英伟达起步：从教科书学习

欢迎来到 Startup School 2026。现在，让我们开始吧。请和我一起欢迎英伟达的创始人兼 CEO 黄仁勋上台。嘿，Jensen。请大家欢迎。

Welcome to Startup School 2026. Now, let's get started. Please join me in welcoming to the stage the founder and CEO of Nvidia, Jensen Huang. Hey Jensen. Please, everybody.

天哪，这对我来说真是个超现实的时刻。谢谢。

Oh my god. This is a surreal moment for me. Thank you.

感谢你来到这里，Jensen。

Thank you for being here, Jensen.

我很高兴来。能在这里真是太好了。显然，如果你在这里，你就能成功。所以，我很高兴我来了。

I'm delighted to do it. It's great to be here. Apparently, if you're here, you are going to make it. So, I'm happy I'm here.

哦，Jensen。对于只知道英伟达是 AI 核心公司的学生来说，他们最需要了解英伟达早期故事的哪一部分？

Oh, Jensen. Well, for the students who only know Nvidia as a company at the center of AI, what part of the early Nvidia story do they most need to understand?

大多数人不相信的是，我们创办公司时选择的技术绝对是错误的。我们最初的想法是重新发明 3D 图形。公司的理念是，通用计算机（CPU）确实很有用，但如果我们能用加速器增强它们，就能解决原本太难解决的问题。我们选择的第一个问题就是 3D 图形。当时是 1993 年，PC 刚传闻要出现。我们的主要想法是把每一台个人电脑都变成游戏机，因为我们在游戏机时代长大。我们想，如果我们能设计一个系统，装进个人电脑，把它变成游戏机呢？所以我们想重新发明算法，那些需要大型超级计算机的算法，并把它装进 PC。我们想出了一些新算法。我们很兴奋，我们相信它。我们进行了深思熟虑的推理。然后我们去创办公司来构建它。结果，算法完全错了。创办公司的技术也完全错了。1995 年，我们意识到了这一点，但几乎为时已晚，因为那时已经有大约 35 到 40 家公司为 PC 构建 3D 图形。我们意识到它行不通。我回到公司，说：‘我们该怎么办？它行不通。’我们都在讨论。我说：‘看，如果我们不面对这行不通的事实，并开始朝着正确的算法努力，我们就不会有公司了。’然后有人告诉我，我们没有人知道正确的方法。我们不仅选错了技术，还不知道正确的方法。那对我来说是重要的一天。我口袋里只有几百美元，所以我去了 Fry's，买了三本教科书。这些教科书是关于 OpenGL 和如何设计流水线的。我把它们带回公司，交给了工程师们，然后我们就走到了今天。我们重新发明了计算机图形学。我们成为现代计算机图形学的世界领导者。我们在过去 25 年里发明了大多数重大突破。大家都会以为英伟达一开始就是 3D 图形领域的领导者，但我们是从教科书里学来的。我们实际上创办了公司，筹集了资金，然后买了教科书。想想看，对我来说，最大的教训是技术总是在变化。只要你能面对现实并学习，技术本身其实并不重要。从那以后，英伟达一直在发明我们从未做过的新技术。我们以同样的态度对待一切：如果这件事很重要，我们就去学，它能有多难？它总是比我们预想的要难得多，但你要带着‘它能有多难？’的态度去开始。

The thing that most people don't believe is that the choice of technology we started the company with was absolutely wrong. We started with the idea that we would reinvent 3D graphics. The company's philosophy was that general-purpose computers, the CPUs, were really useful, but if we could augment them with accelerators, we could solve problems that were otherwise too hard to solve. One of the first problems we chose was 3D graphics. During that time, 1993, the PC was just rumored to be coming. Our big idea was that we would turn every personal computer into a game console because we grew up in the era of game consoles. We thought: what if we could design a system that would fit into the personal computer and turn it into a game console? So we thought we would reinvent the algorithm that required these large supercomputers and fit it into the PC. We came up with some new algorithms. We were excited about it. We believed in it. We reasoned about it in a thoughtful way. We went to start the company to build it. Well, it turns out the algorithm was exactly wrong. The technology that founded the company turned out to be exactly wrong. In 1995, we realized that, and it was almost too late because by then there were some 35 to 40 other companies building 3D graphics for PCs. We realized it didn't work. I went back to the company and said, 'What are we going to do? It doesn't work.' We were all talking about it. I said, 'Look, we won't have a company if we don't confront the fact that this doesn't work and start working towards the right algorithm.' Then somebody told me that none of us knew how to do it the right way. Not only did we choose the wrong technology, we didn't know how to do it the right way. That was a big day for me. I had a couple of hundred dollars in my pocket, so I went down to Fry's and bought three textbooks. The textbooks were about OpenGL and how to design pipelines. I brought them back to the company and gave them to the engineers, and here we are. We reinvented computer graphics. We're the world leader in modern computer graphics. We invented most of the major breakthroughs in the last 25 years. Everybody would have thought that Nvidia started out as world leaders in 3D graphics, but we learned it from a textbook. We actually started the company, raised money, and bought textbooks. When you think about it, the big lesson for me is that technology is changing all the time. So long as you are able to confront reality and learn, the technology itself actually doesn't matter. Since then, Nvidia has been inventing all kinds of technology we had never done before. We approach everything with the same attitude: if it's important to do, we're going to go learn it, and how hard can it be? It always turns out to be much harder than we expect, but you go into it with the attitude, how hard can it be?

我的意思是，在后台我们和一些顶尖的 YC 公司聊过，你说每一家公司都在某个领域有专长，而你和英伟达也在那个领域有专长，它们都只是——我忘了你怎么说的——某种算法领域。所以听起来 3D 图形只是第一个算法领域。

I mean, backstage we were talking with some of the top YC companies, and you were saying that each one has an expertise in a domain that you and Nvidia have an expertise in, and they're all just—I forget what you said—it was like an algorithmic domain of a sort. So it sounds like 3D graphics was merely the first of an algorithmic domain.

没错。粒子物理、流体动力学，是的。

That's right. Particle physics, fluid dynamics, yeah.

但是你创造了人们想要的东西，人们愿意花大价钱购买的最终产品。

But you created the thing that people want, the end product that people want to pay a lot of money for.

公司完全正确的核心理念是，可以增强 CPU 来解决那些原本太难解决的问题。分子动力学是其中之一，图像处理是其中之一，逆物理是另一个。各种不同的算法，当然深度学习是主要之一。为了打造今天的公司，我们很早就意识到，关键不在于制造出色的芯片，而在于加速一个算法领域。我一直坚信，伟大公司的特质是对世界有一个你深信不疑的独特视角。这与其说是技术，甚至不如说是市场。这些都很重要，如果你在正确的时间为正确的市场提供了正确的技术，你的生活会容易得多。但是，一个关于未来某个重要事物的高层次愿景，一个独特的、你深信不疑的视角，并且追求这个愿景是困难的——这些都是不错的组合。就我们而言，我们意识到加速计算将变得重要。而加速计算确实变得非常重要，我们的认识是，一切都与算法有关，而不是芯片。事实证明这是完全正确的。

The big idea of the company that was spot-on is that it is possible to augment the CPU to solve problems that are otherwise too difficult to solve. Molecular dynamics is one of them, image processing is one of them, inverse physics is another one. All kinds of different algorithms, of course deep learning is one of the major ones. To create the company we have today, we realized early on that it's not about building a great chip, it's about accelerating an algorithm domain. One of the things I've always believed is that what makes great companies is a unique perspective about the world that you deeply believe in. It's not so much the technology, it's not so much the market even. Those things all matter, and if you have the right technology for the right market at the right time, your life is going to be a lot easier. But a high-level vision about the future of some important thing, a perspective about it that is somehow unique, that you deeply believe in, and ideally pursuing that vision is hard to do—those are kind of good combinations. In our case, we realized that accelerated computing was going to be important. And accelerated computing turns out to be very important, and our realization is everything to do with algorithm, not the chip. That turns out to be exactly right.

你谈了很多关于创始人的艰辛。有没有一些特别让你难忘的故事？这个房间里的人都想创办一家公司，但他们真的准备好“吃玻璃”、可能不得不关闭公司、事情出错了吗？有哪些关键时刻让你印象深刻？我记得你刚去过日本，对吧？你是在纪念世嘉，是吗？我觉得那是一个很有力量的故事。

So you've said a lot about the hardships of a founder. Are there a few stories that really jump out at you? The people in this room would love to start a company, but are they really prepared for eating glass and possibly having to shut down the company? Things going wrong? What are some of the pivotal moments that really jump out at you? I think you were just in Japan, right? And you were sort of honoring Sega, was it? So I feel like that was a really powerful story.

让我们意识到我们选择的算法是错误的项目是与世嘉的合作。世嘉曾与我们签约，为他们建造在 Saturn 之后的那款游戏机，后来变成了 Dreamcast。我不知道有没有人知道 Dreamcast。好吧。所以我们没有建造 Dreamcast。

The project that led us to realize the algorithm we chose was wrong was a partnership with Sega. Sega had contracted us to build the game console after Saturn that turned out to have been Dreamcast. I don't know if anybody knows what Dreamcast is. Okay. So, we did not build Dreamcast.

我们原本应该造 Dreamcast。但因为我们的算法和技术有根本缺陷，我去日本告诉当时的 CEO Irimajiri 先生，我们无法完成那 1200 万美元的合同，因为技术行不通。我解释了原因。然后我建议他们另找别人。但我也问能不能还是拿到那笔钱。他说：“你告诉我你做不到我签约让你做的事，但你想要全部的钱。”我说：“完全正确。”显然，我很礼貌也很谦逊。他意识到我是诚实的，一切都有道理。如果他不给我们钱，我们就倒闭了。我认为你投资的是人，不是公司。Irimajiri 意识到他信任我们，想让我们活下去。那 500 万美元让我们活了下来，给了我时间去发现下一步该做什么。

We were originally supposed to build Dreamcast. But because our algorithm and technology were fundamentally flawed, I went to Japan and told Irimajiri-san, the CEO at the time, that we couldn't fulfill the $12 million contract because the technology didn't work. I explained why. Then I advised they choose someone else. But I also asked if I could still have the money. He said, 'You're telling me you can't do what I contracted you to do, but you want all the money.' I said, 'That's exactly right.' Obviously, I was polite and humble. He realized I was honest and it all made sense. If he didn't give us the money, we'd be out of business. I think you invest in people, not companies. Irimajiri recognized he trusted us and wanted to see us make it. That $5 million kept us alive and gave me time to discover what to do.

然后我猜他们以 1500 万美元卖掉了，我听说的。

And then I guess they sold it for 15 million, I heard.

对，我们一上市他们就卖了。Nvidia 上市时，估值是 3 亿美元，那是 1999 年。那是笔真金白银。

Yeah, they sold it the moment we went public. When Nvidia went public, our valuation was $300 million in 1999. That was real money.

我觉得现在超过一万亿了吧。

I think it's north of a trillion now or so.

不止一万亿，是的。

It's more than a trillion, yeah.

太疯狂了。所以你是核心——我们喜欢说你是控制香料的人。在此之前，我不认为有人能预测 GPU 和你构建的技术对这场 AI 革命有多重要。你看到了什么？是加速器加上天时地利，还是很多因素让你抓住了这个位置？

That's wild. So, you're sort of the core – we like to say you're the man who controls the spice. Before that, I don't think anyone could have predicted how important GPUs and the technology you built would be for this AI revolution. What did you see? Was it the accelerator and being in the right place at the right time? Or were there many things that led to that position?

是的。我和其他人一样看到了 AlexNet。但记住，我看世界的角度总是在寻找算法。算法可以是 NAMD、VASP、OpenGL、SQL——某些领域特定语言。所以当 AlexNet 出现时，算法就是深度学习。问题是：这是什么算法，为什么这么有效，还能做什么？如果把它扩展到更大，它能解决哪些现在解决不了的问题？我们的突破在于意识到 AlexNet 不仅仅是 AlexNet。它是一种深度学习的方法，能让你学习任何函数。所以 15 年前，我告诉所有人：“我们刚刚发现了通用函数近似器。”我们可以给它几乎任何函数，它就能学会该函数是什么。对于很多函数，不需要精确；事实上，精确是不可能的。大多数有趣的问题都是不精确的。所以当我们意识到我们有了一个通用函数近似器那天，我们问：这对计算栈意味着什么？软件会发生什么？哪些行业会受影响？几乎立刻，我们开始研究计算机视觉、机器人、自动驾驶。那种基础能力能解决那些领域的重要问题。大的突破是：这比 AlexNet 更基础。这是一种新的做软件的方式。对处理器、中间件、算法、应用——五层蛋糕——我在 15 年前就想象要重塑整个工业栈。这不过就是问问题，从第一性原理推理：“如果这样，那么怎样？如果它能变得更好，那又怎样？”

Yeah. I saw AlexNet just like everybody else. But remember, my view of the world was always looking for algorithms. The algorithm could be NAMD, VASP, OpenGL, SQL—some domain-specific language. So when AlexNet came along, the algorithm was deep learning. The question was: what is this algorithm, why is it so effective, and what else can it do? If you scale it beyond that, what could it solve that we can't solve today? The breakthrough for us was realizing AlexNet was not just AlexNet. It was an approach with deep learning that allows you to learn any function. So 15 years ago, I told everyone, 'We just discovered the universal function approximator.' We can give it almost any function and it can learn what the function is. For many functions, you don't have to be precise; in fact, it's impossible to be precise. Most interesting problems are imprecise. So the day we realized we have a universal function approximator, we asked: what does that do to the computing stack? What happens to software? What industries could it impact? Almost right away, we started working on computer vision, robotics, self-driving cars. That fundamental capability could solve important problems in those areas. The big breakthrough was that this is much more foundational than AlexNet. This is a new way of doing software. The implications for the processor, middleware, algorithms, applications—the five-layer cake—I imagined reinventing that entire industrial stack about 15 years ago. It's about asking questions, reasoning from first principles: 'If this, then what? If this can get better, so what?'

有一件事非常突出：你深入到细节里。你读论文，直接与首席科学家交谈。那是真正的创始人模式。与此同时，你有一个组织，有高管们想要保持在图表上。有时候这会惹恼一些人。你对人们关于如何建立能按第一性原理思考的组织有什么建议吗？因为如果财富 500 强也这样做，他们会更像 Nvidia。你建立了一家非常独特的公司。

One thing that really jumps out is how deep you go into the weeds. You read papers, talk directly to principal scientists. That's true founder mode. At the same time, you have an organization with executives who want to stay on the graph. Sometimes that ruffles feathers. Do you have advice for people about building an organization that thinks in first principles? Because if the Fortune 500 did that, they'd look more like Nvidia. You've built a very unique company.

我的心态总是从好奇心开始。我自己有很多问题。和其他人一样，我会寻找最短路径得到答案。但很多时候，我身边的人给出的答案并不令人满意，我会有更多问题。也许他们很忙。所以我的第一倾向是满足自己的好奇心，亲自寻找答案。第二，如果我发现某个领域的信息可能对某人或我们公司很重要，那么我会希望尽可能多学习，以便能为公司服务并与大家分享。这和你分享知识没有什么不同。我看你的播客和视频，非常喜欢。你在与大家分享想法。

My state of mind always starts with curiosity. I have a bunch of questions myself. Like anyone else, I seek the shortest path to an answer. But often the answers from people near me aren't satisfying, and I have more questions. Maybe they're busy. So my first inclination is to discover the answers to my own curiosity. Second, if I find that the information in a domain could be important to someone and to our company, then I want to learn as much as possible so I can be of service and share it. This is no different from you sharing knowledge. I watch your podcasts and videos and really enjoy them. You're sharing ideas with everyone.

从很多方面来说，我认为 CEO 是为公司服务的，是为所有在那里工作的人服务的。你想赋予他们一些洞察力。所以这就是它的来源。与其说是一种管理技巧，不如说是一种人格技巧。我想赋予你力量，这是我刚刚观察到的一件非常重要的事情。我来告诉你为什么它如此重要。之所以要贴近地面、深入一线，是因为技术往往很复杂，或者变化非常快。尤其是在像我们这个世界这样快速变化的情况下，除非你对实际发生的事情有切身感受，否则你可能会觉得它变化太快而无法理解。但如果你随着时间的推移理解了它的基本原理，那么一切就都说得通了。这有点像我设想的冲浪。我不会冲浪，但我能想象它有点像冲浪。你踏上浪尖。对我来说，这看起来很混乱，但对冲浪者来说，他们能读懂波浪，知道如何保持在浪尖上。所以我认为当 CEO 和冲浪非常相似。你必须学会如何冲浪，而为了学会冲浪，你必须理解波浪。你必须能够读懂风向，还要有好的时机，除非你尝试，除非你真正去做，否则你不可能拥有这些。一部分是为了让自己了解情况，一部分是为了尝试分解问题，以便公司能够以他们能够采取行动的方式学习。一部分是为了激励他人。这都是在座所有人具备的基本特质。当你成为 CEO 时，你不必改变你的个性或行为。你完全可以继续做你自己。很久以前我学到的一件事，我不知道在哪里看到的，CEO 或创始人正在建造一辆你将要驾驶的赛车。你要制造一辆 F1 赛车，但你要以你能驾驶的方式去制造它。你应该让车适应你。有人曾经问我：‘黄仁勋，如果你不使用传统的管理技巧和组织技巧，你离开公司后会发生什么？’我说，当我有一天在岗位上倒下时，他们只需要为公司重塑一个适合下一任 CEO 的架构。这是明智的，因为我们是 F1 车手。我们是赛车手。世界竞争非常激烈，我们必须赢。我们必须实现我们的使命。所以，无论需要做什么来让车适应你，无论需要做什么来让组织适应你，那就是你必须做的。下一任 CEO，无论个性如何，他们自己会想办法。

In a lot of ways, I think a CEO is in service of the company, in service of all the people that are working there. And you want to empower them with some insight. So that's really where it's coming from. It's not so much a management technique, but a personality technique. I want to empower you and this is something really important that I just observed. Let me tell you why it's so important. Part of having to be near the ground and be in the weeds is because often the technology is complicated or it's changing really fast. And especially when it's changing fast like our world, unless you have a tactile sensation of what is actually happening, it could either feel like it's moving way too fast to understand. But if you understand the first principles of it over time, then everything kind of makes sense. It's kind of like surfing I would imagine. I don't know how to surf, but I can imagine it's kind of like surfing. You get out on the wave. To me it looks like chaos, but to a surfer, they can read the waves and they know how to stay on top of it. So I think being CEO is very similar to that. You have to learn how to surf, and in order to learn how to surf you have to understand the waves. You have to be able to read the wind and you have to have good timing, and you can't have any of that unless you try, unless you actually do it. Partly it is to inform myself, partly it is to try to break down the problem so that the company can learn it in a way that they can do something about. Part of it is about inspiring other people. It's all those basic traits of all the people in this room. You don't have to change your personality or your behavior when you become CEO. It is possible for you to continue to be yourself. One of the things that I learned a long time ago, and I have no idea where I saw this, but the CEO or the founders are building a car that you are going to race. You're going to build an F1 racer, but you're going to build it in a way that you can drive. You should adapt the car to you. Somebody had asked me, 'Jensen, if you don't use conventional management techniques and organizational techniques, what's going to happen when you leave the company?' Well, when I die on the job someday, I told them they will just have to reshape the company for the next CEO. The reason that's wisdom is because we are the F1 drivers. We are the racers. The world is really competitive and we have to win. We have to achieve our mission. So whatever it takes to fit the car to you, whatever it takes to fit the organization to you, that's what you got to do. And the next CEO, whatever the personality is, they can figure it out.

太棒了。我的意思是，你对车所做的任何改变都只会让你慢下来，让你输掉那些不适合你的比赛。

Amazing. I mean, that does seem like any change you make to the car will just slow you down and lose you races that isn't fit to you.

是的，我们一直在根据我们的需求调整赛车。这就是我一直在做的事情。我不断调整公司，不断重塑业务流程和工作方式，以便为公司更有效地工作。

Yeah, we are constantly tweaking the car to our needs. That's really what I'm doing all the time. I'm constantly tweaking the company, constantly reshaping business processes and the way things work so that I can be more effective for the company.

真正的创始人模式。

True founder mode.

是的，创始人模式。创始人模式可以持续 34 年。

Yeah, founder mode. Founder mode could scale for 34 years.

没错。

That's right.

从零到 5 万亿美元。没有证据。不

From zero to 5 trillion. No evidence. No

我想换个话题，谈谈你现在最感兴趣的前沿算法是什么？我很喜欢你从材料科学深入到应用层面。你是第一个在舞台上谈论 Open Claw 和 Hermes 智能体的人。我想知道你是否可以带我们了解你一天中如何思考不同阶段。从材料到芯片到数据中心，甚至到应用层面，比如人们将如何工作。这有点像全栈 AI 工厂的概念。

I'd love to switch gears to talk about what are the frontier algorithms that you're most interested in now? I love that you're all the way down into the material science, all the way up into the app level. You know, you're the first to speak on stage about Open Claw and now Hermes agent. I wonder if you can sort of walk us through a day in the life of how you think about the different stages. Going from materials to chips to data centers to even the app level, like how people are going to work. There's this idea of a full stack AI factory.

这可能是未来最有用的技能之一。事实上，听你谈论技术以及你如何使用它，最重要的事情之一就是系统理解。系统意识、系统设计、系统组织。系统思考。原因在于大多数需要完成的底层任务都将通过智能体方式完成。它们无论如何都会被自动化。所以，无论是在我这一代，我们编译芯片、合成晶体管、门电路和功能模块，所有这些都是合成出来的。所以我们的设计师大多是系统设计师。在软件方面，大多数软件也将通过智能体方式完成。因此，你必须更能抽象地思考系统。你要解决的问题是什么？约束条件是什么？输入在哪里？输出在哪里？信息从哪里来？信息流入和流出系统的速率是多少？约束条件是什么？是处理器吗？是内存吗？是网络吗？在足够技术层面上的这些系统问题将对在座的所有人非常有帮助。我不认为这种基础知识会变得无用。我认为它会越来越有用。我尽力去理解系统。说到智能体，事实上我们已经有了粗略级别的递归自我改进。每次你使用它，它都会改进 markdown 文件。每次你使用它，它都会更新它的长期记忆。长期记忆被处理，要么压缩，要么变成知识图谱，等等。它一直在被改进，异步进行。所以智能体每次都会变得更聪明。然而，问题在于，我认为这对每个人都很有帮助的问题之一是，我们如何实现非常具体细粒度的控制？如果不是通过 RAG，如果不是通过条件输入，如果不是通过我们所有的提示直接进入输出，那太粗糙了。

This is one of the things that is probably going to be the most useful skill in the future. In fact, just listening to you talk about technology and your use of it, one of the most important things is systems understanding. Systems awareness, system design, system organization. Systems thinking. The reason for that is because most of the low-level things that have to be done are going to be done agentically anyways. They're going to be automated anyhow. So, whether it's in my generation, it's about compiling chips and synthesizing transistors and gates and functional blocks, all of that is now synthesized. So most of our designers are systems designers. In the case of software, most software is going to be done agentically anyhow. So you have to be much more able to think abstractly about systems. What are the problems you're trying to solve? What are the constraints? Where is the input? Where is the output? Where is information coming from? What is the rate of information flowing in and out of the system? What are the constraints? So is it processor? Is it memory? Is it networking? Under these systems problems at a sufficiently technical level, it's going to be very helpful to all of the people in this room. I don't think that fundamental knowledge is ever going to be useless. I think it's going to be more and more useful. I try to understand systems the best I can. Speaking of agents, the fact of the matter is we kind of have coarse-level recursive self-improvement already. Every time you use it, it improves the markdown files. Every time you use it, it updates its long-term memory. The long-term memory is being processed, either compacted or turned into knowledge graphs, and so on. It's being improved all the time, asynchronously. So the agent is getting smarter every time. Still, the problem is, and this is one of the problems that I think it'd be helpful for everybody to solve, is how can we have very specific fine-grained control? If not for RAGs, if not for conditional inputs, if not for all of our prompts directly into output, it was too coarse.

所以我们可以调节、可以控制智能体，一直到底层——当它提出一个计划时，我更改计划文件中的一个词，那个词就带来增量变化。不是完全改变，而是特定改变。可能是一个像素、一个三角形、CAD文件中的一个组件、一层、一个过孔、一个连接。然后它会重新生成所有其他部分。我认为这种对智能体的控制程度和协作水平将具有变革性。我们不需要智能体 100% 准确、100% 高质量才能使用它。它可能只有 80%，然后我们帮助它完成剩下的部分；或者它达到 99%，我们帮助它完成剩下的部分。所以我认为可控性可能是我们在各个层面都需要的最大的突破。

And so the fact that we can condition, the fact that we can control the agents all the way down to eventually when it comes up with a plan, I change one word in a plan file, and that one word makes a delta difference. Not complete difference, but specific difference. Maybe it's one pixel, maybe it's one triangle, maybe it's one component in a CAD file, maybe one layer, one via, one connection. And then it regenerates everything else. I think that level of control and that level of collaboration with agents will be game changing. We don't need the agents to be 100% accurate, 100% high quality in order for us to use it. It could literally be 80% and then we help it the rest of the way, or it could be 99% we help it the rest of the way. And so I think controllability is probably the single biggest breakthrough that we need for agents at every single level.

你认为人们会喜欢吗？我的意思是，Hermes 或 Open Claw，感觉这实际上有点关乎生存。人们应该控制自己的个人 AGI。他们不应该把这个应用外包，让它只在云端，成为别人告诉你该做什么的智能体。你希望它是你自己的。

Do you think people will like — I mean, with Hermes or Open Claw, it feels like that might actually be somewhat existential. Like people should control their own personal AGI. Like they shouldn't outsource that app and have it be just in the cloud and someone else's agent that kind of tells you what to do. Like you want it to be your own.

是的。这是英伟达如此深入参与背后的推力吗？我认为，首先，我需要理解智能体，因为智能体就是新软件。这个新软件如何处理对计算机架构非常重要。我们越深入了解智能体的本质以及它与聊天机器人的不同之处（聊天机器人又与最初的推理不同），当我们思考这些处理层时，我们就越能设计出更好的系统。我们不得不生活在未来 5 到 10 年，因为仅仅构建一个系统就需要大约三年，再用几年才能推广，而且你希望它们能在之后 10 年内被使用。所以你必须在一定程度上生活在未来。因此，对我们来说，从第一性原理出发，智能体系统就是：工作负载是什么？算法是什么？它将如何演变？瓶颈在哪里？阿姆达尔定律问题在哪里？如何扩展？并发会怎样？如何处理沙箱？如何处理 MCP？如何处理工作记忆和长期记忆？如何让所有这些自主系统、异步系统始终运行？那么，什么样的设计架构才最适合？我们必须去发现它。然后，第二件事是我希望我们自己使用智能体来加快英伟达的速度。所以我们有后台工作人员，我们有 Claude Code 在英伟达各地的沙箱中自主运行，这真的很棒。有些人使用 Codex，有些人使用 Claude Code，有些人使用 Cursor，有些人使用 Cognition。我们百花齐放，让人们选择他们想用的工具，然后我们从中学到很多。所以第二部分就是帮助公司更快发展。使用这些工具，他们用得越多，你就越能了解如何让它在未来更好地工作。最后一部分是发现未来解决方案的技术。也许当我们看到斯坦福早期版本的思维链时，现在大概是十年前，也许八年前。问题是它将在推理中起多大作用，它的可扩展性如何？对计算机视觉有何影响，如果我们能利用先验知识进行推理？然后大的突破当然就是在这个小领域思考，你会意识到也许我们不需要那么多数据来训练自动驾驶汽车。这促使我们创造了 Alpaca My，这是世界上第一辆会思考的自动驾驶汽车。只需要大约一百万英里，几百万英里，它就是一个非常棒的自动驾驶汽车。原因是它像我们一样，对吧？我们不需要那么多英里就能在一生中大部分时间驾驶得很好。原因是我们从语言模型中获得先验知识，我们可以分解从未见过的情况，并用我们很好理解的东西来构建它。所以这是一个例子，看到某件事然后意识到它以后的影响。当智能体系统出现时，很明显大型语言模型需要记忆、先验知识、工具以及与其他智能体网络连接的方式。所以一旦你看到一些早期指标，并且能够推理未来，它就能帮助你跃入未来。

Yeah. Is that part of the thrust behind Nvidia being so involved in? I think, well, first of all, I need to understand agents because agents is the new software. How this new software is processed matters a lot to computer architecture. The more intimate we are about the nature of agents and how it's different from chatbots, which is different from maybe inference in the very beginning. However, we think about these processing layers, the more intimate we are about the nature of the processing, the better we can design systems. We kind of have to live in the future 5 to 10 years because it takes about three years just to build a system, takes a couple years to ramp it up, and you would like them to be able to use the computer for 10 years after. So you kind of have to live in the future for a while. And so for us, agentic systems at first principles is just: what is the workload, what's the algorithm, how is it going to evolve, where are the bottlenecks, where are the Amdahl's law problems, how does it scale, what happens to concurrency? How do you deal with sandboxes? How do you deal with MCP? How do you deal with working memory, long-term memory? How do you have all these autonomous systems, asynchronous systems working all the time? And so, what kind of design architecture makes perfect sense for that? So we have to go and discover that. And then, of course, the second thing is I want to use agents ourselves to make NVIDIA go faster. So we have voices in the back and we have Claude Code autonomously running in sandboxes all over NVIDIA, and that's really fantastic. Some people use Codex, some people use Claude Code, some people use Cursor, some people use Cognition. We let a thousand flowers bloom, let people select the tools they want to use, and then we learn from all of that. So the second part is just helping the company move faster. Use the tools, and the more they use it, the more you learn about how to make it work better in the future. And then the last part is discovering the future of solutions technology for the future. And maybe when we saw the early versions of chain of thought out of Stanford, it was probably a decade ago at this point, maybe eight years ago. The question is how effective is that going to be in reasoning, and how scalable is it going to be? And what is the implication, for example, in computer vision, if we can reason from prior knowledge. And then the big breakthrough, of course, is just in thinking through that small little domain, you come to realize that maybe we don't need as much data for cars to train a self-driving car. Which led us to creating Alpaca My which is the world's first thinking self-driving car. With just a million miles or so, a couple million miles, it's an incredibly great self-driving car. The reason for that is it's kind of like us, right? We don't need that many miles before we could drive fairly well most of our lives. The reason is because we have prior knowledge from our language model, and we can decompose a situation we've never seen before and build it up out of things that we understood and know very well. So that's an example of seeing something and then realizing the impacts on sometime later. When agentic systems came along, it's very clear that obviously large language models need memory, need prior knowledge, need tools, need ways to network with other agents. So once you see some early indicators and you are able to reason about the future, it helps you get a leap into the future.

我觉得我在英伟达周围开始看到一种模式。就像你看到一个难题，有一个新算法，有新事物发生，然后你实际上就在那里提供开源。我的意思是，我记得当 OpenCL 出现时，人们说它不安全，但你们推出了一个沙盒工具包，它可以包裹任何框架并使其安全。所以，

I feel like there's this pattern that I'm starting to see around Nvidia. It's like you see a problem, there's a new algorithm, there's some new thing happening, and then actually you're right there with open source. I mean, I remember when OpenCL came out and people said it was unsafe, but you guys came out with a sandboxing sort of toolkit that surrounds any harness and makes it safe. And so,

当我看到 OpenCL 时，我的第一个想法是：首先，我了解了它。然后，不用太多想象力，你就意识到我们刚刚设计了现代计算机。这是将承载大型语言模型的操作系统。在很多方面，OpenCL 对我来说是一个‘Linux 时刻’。现在每个人都可以构建自己的 AI。我对此非常兴奋。我们联系了 Peter，我们说，‘嗨，英伟达的所有工程师都是你的工程师。我就是这么告诉 Peter 的。你家门外有这艘战舰。你可以随意拆分问题，我们将按你的意愿贡献。’Hermes 团队也是如此。我对他们所做的工作感到非常兴奋。我确实认为世界需要每个人都能构建自己的 AI。当然，你完全可以，我也鼓励大家尽可能使用云服务。每个人都应该使用 ChatGPT 和 Claude，对吧，每个人都应该使用。但如果你需要构建自己的 AI，因为你是一家公司，你需要构建你自己的领域特定 AI。

When I saw OpenCL, my first thought was: well, first of all, I learned about it. And then, without much imagination, you just realized we just designed the modern computer. This is the operating system that's going to hold a large language model. And in a lot of ways, OpenCL to me was a 'Linux moment'. And now everybody can build their own AI. I was so excited about that. We contacted Peter, and we said, 'Hey, all of Nvidia's engineers are your engineers. That's what I told Peter. You got this battleship outside your house. Break down the problem as you desire and we'll contribute as you wish.' Same thing with the Hermes team. I'm so excited about the work that they're doing. I do think that the world needs the ability for everybody to build their own AI. And you can, of course, and I encourage everybody to use cloud services as much as possible. Everybody should use ChatGPT and Claude, right, everybody should use that. But if you need to build your own AI because you're a company and you need to build your own domain-specific AIs.

现在你有 Hermes 和 Open Claude，还有 LangChain、Deep Agent 等各种方式可以构建自己的 AI。这相对容易，因为软件本身就很智能。AI 如此聪明，你也能轻松地让它适应你的需求。我希望鼓励每个人、每家公司都构建自己的 AI。谁知道开源会带来什么样的创新呢？我觉得所有价值都在于构建自己的 AI。如果别人用现成的，而你拥有一个可以递归自我改进的系统——搞技术的人对 markdown 文件很不屑，说只是文本而已，但文本就是智能。我们身处一个完全不同的世界。词语就是思想。没错，词语就是思想。而且事实证明，你可以试着不用词语思考。对，就是这样。

Now you have Hermes and Open Claude, you have LangChain, Deep Agent, all these different ways to build your own AI. It's relatively easy because the software is smart. AI is so smart you can adapt it easily. I think we want to encourage everybody and every company to build their own AIs. Who knows what innovation will come from the fact that it's open source? I feel like all the alpha is in building your own AI. If someone else uses off-the-shelf tools, but you have a thing that can recursively self-improve — tech people are flippant about markdown files, saying it's just text, but text is intelligence. We're in a different world. Words are thoughts. Yeah, words are thoughts. And it turns out you can try to think without words. Yeah, that's right.

再次转换话题。每当格局改变，人们总会有些担忧。智能将唾手可得，这非常棒。我认为这对在座各位都是好兆头。你觉得经济会发生什么变化？从更宏观的角度看，会发生什么？

So switching gears again, a lot of people are anytime you move the cheese, people get a little worried. Intelligence is going to be on tap, which is really awesome. I think it bodes well for everyone in this room. What do you think changes about the economy? What do you think happens in a broader sense?

显然，我的说法并不均衡。我们会自动化任务，特别是认知任务。如果一项任务是接听电话，根据掌握的信息提供答案，那么这项任务就会被自动化。先不谈这个，我想谈谈巨大的机遇。许多任务会被自动化，每个工作都会改变，同时会出现大量新工作。关键在于：证据显示 AI 和自动化正在各地创造就业。AI 消灭工作的说法完全是反的。AI 消灭的是任务，但不一定消灭工作。一个人的工作有其目的，包含许多任务，有些可自动化，很多则不能。证据：我们自动化了编程，但软件工程师岗位每年增长 10%。我们自动化了放射扫描，但放射科岗位在过去几年增长了 20%，即使 AI 已覆盖该领域。原因在于患者积压严重，医生可以接诊更多病人，从而需要更多护士和放射科医生。软件行业同理：想法积压很多；自动化编程让我们能雇佣更多工程师。有人曾预言 Harvey 会淘汰法律助理，但法律助理增长迅猛，因为诉讼积压严重，律所雇佣更多人。这是生产力推动增长、增长带动就业的经典例子。这就是为什么现在的就业比我刚毕业时更多。

Obviously, what I'm going to say is uneven. We're going to automate tasks, cognitive tasks. If a task is somebody makes a phone call and sends words to you, and your job is to provide a response with all information at your fingertip, that task will be automated away. Ignoring that for a second, my point is about the great opportunity. Many tasks will be automated away. Every job will change, and there will be a whole bunch of new jobs. The bottom line is: evidence shows AI and automation are creating jobs everywhere. The narrative about AI destroying jobs is exactly backwards. AI eliminates tasks, but not necessarily jobs. A person's job has a purpose with many tasks. Some can be automated, many cannot. Evidence: we automated coding, but software engineer jobs grew 10% year over year. We automated reading radiology scans, but radiology jobs increased 20% in the last several years, even though AI took over. Reason: backlog of patients is high, so doctors admit more patients, needing more nurses and radiologists. Same with software: backlog of ideas is high; automating programming lets us hire more engineers. Harvey was supposed to eliminate paralegal jobs, but paralegals are growing like crazy because lawsuit backlog is high, and law firms hire more people. This is a classic example: productivity increases growth, which drives more employment. That's why there is more employment today than when I first came out of school.

我们一直在讨论软件和智能体。另一个令人兴奋的领域是实体机器人，Nvidia 在这方面走在前沿。距离还有多远？我记得你以前说过今年内就能实现。关于实际机器人何时可用，最新的想法是什么？

So we've been talking about software and agents. Another exciting thing Nvidia is on the edge of is physical robots. How far out? I think you once said as soon as this year. What's the latest thinking on when we can expect practical robotics?

当我看到生成式视频时，那是一个伟大的时刻。我们在渐进式 GAN 和条件 GAN 上做了原创工作。在人们看到外部生成的视频多年前，我们实验室内部就已经在用完全由视频和神经网络驱动的模拟器了。当我看到我们能生成关节动作——如果能生成手指移动、手拿杯子的视频——为什么不能让机器人也做到呢？我意识到机器人关节动作即将到来。现在的问题是机器人如何理解生成符合物理、因果、摩擦、张力的运动。这促使我们开始创建所谓的物理 AI。我们开始研究世界基础模型，它们理解物理和世界运作方式。机器人的 ChatGPT 时刻发生在几年前。ChatGPT 刚出来时没做什么有用的事，但开启了我们的想象力。几年前，通过强化学习、微调并基于物理的行走机器人已经实现。现在我们需要做与智能体系统相同的事情：为它们创造学习和评估的环境。我们必须实现真实到模拟，生成基于模拟的、有物理基础的模拟器，以及生成式物理模拟。

The moment I saw generative video was a great moment. We did original work on progressive GANs and conditional GANs. Years before people saw generated videos outside, inside our labs we were driving a simulator completely generated by video, by neural networks. When I saw us generating articulation — if I can generate video of a finger moving, of a hand picking up a glass — why can't a robot do the same? I realized robotics articulation was around the corner. Now the question is how the robot understands to generate motions obeying physics, causality, friction, tension. That started us on creating what we call physical AI. We started working on world foundation models that understand physics and how the world works. The ChatGPT moment of robots happened a couple of years ago. When ChatGPT first came out, it didn't do anything productive, but opened our imagination. A couple of years ago, robots walking with reinforcement learning, fine-tuned and grounded in physics, really happened. Now we need to do the same as for agentic systems: create environments for learning and evaluation. We have to do real-to-sim, generate simulators based on simulation, grounded physics simulation, and generative physics simulations.

因此，Isaac Sim、Cosmos 以及我们在这个领域所做的一切工作都与仿真相关。最后一部分是从仿真到现实。这部分与强化学习有关，将其建立在物理基础上，建立在机器人所需的所有机电系统基础上。但我想，这三个基本系统构成了机器人技术的评估（如果你愿意这么说的话），也就是机器人技术的后训练。我认为我们很快就能看到它。

And so, Isaac Sim, Cosmos, and all the work that we do in that area is related to simulation. And then the last part is sim to real. And so, that part has something to do with reinforcement learning, grounding it on physics, grounding it on all the electromechanical systems that robots require. And so, but these three basic systems, I think, build up the evaluation, if you will, the post-training of robotics. And I think we're going to see it right around the corner.

太棒了。物理 AI 首先出现在哪个领域，并且真正具有经济现实意义？你已经看到了吗？

Amazing. Where does physical AI show up first in a way that's really economically real? Are you seeing that already?

我们推测机器人技术会发展起来，并决定机器人技术的第一个应用既要具有足够大的市场，又要具有相对标准化的技术以便我们能够扩展并获得飞轮效应，还要具有真正的经济价值，那就是自动驾驶汽车。所以，在 Waymo 内部，我们使用了 Nvidia 的芯片。在 Tesla，我们的芯片也在车里。现在我们在数据中心里。梅赛德斯，我们在数据中心里，也在车里提供软件栈。我们开发了 Alpaca Myo，并将其开源。我们开源自动驾驶汽车软件栈的原因是，农业、邮件投递、仓库 AMR 都需要它。有太多不同的方式可以应用自主导航，而这些市场没有一个是足够大的自动驾驶汽车市场，我们认为它足够多样化，因此我们为它创建了整个软件栈。所以我们正在各种不同的地方与自动驾驶车辆合作。我们的机器人业务、自动驾驶汽车业务，基本上就是物理 AI 业务，可能已经接近 100 亿美元了。所以它已经非常庞大了。这很可能将成为世界上最大的产业之一，时间不会只是两三年，但也不会超过十年。所以这将是我们下一个 1000 亿美元的业务。

We conjectured that robotics was going to come along and decided that the first application of robotics that has both a large enough market, relatively standardized technology so that we could scale and get the flywheel going, and has real economic value was self-driving cars. And so, inside Waymo, our chips from Nvidia. At Tesla, we were in the car. Now we're in the data center. Mercedes, we're in the data center, we're in the car with a software stack. We worked on Alpaca Myo, and we open-sourced it. And the reason why we open-sourced the self-driving car stack is because you need it for agriculture, you need it for mail delivery, you need it for warehouse AMRs. There's so many different ways that you could apply autonomous navigation, and none of those markets are big enough to be a self-driving car market and we thought it was sufficiently diverse that we would create the whole stack for it. And so we're working with autonomous vehicles in all kinds of different places. Our robotics business, autonomous vehicle business, basically physical AI business is probably almost like $10 billion. So it's really, really big already. Likely this will be one of the largest industries in the world and it'll take longer than a couple two, three years. It'll take less than 10. And so this will be our next $100 billion business.

太棒了。嗯，我想花点时间。我认为在场的观众非常合适，也许我们可以在这个舞台上欢迎 Jensen 加入 X。欢迎来到 X。我是说，你发布了第一条帖子，感谢你的领导力。

Amazing. Um, I want to take a moment. I think this is the exact right crowd to, you know, maybe as an arena we can welcome Jensen to X. Welcome to X. I mean, you made your first post, and thank you for your leadership.

你知道，这正好说明我有多内向。我直到 2026 年才在 X 上发布了第一条帖子。我可能是地球上最后做这件事的人。但我发的内容对我来说太重要了，对整个行业和世界也太重要了。所以我克服了害羞，在 X 上发布了我的第一条内容。

You know, that just shows you how introverted I am. It took me until 2026 to have the first post on X. You know, I'm probably the last human on Earth that did it. But what I posted was too important to me and too important to the industry and too important to the world. And so I overcame my shyness and put my first thing out on X.

不，感谢你的领导力。我的意思是，开源、开放权重、开源模型对于在座所有人想要做的事情来说非常重要。我们想要创造产品。

No, thank you for your leadership. I mean, open source, open weights, open source models are incredibly important for what all of us in this room want to do. Like we want to create products.

如果没有开源，移动云产业就永远不会出现。如果没有 Linux，如果没有 Kubernetes，如果没有所有这些平台，如果没有 TensorFlow 或更重要的 PyTorch，以及 Caffe 和 Torch 的早期版本，所有 Theano。还记得所有这些的早期版本都是开源的。如果没有这些，我们怎么会有现代 AI？

If not for open source, the mobile cloud industry would have never happened. If not for Linux, if not for Kubernetes, if not for all of these platforms, if not for TensorFlow or more importantly, PyTorch, and the early versions of Caffe, Torch, all of Theano. Remember the early versions of all those were all open source. If not for all of that, how would we have modern AI?

好吧，感谢你的领导力，你的声音在这里非常重要。谢谢你。

Well, thank you for your leadership and your voice is incredibly important here. Thank you.

在我们结束之前，我觉得与你的故事产生了强烈共鸣。我想这里的每个人，我的意思是，都想听听你一路走来的智慧。我看到你看到了很多东西，所有将在社会中占据主导地位的算法。基于你所看到的，一个年轻人现在应该学习什么，才能在未来仍然重要？

Before we go, I feel like I just really resonate with your story. I think that everyone here, but I mean, would love the wisdom of your journey coming here. I mean, what should a young person learn now, given all the things that you're seeing, all the algorithms that are going to take hold in society? What should a young person learn now that will still matter, based on what you're seeing?

嗯，我今天看到的一些东西和遇到的一些初创公司令人非常鼓舞。重要的收获是，简单的工作当然会被自动化掉。当我说简单的工作时，我指的是软件，比如编程。你坐在电脑前实际编写代码来解决问题的想法，显然会被自动化掉。你知道，在我那一代人成长的过程中，我们必须做长除法。天哪，谁还需要学长除法？所以，那被编码掉了，被自动化了。所以，我认为简单的工作会被自动化，但困难的问题、硬科学——物理、化学、生物学、计算机科学、计算机工程、系统思维，特别是交叉领域，这些难题永远不会消失。所以，AI 只是一个不可思议的工具，帮助我们变得比以前更有雄心，更急于解决这些极其庞大和难以置信的难题。所以，如果你看看我这一代，在我刚毕业时，一个芯片设计师可能设计一个只有一千个晶体管的芯片，那已经是很庞大的芯片了。现在设计一个万亿晶体管的芯片甚至不——如果有人告诉我，'Jensen，我们的下一个芯片有一万亿个晶体管'，我会说，'好的。'这不是什么大事。原因是我们现在非常有雄心，问题的规模、任务的规模不再是问题。所以，你不需要担心要写多少代码、需要多少工程师。你不需要再考虑那些事情了。你只需要考虑要解决什么问题。所以，我认为深科技、深科学，理解技术和社会问题的交叉点，理解市场空白和机会，所有这些仍然存在。你越擅长系统思维，能够协调数百万个智能体自主解决问题，你就越有优势。这就是为什么系统思维如此重要。但除此之外，我认为世界将继续有许多伟大的挑战等待我们去解决。按老办法上学吧。待在学校里。

Well, some of the things that I saw today and some of the starters I met today was really quite encouraging. And the big takeaway is, of course, the simple stuff is going to get automated away. And when I say simple stuff, I mean software, you know, coding. The idea that you would solve a problem by sitting in front of a computer and actually writing code, that concept is obviously going to get automated away. You know, in my generation, when I was growing up, we had to do long division. For God's sakes, who has to learn long division? And so, that got coded away, that got automated away. And so, I think the simple stuff is going to get automated away, but the hard problems, the hard sciences, physics, chemistry, biology, computer science, computer engineering, systems thinking, and particularly the domains that are intersecting, those hard problems will never go away. And so, AI is just an incredible tool that helps us become even more ambitious, even more impatient about solving these extraordinarily large and incredibly hard problems than before. And so, if you look at my generation, when I first graduated, a chip designer would design a chip with maybe a thousand transistors, and that would be a very large chip. Now designing a trillion transistor chips is not even — if somebody would have told me, 'Jensen, our next chip is a trillion transistor,' I said, 'Okay.' It's not a thing. And the reason for that is because we are so ambitious now, the scale of the problem, the scale of the task is no longer a matter. And so, you don't have to worry about how much coding, how many engineers. You don't have to think about those things anymore. You just have to think about what is the problem you have to solve. And so, I think that the deep tech stuff, the deep science stuff, understanding the intersection between technology and social issues, understanding market gaps and holes, opportunities, I think all of that still exists. And the better you are at systems thinking so that you could orchestrate millions of agents solving problems autonomously, the better off you are. And so, that's why systems thinking is going to be so important. But otherwise, I think the world's going to continue to have a lot of great challenges for us to solve. Go to school the same old way. Stay in school.

待在学校里。

Stay in school.

我想我通常喜欢以你看着人群来结束。有这么多人——我的意思是，我在开场时就说过，我真诚地看着人群，我看到的人和我们本质上没有区别。你知道，我们实际上只是技术人员，热爱系统。你怎么知道——

I guess I usually like to end with you looking out on the crowd. There are a lot of people who — I mean, I started the opener with like I honestly look in the crowd and I see people who are not different than us per se. You know, we actually just are technical and like love systems. How you know —

谢谢你。

Thank you.

你会给这个房间里的人什么建议？你在这个房间里看到了自己，我很好奇你会说什么。如果你能给 18 到 22 岁的自己发一封电报、一条消息，那会是什么？

What advice would you give to this room of, you know, and you see yourself in this room and I'm curious what you would say. If you could send a telegram, a message to the 18 to 22-year-old version of yourself, what would that be?

我可以准确告诉你英伟达成立、我们三个人开始时我的感受。当时我觉得有太多东西需要我知道，太多东西需要我学习，而我当时并不知道。我早些时候告诉过你，那时没有 YouTube，没有 YC，没有人教你如何创业。所以我去了书店，买了一本书，书名是《如何创业？》。不幸的是，那本书有 500 页。我想等我读完，公司早就倒闭了，我和 Lori 的钱也花光了。所以读它毫无意义。但我记忆犹新的是，去筹钱时我有多害怕，因为我觉得要跟一群人谈话，却不知道如何回答他们的问题。这是真的。即使到今天，我也几乎不知道如何回答他们的问题。但我学到的是，那些都不重要，事实证明。你永远会有你不知道的事情。每天世界在变，技术在变。显然，这是过去 60 年里创办公司最好的时机。整个行业都变了。从技术角度来看，这是一个彻底的重新开始。人类历史上最重要的技术——计算机，被完全重置了。所以这绝对是创办公司的最佳时机。我嫉妒你们所有人以及你们面前的机会。这将是不可思议的。所以一方面这是完美时机。另一方面，技术变化如此之快。所以问题是，对你来说正确的感受是什么？最终，我跟你讲过买另一本教科书的故事。我认为我今天对所有新体验、新技术、新市场和新动态的心态和感受是，我看着它说，这很重要。我必须去学习它。我必须去做点什么。而且我最好尽快开始。能有多难？我一直有这种感觉，能有多难？老实说，它比你想象的要难得多。但你不希望你的头脑停留在那里。你希望你的头脑想，能有多难？让痛苦一点点到来。不要想象它会有多难，让那变成焦虑而不去行动。你希望在你的脑海里想，能有多难？反正我有一堆 AI 智能体在帮我。所以能有多难？然后你就开始着手做。这大概就是企业家的态度。你必须在过程中学习很多东西。你相信自己的学习能力。学习是最大的超能力。如果你带着“能有多难”的态度去做，如果别人能做到，我也能做到。并且要意识到它会很难，你必须有韧性每天克服它。你不需要一天克服一生。你只需要克服那个早晨。那个早晨，你需要克服今天。所以没什么大不了的。只要熬过今天。为明天努力。继续追随你的梦想。如果你坚持得足够久，英伟达就会发生。所以我认为我能分享的智慧，如果有的话，就是韧性可能是最重要的东西。如果你相信某件事，就去做，把你的头脑从因为恐惧、焦虑或缺乏自信等而阻止你追求它的状态中拉出来。然后你只需要告诉自己，我会在学习中找到路。

I could tell you exactly how I felt when NVIDIA was founded and the three of us started. The thing I felt at the time is there was so much for me to know and so much for me to learn. And I didn't know it. I was telling you earlier, at the time there were no YouTube, no YC, nobody teaching you how to start a company. So I went to the bookstore and bought a book that said "How to start a company?" Unfortunately, the book was like 500 pages long. And I figured by the time I read it, I'd be out of business and Lori and I would be out of money. So there's no sense reading it. But the thing I remember very vividly is how scared I was to go raise money because I felt I was about to talk to a bunch of people and I didn't know how to answer their questions. And it's true. I barely know how to answer their questions even today. But the thing I learned is none of that stuff matters, as it turns out. And you're always going to have things you don't know. Every single day the world's changing, technology changing. Obviously, this is the greatest time in the last 60 years to start a company. The whole industry has changed. It's a complete reset from a technology perspective. The single most important technology in human history, the computer, has been completely reset. So this is absolutely the single greatest time to start a company. I'm jealous of all of you and the opportunities you have ahead. It's going to be incredible. So it's the perfect time on the one hand. On the other hand, the technology is changing so fast. So the question is, what's the right feeling for you? Eventually, I told you the story of me buying the other book, the textbook. I think the psychology and the feeling I have today on all the new experiences and new technology and new markets and new dynamics, I look at it and say, this is important. I've got to go learn it. I've got to go do something about it. And I better get to it as fast as I can. And how hard can it be? I always had this feeling, how hard can it be? And truth be told, it is way harder than you think. But you don't want your mind to be there. You want your mind to be, how hard can it be? And let the suffering come to you a little bit at a time. Don't imagine how hard it's going to be and let all that turn into anxiety and not doing something about it. You want to imagine in your head, how hard can it be? I've got a bunch of AI agents helping me anyway. So how hard can it be? And then you get going on working on it. So that's probably the attitude of an entrepreneur. You have to learn a bunch of stuff along the way. You believe in your ability to learn. Learning is the single greatest superpower. And if you go into it with the attitude, how hard can it be? If anybody can do it, I can do it. And just realize that it will be hard and you just have to have the resilience to overcome it every single day. You don't have to overcome life in one day. You just have to overcome that morning. That morning, you have to overcome today. So it's not a big deal. Just get through today. Work towards tomorrow. Keep following your dreams. And if you stick with it long enough, NVIDIA happens. So I think the wisdom I can share, if anything, is resilience is probably the single most important thing. And if you believe in something, just get going on it and get your mind out of keeping yourself from pursuing it because of fear or anxiety or lack of confidence or whatever. And then you're just going to tell yourself I'm going to learn my way there.

Jensen Huang，各位。好了，谢谢大家。

Jensen Huang everybody. All right, guys. Thank you.

非常感谢。是的，确实如此。谢谢大家。

Thank you so much. Yes, it was. Thank you guys.