world who have a lot more in common that uh we have this different and it’s it’s good to understand each other, to collaborate, to be friends with each other, to work together and compete as friends versus uh dividing each other and making enemies of each other. >> And when you travel, you really it really get hit with that. >> Especially when you hitchhike. >> [laughter] >> Especially when you hitchhike, you realize that so much. >> Jason, thank you so much for the delicious food. This is amazing.

11:57 >> was that? >> That was fun. >> How fun was that? >> And maybe some night market tonight as well. >> Okay. Okay, I’m going to take you to a different one. >> Yeah. Even crazier. Even crazier. >> it’s going to be incredible.