黄仁勋谈分解推理与 AI 工厂

本周是特别节目。我们暂停了常规节目，只有三个人能让我们这么做：特朗普总统、耶稣和 Jensen。顺序由你定。但你这一路走来真是太精彩了，活动也很棒。

Special episode this week. We've preempted the weekly show and there's only three people we preempt the show for. President Trump, Jesus, and Jensen. And I'll let you pick which order we do that. But what an amazing run you've had and a great event.

每个行业都来了，每家科技公司都来了，每家 AI 公司都来了。太不可思议了。

Every industry is here. Every tech company is here. Every AI company is here. Incredible.

确实不可思议，非同凡响。过去一年最重大的公告之一就是 Grok。你收购 Grok 的时候，有没有意识到 Chamath 会变得这么让人受不了？

Incredible. Extraordinary. And one of the great announcements of the past year has been Grok. When you made the purchase of Grok, did you realize how insufferable Chamath would become?

我隐约有点感觉……

I had an inkling that that that...

我们是他的朋友，每周都得应付他。

We're his friends. We have to deal with him every week.

我知道。

I know it.

还得忍受他六周的闭市期。

To deal with him for the 6-week close.

我知道。两周。现在全想起来了，让我挺不舒服的。问题是，我们的很多策略在 GTC 上都是提前几年公开的。两年半前，我介绍了 AI 工厂的操作系统，叫 Dynamo。Dynamo，你知道，是西门子发明的一种机器，把水变成电。Dynamo 驱动了上一次工业革命的工厂。所以我觉得它是下一次工业革命操作系统——也就是 AI 工厂操作系统——的完美名字。在 Dynamo 内部，核心技术是解耦推理。Jason，我知道你技术很强。

I know it. 2 weeks. It's all coming back to me now. It's making me rather uncomfortable. The thing is many of our strategies are presented in broad daylight at GTC years in advance of when we do it. 2 and a half years ago, I introduced the operating system of the AI factory and it's called Dynamo. Dynamo, as you know, is a piece of instrument, a machine that was created by Siemens to turn essentially water into electricity. And Dynamo powered the factory of the last industrial revolution. So I thought it was the perfect name for the operating system of the next industrial revolution, the factory of that. And so inside Dynamo, the fundamental technology is disaggregated inference. Jason, I know you're super technical.

当然。

Absolutely.

这个你来吧。给观众解释一下，我不想抢你的话。

I'll let you take this one. Go ahead and define it for the audience. I don't want to step on you.

谢谢。我知道你刚才想插话。但这就是解耦推理，意思是推理的处理流水线极其复杂。事实上，这是当今最复杂的计算问题。规模惊人，有各种形状和大小的数学运算。我们想到，可以把处理过程的不同部分解耦，让一部分在某个 GPU 上运行，其余部分在不同的 GPU 上运行。这让我们意识到，也许解耦计算本身也有意义。我们可以拥有异构的计算特性。同样的思路也让我们走向了 Mellanox。

Yeah, thank you. I knew you wanted to jump in there for a second. But it's disaggregated inference, which means the pipeline, the processing pipeline of inference is extremely complicated. In fact, it is the most complicated computing problem today. Incredible scale, lots of mathematics of different shapes and sizes. And we came up with the idea that you would disaggregate parts of the processing such that some of it can run on some GPUs, the rest of it can run on different GPUs, and that led to us realizing that maybe even disaggregated computing could make sense. That we could have different heterogeneous nature of computing. That same sensibility led us to Mellanox.

没错。如今 Nvidia 的计算分布在 GPU、CPU、交换机、纵向扩展交换机、横向扩展交换机、网络处理器上，现在还要加上 Groq，我们要把合适的工作负载放在合适的芯片上。我们真的从一家 GPU 公司进化成了一家 AI 工厂公司。

Yep. Today Nvidia's computing is spread across GPU, CPUs, switches, scale-up switches, scale-out switches, networking processors, and now we're going to add Groq to that, and we're going to put the right workload on the right chips. We just really evolved from a GPU company to an AI factory company.

我想这大概是我最大的收获。你看到了这种根本性的解耦，我们从 GPU 出发，现在有了所有这些不同选项的复杂组合，最终会存在。你们在台上说的，或者你在台上说的：我希望高价值推理的人听好了，你们数据中心 25%的空间应该分配给这种 Groq LPU GPU 组合。

I think that was probably the biggest takeaway that I had. You're seeing this fundamental disaggregation where we've gone from a GPU, and now you have this complexion of all these different options that will eventually exist. The thing that you guys said on stage, or you said on stage, was: I would like the high-value inference people to take a listen to this, and 25% of your data center space you said should be allocated to this Groq LPU GPU combo.

大约是数据中心里 Vera Rubin 的 25%。

To about 25% of the Vera Rubins in the data center.

那么，你能告诉我们业界如何看待这个想法吗？就是现在基本上创建下一代解耦的预填充、解码，你觉得人们会有什么反应？

So, can you tell us about how the industry looks at this idea of now basically creating this next generation form of disaggregated prefill, decode, and how people do you think will react to it?

是的。退一步看。当我们加入这个时，我们从大语言模型处理转向了智能体式处理。现在，当你运行一个智能体时，你在访问工作记忆、长期记忆，使用工具，真的在狠狠地压榨存储。智能体与其他智能体协作。有些智能体是非常大的模型，有些是较小的模型，有些是扩散模型，有些是自回归模型。所以数据中心里有各种不同类型的模型。我们创建 Vera Rubin 就是为了能够运行这种极其多样化的工作负载。我的感觉是，我们增加了一个原本只有一机架的公司，现在又增加了四个机架。

Yeah. Take a step back. At the time that we added this, we went from large language model processing to agentic processing. Now, when you're running an agent, you're accessing working memory, you're accessing long-term memory, you're using tools, you're really beating up on storage really hard. You have agents working with other agents. Some of the agents are very large models, some of them are smaller models, some of them are diffusion models, some of them are auto regressive models. And so there's all kinds of different types of models inside this data center. We created Vera Rubin to be able to run this extraordinarily diverse workload. My sense is, we added what used to be a one rack company, we now add four more racks.

对。

Right.

所以 Nvidia 的 TAM，如果你愿意这么说的话，从原来的水平增加了大概 33%或 50%。其中 33%或 50%里，很大一部分将是存储处理器，叫 Bluefield。一部分，我希望很大一部分将是 Groq 处理器，还有一些是 CPU。它们都将，很大一部分将是网络处理器。所有这些基本上都在运行 AI 革命的计算机，也就是智能体。现代工业的操作系统。

So Nvidia's TAM, if you will, increased from whatever it was to probably something, call it 33% or 50% higher. Now, part of that 33% or 50%, a lot of it's going to be storage processors, it's called Bluefield. Some of it will be, a lot of it I'm hoping will be Groq processors, and some of it will be CPUs. And they're all going to, a lot of it's going to be networking processors. And so all of this is going to be running basically the computer of the AI revolution called agents. The operating system of modern industry.

嵌入式应用呢？比如，我女儿在家里的泰迪熊想和她说话。里面用什么？是定制 ASIC，还是最终会变成更广泛的 TAM，配合针对边缘和嵌入式应用不同用例而开发的工具？

What about embedded applications? So, my daughter's teddy bear at home wants to talk to her. What goes in there? Is it a custom ASIC or does there end up becoming much more kind of a broader set of TAM with developing tools that are maybe different for different use cases at the edge and in embedded applications?

这个问题涉及三台计算机。从最大规模来看，退一步讲。有一台计算机是真正用来训练 AI 模型、开发和创造 AI 的。另一台用于评估它。取决于你遇到的问题类型，比如你看看周围，有各种机器人和汽车之类的东西。你必须在代表物理世界的虚拟训练场中评估这些机器人。所以它必须是遵循物理定律的软件。那是第二台计算机。我们称之为 Omniverse。第三台计算机是边缘计算机，机器人计算机。那个机器人计算机，其中一个可以是自动驾驶汽车，另一个是机器人，另一个可以是泰迪熊。给泰迪熊用的很小的一个。最重要的之一是我们正在研发的，它基本上把电信基站变成了 AI 基础设施的一部分。所以现在所有的……这是一个 2 万亿美元的产业。所有这一切最终都将转变为 AI 基础设施的延伸。所以，无线电设备将成为边缘设备。工厂、仓库，等等。所以有这三台基本计算机，它们都是必需的。

There are three computers in the problem. At the largest scale, when you take a step back. There's one computer that's really about training the AI model, developing, creating the AI. Another computer for evaluating it. Depending on the type of problem you're having, like for example, you look around, there's all kinds of robots and cars and things like that. You have to evaluate these robots inside a virtual gym that represents the physical world. So it has to be software that obeys the laws of physics. And that's a second computer. We call that Omniverse. The third computer is the computer at the edge, the robotics computer. That robotics computer, one of them could be self-driving car, another one's a robot, another one could be a teddy bear. Little tiny one for a teddy bear. One of the most important ones is one that we're working on that basically turns the telecommunications base stations into part of the AI infrastructure. So now all of the... It's a $2 trillion industry. All of that in time will be transformed into an extension of the AI infrastructure. And so, radios will become edge devices. Factories, warehouses, you name it. So there are these three basic computers, all of them are going to be necessary.

Jensen，去年，我觉得你领先于全世界，你说推理不会只增长一千倍……

Jensen, last year, I think you were ahead of the rest of the world in saying inference isn't going to a thousand...

就在去年。

Just last year.

是的。它会增长一百万倍吗？十亿倍？

Yes. Is it going to 1 million X? Is it going to 1 billion X?

是的。

Yeah.

对。我想当时人们觉得这相当夸张，因为世界仍然专注于预训练 Scaling，专注于训练。现在我们到了这里。推理已经爆发了。我们受推理限制。

Right. And I think people at the time thought it was pretty hyperbolic because the world was still focused on pre-scaling, on training. Here we are. Now inference has exploded. We're inference constrained.

你宣布了一个我认为是领先的推理工厂，吞吐量将比下一个工厂好 10 倍。但外界传言说你的推理工厂要花 400-500 亿美元，而定制 ASIC、AMD 等替代方案只需 250-300 亿美元，你会失去市场份额。跟我们聊聊吧？你看到了什么？你怎么看待市场份额？大家支付两倍的溢价合理吗？

You announced an inference factory that I think is leading edge, going to be 10x better in terms of throughput than the next factory. But the chatter is that your inference factory will cost $40-50 billion, while alternatives like custom ASICs and AMD will cost $25-30 billion, and you'll lose share. Why don't you talk to us? What have you seen? How do you think about share? Does it make sense for folks to pay a 2x premium?

关键点在于，你不应该把工厂的价格和 token 的成本等同起来。很可能 500 亿美元的工厂会带来最低成本的 token。我们以极高的效率生产这些 token。500 亿和 200 亿的差别只是土地、电力和外壳。除此之外，存储、网络、CPU、服务器、冷却都是必需的。GPU 价格是 1 倍还是半倍，差别不是 500 亿对 300 亿，而是 500 亿对 400 亿。当 500 亿美元的数据中心实际吞吐量是 10 倍时，这个百分比并不大。

The big takeaway is that you should not equate the price of the factory with the cost of tokens. It's very likely that the $50 billion factory will generate the lowest cost tokens. We produce these tokens with extraordinary efficiency. The difference between $50 billion and $20 billion is just land, power, and shell. On top of that, you have storage, networking, CPUs, servers, cooling anyway. The difference between the GPU being 1x price or half price is not $50 billion vs $30 billion, but maybe $50 billion vs $40 billion. That's not a large percentage when the $50 billion data center actually has 10x throughput.

对。

Right.

这就是为什么我说，即使对大多数芯片来说，如果你跟不上技术状态和我们前进的步伐，即使芯片免费，也不够便宜。

That's why I said that even for most chips, if you can't keep up with the state of technology and our pace, even when chips are free, it's not cheap enough.

是的。我能问一个战略问题吗？你经营着世界上最有价值的公司，明年收入超过 3500 亿美元，自由现金流 2000 亿美元，以疯狂的速度复合增长。你怎么决定做什么？你怎么获取信息？如何塑造市场、加倍下注、撤退或进入新领域？

Yeah. Can I ask a general strategy question? You're running the most valuable company in the world, with $350+ billion revenue next year, $200 billion free cash flow, compounding at crazy rates. How do you decide what to do? How do you get information? How do you shape the market, double down, pull back, or go into green fields?

最终，这是 CEO 的工作。我们的工作是定义战略和愿景。我们受到杰出的计算机科学家和技术专家的启发，但我们必须塑造未来。部分考量是：这件事是否极其困难？如果容易，我们应该退后，因为会有很多竞争对手。这是否是前所未有、能发挥我们特殊超能力的事情？我们也知道其中会有很多痛苦和磨难。没有伟大的事物是因为容易而发明的。如果它超级困难且从未有人做过，你会经历痛苦，所以你最好享受它。

Ultimately, that's the CEO's job. Our job is to define the strategy and vision. We're informed by amazing computer scientists and technologists, but we have to shape the future. Part of it is: is this insanely hard to do? If it's easy, we should back away because there will be many competitors. Is it something never done before that taps into our special superpowers? We also know a lot of pain and suffering will go into it. No great things are invented because they were easy. If it's super hard and never done, you'll have pain, so you better enjoy it.

那么你能看看你宣布的三四个长尾项目，谈谈它们的长期可行性吗？无论是太空数据中心、汽车 ADAS 还是生物学？让我们了解你如何看待这些曲线的拐点。

So can you look at three or four of the longer-tail things you announced and talk about their long-term viability—whether it's the data center in space, ADAS in autos, or biology? Give us a sense of how you see these curves inflecting.

很好。物理 AI 是一个大类别。我们有三个计算系统和软件平台。物理 AI 是科技行业首次有机会涉足一个 50 万亿美元的行业，这个行业此前基本没有技术。我们 10 年前开始，现在正在拐点。这是一个数十亿美元的业务，接近每年 100 亿美元，呈指数增长。对于数字生物学，我们接近 ChatGPT 时刻。我们即将理解如何表示基因、蛋白质、细胞。我们已经理解化学物质。在两到五年内，我们将能够表示和理解生物学构建块的动态。五年内，医疗保健将受到数字生物学的影响。这些都是很好的机会，而且就在我们身边。

Excellent. Physical AI is a large category. We have three computing systems and software platforms. Physical AI is the tech industry's first opportunity to address a $50 trillion industry that has been largely void of technology. We started 10 years ago, and it's inflecting now. It's a multi-billion dollar business, close to $10 billion a year, growing exponentially. For digital biology, we are near the ChatGPT moment. We're about to understand how to represent genes, proteins, cells. We already understand chemicals. In two to five years, we'll represent and understand the dynamics of biology's building blocks. In five years, healthcare will be impacted by digital biology. These are great opportunities, and they're all around us.

农业。

Agriculture.

正在拐点。

Inflecting now.

毫无疑问。Jensen，我想把你从数据中心带到桌面。公司建立在爱好者、视频游戏玩家和显卡之上。你提到了 Claude Code 和智能体的革命。爱好者想要桌面。你发布了戴尔 6800 工作站，750GB 内存。Mac Studio 卖光了。我的公司正在转向开放 Claude。这个桌面上的开源智能体运动对你意味着什么？它会走向何方？

No question. Jensen, I want to take you from the data center to the desktop. The company was built on hobbyists, video gamers, and graphics cards. You mentioned Claude Code and the revolution of agents. Hobbyists want desktops. You announced the Dell 6800 workstation with 750GB RAM. The Mac Studio sold out. My company is moving to open Claude. What does this movement of open-source agents on the desktop mean to you? Where is it going?

很好。首先，让我们退一步。在过去两年里，我们看到了三个拐点。第一个是生成式 AI。ChatGPT 让 AI 进入了每个人的意识。但这项技术早在几个月前就存在了；ChatGPT 为它加上了用户界面，使其易于使用。生成式 AI 生成用于内部和外部消费的 token。

Great. First, let's step back. In the last two years, we saw three inflection points. The first was generative AI. ChatGPT brought AI to everyone's awareness. But the technology existed months before; ChatGPT put a user interface around it, making it easy to use. Generative AI generates tokens for internal and external consumption.

内部消耗就是思考，这导致了推理。01 和 ChatGPT 的浪潮让信息更接地气，使 AI 不仅能回答问题，还能以更实用、更有用的方式回答。我们开始看到 OpenAI 的收入和经济模式出现转折。第三个浪潮是行业内部才看到的：Claude Code，第一个非常有用的智能体系统，真正革命性的东西。但 Claude Code 只对企业开放，外界大多数人直到 Open Claw 出现才了解它。Open Claw 基本上把 AI 智能体可以做什么带入了大众意识。这就是为什么 Open Claw 从文化角度如此重要。

Internal consumption is thinking, which led to reasoning. 01 and continue that wave of ChatGPT grounded information, made AI not only answer questions, but answer questions in a more grounded way, useful. We started seeing the revenues and the economic model of OpenAI start to inflect. Then the third one was only inside the industry that we saw, Claude Code, the first agentic system that was very useful, really revolutionary stuff. But Claude Code was only available for enterprises. Most people outside never saw anything about Claude Code until Open Claw. Open Claw basically put into the popular consciousness what an AI agent can do. That's the reason why Open Claw is so important from a cultural perspective.

现在，它如此重要的第二个原因是 Open Claw 是开放的，但它构建了一种计算模型，基本上是在重新发明计算机。它有一个记忆系统，有一个短期记忆文件系统。它有技能，有规模。所以，首先它拥有资源并管理资源。它做调度。还有定时任务，它可以衍生出智能体，可以分解任务并在调度中解决问题。它有 IO 子系统，可以输入、输出、连接到 WhatsApp。它还有一个 API，可以运行多种类型的应用程序，称为技能。这四个要素从根本上定义了一台计算机。因此，我们得到了什么？我们第一次拥有了一台个人人工智能计算机。

Now, the second reason why it's so important is that Open Claw is open, but it formulates and structures a type of computing model that is basically reinventing the computer altogether. It has a memory system. It has a short-term memory file system. It has skills. It has scales. So, the first thing is it has resources and manages resources. It does scheduling. And cron jobs, it could spawn off agents, it could decompose a task and solve problems as it does scheduling. It has IO subsystems. It could input, output, connect to WhatsApp. Also, it has an API that allows it to run multiple types of applications called skills. These four elements fundamentally define a computer. And therefore, what do we have? We have a personal artificial intelligence computer for the very first time.

它是开源的，几乎可以在任何地方运行。所以，这基本上是现代计算的蓝图和操作系统。它将无处不在。当然，我们需要帮助它做到的一点是，每当你有智能体软件时，你必须确保智能体软件能访问敏感信息、能执行代码、能对外通信。我们必须确保所有这些都受到治理，都是安全的，并且我们有策略让这些智能体拥有三件事中的两件，但不能同时拥有全部三件。所以，治理部分我们贡献给了 Peter。Peter Steinberger 在这里，我们有一大群优秀的工程师与他合作，帮助保护这个东西，使其能够保护我们的隐私和安全。

It's open source. It runs literally everywhere. And so, this is basically the blueprint, the operating system of modern computing. And it's going to run literally everywhere. Now, of course, one of the things that we have to help it do is whenever you have agentic software, you have to make sure that an agentic software has access to sensitive information, it could execute code, it could communicate externally. We have to make sure that all of it has to be governed, all of it has to be secure, and that we have policies that give these agents two of the three things, but not all three things at the same time. And so, the governance part of it, we contributed to Peter. Peter Steinberger was here, and we've got a mound of great engineers working with him to help secure and keep that thing so that it could protect our privacy, protect our security.

Jensen，这种范式转变使得国内通过的一些 AI 监管立法以及许多拟议的立法实际上变得毫无意义，不是吗？你能简单评论一下，范式转变如何迅速使许多 AI 监管监督模型变得过时？这在政治上正成为一个非常热门的话题。

Jensen, that paradigm shift makes some of the AI legislation that has passed around the country to regulate AI, and a lot of the proposed legislation effectively moot, doesn't it? Can you just comment for a second on how quickly the paradigm shift kind of obviates a lot of the models for regulatory oversight of AI, which is becoming a very hot topic in politics right now?

嗯，这就是我们与政策制定者打交道时，需要始终走在他们前面，Brad，你在这方面做得很好。我们必须主动向他们介绍技术的现状，它是什么，不是什么。它不是生物体，不是外星人，没有意识。它是计算机软件。而且，我们并非完全不了解它。我们了解这项技术的很多方面。所以我认为，第一，我们必须确保继续向政策制定者提供信息，不让末日论和极端主义影响他们对这项技术的看法和理解。然而，我们仍然必须认识到这项技术发展非常快，不要让政策过于超前于技术。我们作为一个国家面临的风险，在 AI 方面最大的国家安全担忧，是其他国家采用这项技术，而我们却对它愤怒、恐惧或偏执，导致我们的产业和社会没有利用好 AI。所以我主要担心的是 AI 在美国的普及。

Well, this is the part where we just with policy makers, we need to always get in front of them, and Brad, you do a great job doing this. We have to get in front of them and inform them about the state of the technology, what it is, what it is not. It is not a biological being. It is not alien. It is not conscious. It is computer software. And it is not something that we say things like we don't understand it at all. It is not true we don't understand it all. We understand a lot of things about this technology. And so I think one, we have to make sure that we continue to inform the policy makers and not allow doomerism and extremism to affect how policy makers think and understand about this technology. However, we still have to recognize this technology is moving really fast and don't get policy ahead of the technology too quickly. And the risk that we run as a nation, our greatest source of national security concern with respect to AI, is that other countries adopt this technology while we are so angry at it or afraid of it or somehow paranoid of it that our industries, our society don't take advantage of AI. And so I'm just mostly worried about the diffusion of AI here in the United States.

你能详细谈谈吗？如果你坐在 Anthropic 的董事会里，面对与国防部的那些传闻，你会怎么做？这加剧了人们不知道该怎么想的状况，增加了人们对 AI 软件层面的怨恨、恐惧或普遍不信任。你认为你会告诉 Dario 和他的团队做些什么不同的事情，来改变一些结果和看法？

Can you just double click if you were in the seat in the boardroom of Anthropic over that whole scuttlebutt with the Department of War? It sort of builds on this idea of people didn't know what to think. It's sort of added to this layer of either resentment or fear or just general mistrust that people have sometimes at the software levels of AI. What you would do you think you would have told Dario and that team to do maybe differently to try to change some of this outcome and some of this perception?

关于 Anthropic，我首先要说的是，他们的技术令人难以置信。我们是 Anthropic 技术的大用户。非常钦佩他们对安全的关注，对安全性的重视。他们做事的方式，技术卓越性，都非常棒。我想说，他们希望警告人们技术能力的愿望也非常好。我们只需要确保理解世界是多样化的，警告是好的，但恐吓就不太好了。因为这项技术对我们太重要了。我认为预测未来是可以的，但我们需要更加谨慎。我们需要多一点谦逊，事实上我们无法完全预测未来。说出非常极端、灾难性的事情，却没有证据表明会发生，可能比人们想象的更具破坏性。当然，我们是技术领袖。曾经有一段时间没人听我们的。但现在，因为技术在社会结构、重要产业和国家安全中如此重要，我们的话确实有分量。我认为我们必须更加谨慎，更加温和，更加平衡，更加深思熟虑。

The first thing that I would say about Anthropic is first of all the technology is incredible. We are a large consumer of Anthropic technology. Really admire their focus on security. Really admire their focus on safety. The culture by which they went about it, the technology excellence by which they went about it, really fantastic. I would say that the desire to warn people about the capability of the technology is also really terrific. We just have to make sure that we understand that the world has a spectrum and that warning is good, scaring is less good. And because this technology is too important to us. And I think that it is fine to predict the future, but we need to be a little bit more circumspect. We need to have a little bit more humility that in fact we can't completely predict the future. And the ability to say things that are quite extreme, quite catastrophic, that there's no evidence of it happening, could be more damaging than people think. And of course we are technology leaders. There was a time when nobody listened to us. But now because technology is so important in the social fabric, such an important industry, so important to national security, our words do matter. And I think we have to be much more circumspect, we have to be more moderate, we have to be more balanced, we have to be more thoughtful.

嗯，我会提名你。我认为整个行业必须团结起来。AI 在美国的受欢迎程度只有 17%。我的意思是，我们看到了核能发生了什么，对吧？我们基本上关闭了整个核工业，现在中国在建 100 座裂变反应堆，而美国是零。我们听说数据中心暂停建设，所以我认为我们必须更加积极主动。

Well, I would nominate you. I think the industry's got to get together. 17% popularity of AI in the United States. I mean, we see what happened to nuclear, right? We basically shut down the entire nuclear industry and now we have 100 fission reactors being built in China and zero in the United States. We hear about moratoriums on data centers, so I think we have to be a lot more proactive about that.

但我想回到你公司内部看到的智能体式爆发，效率和生产力提升。关于是否看到投资回报有很多争论。今年年初，大问题是：收入会出现吗？会像智能一样规模化吗？然后二月份 Anthropic 出现了每月 50-60 亿美元的奥本海默时刻。当你看到 Anthropic 和 OpenAI 发生这种情况时，你认为我们现在是否正处于收入像智能一样规模化的曲线上？

But I want to go back to this agentic explosion that you're seeing inside your company, the efficiencies, the productivity gains. There's a lot of debate whether we're seeing ROI. Entering this year, the big question was: are the revenues going to show up? Are they going to scale like intelligence? Then we had this Oppenheimer moment of $5-6 billion month by Anthropic in February. When you see this happening at Anthropic and OpenAI, do you think we're on that curve now where revenues scale the way intelligence is scaling?

当你环顾这些观众，你会看到 Anthropic 和 OpenAI 在这里有代表，但实际上，这里 99% 的东西都是 AI，而不是 Anthropic 和 OpenAI。

When you look around this audience, you will see that Anthropic and OpenAI are represented here, but in fact, 99% of everything here is all AI and it's not Anthropic and OpenAI.

对。对。

Right. Right.

原因在于 AI 非常多样化。第二受欢迎的模型类别是开放模型。第一是开源。OpenAI 是第一，开源是第二。非常遥远的第三是 Anthropic，这说明了这里所有 AI 公司的规模。

The reason for that is because AI is very diverse. The second most popular model as a category is open models. Number one is open source. OpenAI is number one, open source is number two. Very distant third is Anthropic, and that tells you something about the scale of all the AI companies here.

让我回来说几点。第一，当我们从生成式转向推理时，所需的算力增加了大约一百倍。当我们从推理转向智能体式时，算力可能又增加了一百倍。现在，仅仅两年内，算力就增加了一万倍。与此同时，人们为信息付费，但人们主要是为工作付费。

Let me come back and say a couple things. One, when we went from generative to reasoning, the amount of computation we needed was about a hundred times. When we went from reasoning to agentic, the computation is probably another hundred times. Now we're looking at in just two years computation went up by a factor of 10,000x. Meanwhile, people pay for information, but people mostly pay for work.

是的。

Yes.

和聊天机器人对话并得到答案非常棒。帮我做些研究，难以置信。但完成工作，我会付费。

Talking to a chatbot and getting an answer is super great. Helping me do some research, unbelievable. But getting work done, I'll pay for.

确实。

Indeed.

这就是我们现在的情况。智能体式系统完成工作。它们帮助我们的软件工程师完成工作。所以你看：算力增加了一万倍，现在消费可能增加了一百倍。我们甚至还没有开始规模化。我们绝对处于一百万倍的水平。

And so that's where we are. Agentic systems get work done. They're helping our software engineers get work done. So then you take that: you got 10,000x more compute, you get probably at this point 100x more consumption now. And we haven't even started scaling yet. We are absolutely at a million x.

这正好可以谈谈员工数量。你们公司有 2 万、3 万人？

Which is a great place to talk about the number of employees. You have 20,000, 30,000 at the company?

我们有 43,000 名员工。我会说 38,000 人是工程师。

We have 43,000 employees. I would say 38,000 are engineers.

我们在播客中多次讨论过：看看我们公司的 token 使用量。它在大幅增长。有些人问：‘当我加入一家公司时，我能得到多少 token，因为我想成为一名高效的员工？’你在两个半小时的主题演讲中假设你花费了……

Conversation we've had on the pod a number of times is: look at the token usage in our companies. It is growing massively. Some people are asking, 'When I join a company, how many tokens do I get because I want to be an effective employee?' And you postulated during your two and a half hour keynote that you were spending...

干得好，但应该更短。

Well done, it would be shorter.

是的。你没有时间做……

Yeah. You didn't have time to do a...

所以你们知道没有排练，所以就是即兴发挥。

So you guys know there is no practice, and so it's a grip and a rip.

是的。是的。

Yeah. Yeah.

我只是想让你知道，我在演讲的同时也在写演讲稿。

I just wanted to let you know I was writing the speech while I was giving the speech.

但这是否意味着如果我们粗略计算，每个工程师大约 75,000 个 token？那么你现在在英伟达为你的工程团队花费了 10 亿、20 亿美元在 token 上？

But does that mean if we do back-of-the-envelope math, 75,000 tokens for each engineer something like that? So are you spending in Nvidia a billion, two billion dollars on tokens for your engineering team right now?

我们正在努力。让我给你一个思想实验。假设你有一个软件工程师或 AI 研究员，你每年付给他们 50 万美元。我们经常这样做。年底我会问那个 50 万美元的工程师他用了多少 token，花了多少钱在 token 上，如果那个人说 5,000 美元，我会发疯的。如果那个 50 万美元的工程师没有消耗至少 25 万美元的 token，我会非常震惊。这和我们的一位芯片设计师说‘猜怎么着？我打算只用纸和笔。我不认为我需要任何 CAD 工具’没什么不同。

We're trying to. Let me give you a thought experiment. Let's say you have a software engineer or AI researcher and you pay them $500,000 a year. We do that all the time. That $500,000 engineer at the end of the year I'm going to ask him how many tokens, how much did you spend in tokens, and that person said $5,000 I will go ape something else. If that $500,000 engineer did not consume at least $250,000 worth of tokens, I am going to be deeply alarmed. This is no different than one of our chip designers who says, 'Guess what? I'm just going to use paper and pencil. I don't think I'm going to need any CAD tools.'

这是思考这些全明星员工的一个真正的范式转变。这让我想起我们在 NBA 中学到的，当勒布朗·詹姆斯开始每年花费一百万美元来保持身体健康。他现在这个年龄还在打球。这真的是，‘嘿，如果这些是不可思议的知识工作者，为什么我们不给他们超人的能力呢？’

And this is a real paradigm shift in thinking about these all-star employees. It reminds me of what we learned in the NBA when LeBron James started spending a million dollars a year just on his health and maintaining his body. Here he is at age still playing. It really is, 'Hey, if these are incredible knowledge workers, why wouldn't we give them superhuman abilities?'

这正是……

That's exactly what...

如果我们从现在开始推断两三年，那个全明星员工在英伟达的效率以及他们能完成什么？

Where does that go if we extrapolate out two or three years from now, what is the efficiency of that all-star at an Nvidia and what they're able to accomplish?

他们会看起来……首先，‘太难了’——那个想法消失了。‘这需要很长时间’——那个想法消失了。‘我们需要很多人’——那个想法消失了。这和上次工业革命没什么不同，有人说，‘天哪，那栋建筑看起来很重。’没人这么说。‘哇，那座山看起来太大了。’没人这么说。所有太大、太重、太耗时的事情——那些想法都消失了。

They're going to look like... Well, first of all, things that are 'too hard' — that thought is gone. 'This is going to take a long time' — that thought is gone. 'We're going to need a lot of people' — that thought is gone. This is no different than in the last industrial revolution, somebody goes, 'Boy, that building really looks heavy.' Nobody says that. 'Wow, that mountain looks too big.' Nobody says that. Everything that's too big, too heavy, takes too long — those thoughts are all gone.

转向创造力。

To creativity.

没错。

That's right.

你能想出什么？

What can you come up with?

正是。

Exactly.

这意味着现在的问题是，你如何与这些智能体合作？

Which means now the question is, how do you work with these agents?

嗯，这只是计算机编程的一种新方式。过去我们编码，未来我们将编写想法、架构、规范。我们将组织团队。我们将帮助他们定义如何评估好与坏的标准。什么是一个好结果的样子？如何与你迭代，如何头脑风暴。这真的是你想要的。我认为每个工程师都将拥有一百个智能体。

Well, it's just a new way of doing computer programming. In the past we code, in the future we're going to write ideas, architectures, specifications. We're going to organize teams. We're going to help them define how to evaluate the definition of good versus bad. What does it look like when something is a great outcome? How to iterate with you, how to brainstorm. That's really what you're looking for. I think that every engineer is going to have a hundred agents.

回到行业目前面临的公关问题，像 Ohalo 的 David Freiberg 这样的高管，正在通过使用技术、你的技术和 AI，来增加卡路里产量并制造高质量卡路里。你认为你能把成本降低多少倍，Freiberg，这个愿景对你正在做的事情有什么影响？

Back to the PR problem the industry has right now, you have executives like David Freiberg with Ohalo, who's looking at literally taking through the use of technology, your technology and AI, the number of calories produced and making high-quality calories. What is the factor you think you can bring the cost down, Freiberg, and what impact does this vision have for what you're doing?

零样本基因组建模，而且它有效。

Zero-shot genomic modeling, and it works.

是的。

Yeah.

然后你有了那个时刻，你会想，‘天哪——’老实说，就像人们在晚上替换整个企业软件堆栈之后。

Then you have that moment and you're like, 'Holy sh—' Honestly, like and that's after people are replacing entire enterprise software stacks in a night.

我 90 分钟里做了件事，跟伙计们说了。替换了整个软件栈和一大堆工作负载。用 Claude 花了 90 分钟，运行了这个智能体系统，构建了整个东西，部署了它，在周日晚上。晚上 10 点我就搞定了。作为 CEO，你替换了整个软件栈？是的。我的管理团队每个人都得做类似的练习。周末过后，周一我们看到的情况，我当时就想，‘完了。’

I did something in 90 minutes I was telling the guys about. Replaced the whole software stack and like a whole bunch of workload. 90 minutes on Claude, ran this agentic system, built the whole thing, deployed it, on a Sunday night. 10:00 p.m. I was done. As the CEO, you replaced the whole software stack? Yeah. And everyone on my management team had to do a similar exercise. Over the weekend, what we saw on Monday, I was like, 'It's over.'

但是，技术方面、科学方面，我们用 auto research 在 30 分钟内做了一件事，我想听听你对 auto research 的看法，以及这告诉我们效率方面还有多远的路要走。但，用 auto research 和一大块数据，内部发表了一些东西，我们都说，‘哦，天哪。’这通常是一篇需要 7 年的博士论文。它会是我们在这个领域见过的最著名的博士论文之一，会发表在《科学》杂志上，而它是在 30 分钟内完成的，在一台台式电脑上运行 auto research，用我们刚摄入的所有数据。我们周五拿到它，然后说，‘嘿，试试看。’启动它，上 GitHub，下载 auto research，然后运行。你看到每个人的表情都像……然后这为我们解锁的潜力，就像那种需要 7 年的事情，却在 30 分钟内发生了。我们在基因组学中体验到了，我们觉得，‘这太不可思议了。’所以，我认为这种加速正在以你几年前无法想象的方式拓宽每个人的视野。但回到 auto research 这一点，你能评论一下吗？这个东西在一个周末内用 600 行代码发布，并且能够在本地运行，用所有这些多样化的数据集实现它所实现的，这告诉我们算法和硬件优化还处于早期阶段。

But, the technical stuff, the science stuff, we did something in 30 minutes using auto research, and I'd love your view on auto research and what that tells us about how far we still have to go in terms of efficiency. But, using auto research and a chunk of data, something was published internally that we said, 'Oh my god.' And that would normally be a PhD thesis that would take 7 years. It would be one of the most celebrated PhD theses we've ever seen in this field, and it would be in the journal Science, and it was done in 30 minutes on a desktop computer running on auto research with all the data we just ingested. We got it on Friday, and we're like, 'Hey, let's try it.' Boot it up, went to GitHub, downloaded auto research, and ran it. And you see everyone's face just go like... And then the potential of what this is unlocking for us is like the kind of thing that would take 7 years, and it happened in 30 minutes. And we're experiencing it in genomics, and we're like, 'This is unbelievable.' So, I think the acceleration is widening the aperture for everyone in a way that you didn't imagine a few years ago. But just going back to the auto research point, can you just comment on what you think about the fact that this thing got published with 600 lines of code in a weekend and the capacity that it has to run locally and achieve what it can achieve with all of these diverse data sets and what that tells us about the early stages we are in terms of optimization on algorithms and hardware.

Open Claw 如此不可思议的根本原因，第一，是它的汇合，它与大语言模型突破的时机。时机完美，无可挑剔。现在，在很多方面，如果不是因为 Claude、GPT 和 ChatGPT 已经达到了非常好的水平，Peter 可能不会想出它。这也是一种新能力，允许这些模型使用工具。我们长期以来创造的工具，比如网页浏览器、Excel 电子表格，在芯片设计领域，Synopsys、Cadence、Omniverse、Blender、Autodesk，所有这些工具都将继续被使用。有些人说企业 IT 软件行业会被摧毁。让我给你另一种观点。企业软件行业受限于座位数。它即将迎来 100 倍更多的智能体来敲打这些工具。会有智能体敲打 SQL，敲打向量数据库，敲打 Blender，敲打 Photoshop。原因在于，首先，这些工具做得很好。其次，这些工具是我们之间的管道。归根结底，当工作完成时，它必须以我能控制的方式呈现给我。我知道如何控制这些工具。所以，我需要所有东西都放回 Synopsys。我希望所有东西都放回 Cadence，因为那是我控制的方式。那是我验证事实的方式。

The fundamental reason why Open Claw is so incredible, number one, is it's confluence, it's timing with the breakthroughs in large language model. Its timing was perfect. It was impeccable. Now, in a lot of ways, Peter wouldn't have come up with it probably if not for the fact that Claude and GPT and ChatGPT have reached a level that is really very good. It is also a new capability that allows these models to tool use. The tools that we've created over time, web browsers and Excel spreadsheets and, in the case of chip design, Synopsys and Cadence and Omniverse and Blender and Autodesk and all of these tools are going to continue to be used. Some people say that the enterprise IT software industry is going to get destroyed. Let me give you the alternative view. The enterprise software industry is limited by butts and seats. It's about to get 100 times more agents banging on those tools. They're going to be agents banging on SQL, they're going to be agents banging on vector databases, agents banging on Blender, agents banging on Photoshop. And the reason for that is because those tools are, first of all, do a very good job. Second, those tools are the conduit between us. In the final analysis, when the work is done, it has to be represented back to me in a way that I can control. And I know how to control those tools. And so, I need everything to be put back into Synopsys. I want everything to be put back into Cadence, because that's how I control it. That's how I ground truth.

让我问你一个关于开源的问题。我们有这些闭源模型，它们很优秀。我们有这些开放权重模型，很多中国模型令人难以置信。绝对令人难以置信。两天前，你可能没看到，因为你忙着上台，但有一个训练运行发生在名为 BitTensor subnet 3 的加密项目中。他们设法训练了一个 40 亿参数的 Llama 模型，完全分布式，由一群人贡献多余算力，但他们能够有状态地管理训练运行，我认为这是一个相当疯狂的技术成就。我们现代版的 folding@home。那么，你对开源的最终状态怎么看？你是否也看到了架构的去中心化和算力的去中心化，以支持开放权重和完全开源的方法，确保 AI 广泛可用？

Let me ask you a question about open source. So, we have these closed source models, they're excellent. We have these open weight models, many of the Chinese models are incredible. Absolutely incredible. Two days ago, you may not have seen this because you were busy on stage, but there was a training run that happened in this crypto project called BitTensor subnet 3. They managed to train a 4 billion parameter llama model totally distributed with a bunch of people contributing excess compute, but they were able to do it statefully and manage a training run, which I thought was like a pretty crazy technical accomplishment. Our modern version of folding at home. So, what do you think about the end state of open source? Do you see this decentralization of architecture as well and decentralization of compute to support open weights and a totally open source approach to making sure AI is broadly available to everyone?

我相信我们从根本上需要模型作为一流产品、专有产品，以及模型作为开源。这两件事不是 A 或 B，而是 A 和 B。毫无疑问。原因在于模型是一种技术，而不是产品。模型是一种技术，而不是服务。对于绝大多数消费者来说，水平层、通用智能，我真的非常不想自己去微调。我真的想继续使用 ChatGPT。我喜欢用 Claude。我喜欢用 Gemini。我喜欢用 X。它们都有各自的个性，你知道，这取决于我的心情和我想解决的问题。你知道，我可能会在 X 上做，或者在 ChatGPT 上做。所以，行业的这个部分正在蓬勃发展。它会很棒。然而，所有这些行业，它们的领域专长、它们的专业化必须被引导，必须以它们能控制的方式被捕捉。而这只能来自开放模型。我们正在为开放模型行业做出巨大贡献。它接近前沿。坦率地说，即使它达到前沿，我认为产品即服务、世界级产品即模型将继续蓬勃发展。

I believe we fundamentally need models as a first-class product, proprietary product, as well as models as open source. These two things are not A or B, it's A and B. There's no question about it. And the reason for that is because models is a technology, not a product. Models a technology, not a service. For the vast majority of consumers, the horizontal layer, the general intelligence, I would really, really love not to go fine-tune my own. I would really love to keep using ChatGPT. I love to use Claude. I love to use Gemini. I love to use X. And they all have their own personalities, as you know, which just kind of depends on my mood and depends on what problem I'm trying to solve. You know, I might do it on X or I might do it on ChatGPT. And so, that segment of the industry is thriving. It's going to be great. However, there are all these industries, their domain expertise, their specialization has to be channeled, has to be captured in a way that they can control. And that can only come from open models. The open model industry we're contributing tremendously to. It is near the frontier. And quite frankly, even if it reaches the frontier, I think that products as a service, world-class products as models as a product is going to continue to thrive.

我们现在投资的每个初创公司都是先开源，然后再转向专有模型。

Every startup we're investing in now is open source first and then going to the proprietary model.

是的，美妙之处在于，因为你有一个很好的路由器，你在第一天就能连接两者，每一天，你都能访问世界上最好的模型。然后它给你时间降低成本、微调和专业化。所以你将每次都能获得世界级的能力。

Yeah, and the beautiful thing is because you have a great router, you connect the two by on first day, every single day, you're going to have access to the world's best model. And then it gives you time to cost reduce and fine-tune and specialize. And so you're going to have world-class capabilities out to shoot every single time.

Jensen，我能问个问题吗？没有人比你更希望美国赢得全球 AI 竞赛。对吧？但一年前，拜登时代的扩散规则实际上是反美国的 AI 全球扩散。那么，新政府上任一年了，给我们打个分吧。

Jensen, can I ask a question? Nobody wants the US to win the global AI race more than you. Right? But a year ago, the Biden era diffusion rule really was an anti-American diffusion of AI around the world. So here we are a year into the new administration. Give us a grade.

在全球扩散以及我们向世界传播美国 AI 技术的速度方面，我们处于什么位置？我们是 A 级？B 级？还是 C 级？哪些有效？哪些无效？

Where are we in terms of global diffusion and the rate at which we're spreading US AI technology around the world? Are we an A? Are we a B? Are we a C? What's working? What's not working?

首先，特朗普总统希望美国工业领先。他希望美国科技产业领先。他希望美国科技产业获胜。他希望我们把美国技术传播到世界各地。他希望美国成为世界上最富有的国家。他希望这一切。目前，就在我们说话的时候，英伟达在全球第二大市场放弃了 95% 的市场份额，我们现在是 0%。特朗普总统希望我们重新进入那里，第一件事就是为我们将要销售的公司获得许可证。我们有很多公司已经申请了许可证。我们为他们申请了许可证，并且已经从卢特尼克部长那里获得了批准的许可证。现在我们已经通知了中国公司，其中许多公司已经给了我们采购订单。所以我们正在重新启动供应链以进行发货。我认为在最高层面上，布拉德，我们应该承认这一点。当我们无法获得微型电机、稀土矿物时，我们的国家安全就会削弱。当我们无法控制我们的电信网络时，它就会削弱。当我们无法为我国提供可持续能源时，它就会削弱。从根本上说，它被削弱了。这些行业中的每一个都是我不希望 AI 行业变成的样子的例子。

Well, first of all, President Trump wants American industry to lead. He wants American technology industry to lead. He wants American technology industry to win. He wants us to spread American technology around the world. He wants the United States to be the wealthiest country in the world. He wants all of that. At the current moment, as we speak, Nvidia gave up a 95% market share in the second largest market in the world and we're at 0%. President Trump wants us to get back in there and the first thing is to get licenses for the companies that we're going to be able to sell to. We've got many companies who have requested for licenses. We've applied for licenses for them and we've got approved licenses from Secretary Lutnick. Now we've informed the Chinese companies and many of them have given us purchase orders. And so we're in the process of cranking up our supply chain again to go ship. I think at the highest level, Brad, one of the things that we should acknowledge is this. Our national security is diminished when we don't have access to miniature motors, rare earth minerals. It's diminished when we don't control our telecommunications networks. It's diminished when we can't provide for sustainable energy for our country. It is fundamentally diminished. Every single one of these industries is an example of what I don't want the AI industry to be.

对。

Right.

当我们展望未来，问自己‘我们想要什么？当美国科技产业、美国 AI 产业引领世界时，那会是什么样子？’我们都可以承认，AI 模型不可能被普遍采用。那是一个毫无意义的结果。然而，我们都可以想象，从芯片到计算系统再到平台的美国技术栈被世界广泛使用，他们构建自己的 AI，使用公共 AI，使用私有 AI，等等，他们可以在自己的社会中构建应用程序。我希望美国技术栈占世界的 90%。是的，我希望如此。另一种情况，如果它像太阳能、稀土、磁铁、电机、电信那样，我认为这对国家安全来说是一个非常糟糕的结果。

When we look forward in time and we say, 'What do we want? What does it look like when American technology industry, American AI industry leads the world?' We can all acknowledge that there is no way that AI models is won universally. It is an outcome that makes no sense. However, we can all imagine that the American tech stack from chips to computing systems to the platforms are used broadly by the world where they build their own AI, they use public AI, they use private AI, whatever, and they can build their applications in their society. I would love that the American tech stack is 90% of the world. Yes. I would love that. The alternative, if it looks like solar, rare earth, magnets, motors, telecommunications, I consider that a very bad outcome for national security.

同意。

Agreed.

是的。

Yeah.

你目前对全球冲突局势的关注程度如何？这些冲突让你有多担心，黄仁勋？中国和台湾，以及我理解的中东氦气供应，可能成为半导体制造的供应链风险。这些情况让你有多担心？你在这方面投入了多少精力？

How much are you monitoring the situation with the conflicts around the world right now, and how much does it worry you, Jensen? So, China and Taiwan, and then helium availability coming out of the Middle East, I understand, can be a supply chain risk to semiconductor manufacturing. How much do these situations worry you? How much are you spending on them?

首先，我认为在中东，我有 6000 个家庭在那里。英伟达有很多伊朗人，他们的家人还在伊朗。所以我们在那里有很多家庭。首先，他们非常焦虑，非常担心，非常害怕。我们一直在想着他们。我们一直在监控和关注他们。他们得到了我们 100% 的支持。有人多次问我，我们是否还在考虑留在以色列？我们 100% 在以色列。我们 100% 支持那里的家庭。我们 100% 在中东。也有人问我，鉴于中东正在发生的事情，我们是否认为那里是可以扩展人工智能的领域？我相信我们开战是有原因的，我相信战争结束后，中东会比以前更稳定。所以，如果我们以前考虑过，那么之后绝对应该考虑。所以，我 100% 支持。关于台湾，我们必须做三件事。第一，我们必须确保尽快实现美国的再工业化。

Well, first of all, I think in the Middle East, I have 6,000 families there. We have a lot of Iranians at Nvidia, and their families are still in Iran. So we have a lot of families there. The first thing is they are quite anxious, they're quite concerned, quite scared. We're thinking about them all the time. We're monitoring and keeping an eye on them all the time. They have 100% of our support. I've been asked several times, are we still considering being in Israel? We are 100% in Israel. We are 100% behind the families there. We are 100% in the Middle East. I was also asked, given what's happening in the Middle East, is that an area where we believe that we can expand artificial intelligence to? I believe that there's a reason we went to war, and I believe that at the end of the war, the Middle East will be more stable than before. And so, if we were considering it before, we should absolutely be considering it after. And so, I'm 100% in on that. With respect to Taiwan, we have to do three things. One, we have to make sure that we re-industrialize the United States as fast as we can.

是的。

Yeah.

无论是芯片制造厂、计算机制造厂还是 AI 工厂，我们都做得很好。通过获得战略支持，通过赢得台湾供应链的友谊，通过赢得他们的友谊和支持，我们能够以惊人的速度建设亚利桑那州、德克萨斯州和加利福尼亚州。他们确实是战略合作伙伴。他们值得我们的支持，值得我们的友谊，值得我们的慷慨。他们正在尽一切努力为我们加速制造过程。所以我认为这是第一点。第二，我们应该使制造供应链多样化。无论是韩国、日本还是欧洲，我们都应该使供应链多样化，使其更具韧性。第三，让我们表现出克制。在我们增加多样性和韧性的同时，我们不要不必要地施压。所以我们需要耐心。

And whether it's the chip manufacturing plants, the computer manufacturing plants, or the AI factories, we're doing excellent. By gaining the strategic support, by gaining the friendship of the supply chain of Taiwan, by gaining their friendship, by gaining their support, we were able to build Arizona and Texas, California at incredible rates. They are genuinely a strategic partner. They deserve our support, they deserve our friendship, they deserve our generosity. And they're doing everything they can to accelerate the manufacturing process for us. So I think that's number one. Number two, we ought to diversify the manufacturing supply chain. And whether it's South Korea, Japan, or Europe, we ought to diversify the supply chain, make it more resilient. And number three, let's demonstrate restraint. And while we're increasing our diversity and resilience, let's not push unnecessarily. So we need to be patient.

考虑得很周到。氦气是个问题吗？有很多报道。

It's thoughtful. Is helium a problem? A lot of reports out there.

我认为氦气可能是个问题，但供应链中可能有很多缓冲。这类事情往往有很多缓冲。但你知道。

I think helium could be a problem, but it's also the case that the supply chain probably has a lot of buffer in it. These kind of things tend to have a lot of buffer. But you know.

是的。

Yeah.

你在自动驾驶方面取得了巨大进展。做了一个重大宣布。你增加了更多合作伙伴，包括比亚迪。刚刚有一段视频显示你驾驶一辆奔驰车。还有与 Uber 的重大宣布，你将让来自许多不同制造商的汽车上路。我相信你的赌注是，将会有一个类似 Android 的开源平台，你将在其中扮演重要角色，与数十家汽车供应商合作。然后另一边可能有一个类似 iOS 的平台，由特斯拉或 Waymo 主导。你的战略思考是什么？这个棋局如何展开？因为你感觉拥有相当深的栈，在某些方面你在竞争，在其他方面你在合作。

You've made massive progress in self-driving. Made a big announcement. You've added many more partners including BYD. There was just a video of you driving around in a Mercedes. And huge announcement with Uber that you're going to have a number of cars on the road from many different manufacturers. Your bet is I believe that there's going to be an Android type open-source platform that you're going to play a major part in with dozens of car providers. And then maybe on the other side there could be an iOS with Tesla or Waymo. What's your strategy thinking there and how that chessboard emerges because it feels like you have a pretty deep stack, and in some ways you're competing, and in other places you're collaborative.

是的。退一步说，我们相信所有移动的东西有一天都会完全或部分自主。第一。第二，我们不想制造自动驾驶汽车，但我们希望让世界上每一家汽车公司都能制造自动驾驶汽车。所以我们构建了所有三台计算机：训练计算机、仿真计算机、评估计算机，以及车载计算机。我们开发了世界上最安全的驾驶操作系统。

Yeah. Taking a step back, we believe that everything that moves will be autonomous completely or partly someday. Number one. Number two, we don't want to build self-driving cars, but we want to enable every car company in the world to build self-driving cars. And so we built all three computers: the training computer, the simulation computer, the evaluation computer, as well as the car computer. We develop the world's safest driving operating system.

我们还创造了世界上第一辆推理型自动驾驶汽车，它能够将复杂场景分解成更简单的场景，就像我们一样，通过推理系统来导航。这个名为 Alpaca IO 的推理系统让我们取得了令人难以置信的成果。我们开放它，纵向优化，横向创新，让每个人自己决定。你是想从我们这里买一台计算机？以埃隆和特斯拉为例，他们购买我们的训练计算机。他们是想买我们的训练计算机和模拟计算机？还是想与我们合作，三者都做，甚至把车载计算机装到你的车里？我们的态度是，我们想解决问题。我们不是解决方案提供商。无论你如何与我们合作，我们都很高兴。

We also created the world's first reasoning autonomous vehicle, so it could decompose complicated scenarios into simpler scenarios that it knows how to navigate through, just like us, reasoning systems. That reasoning system, called Alpaca IO, has enabled us to achieve incredible results. We open this, we vertically optimize, we horizontally innovate, and we let everybody decide. Do you want to buy one computer from us? In the case of Elon and Tesla, they buy our training computers. Do they want to buy our training computer and our simulation computers? Or do you want to work with us to do all three and even put the car computer in your car? Our attitude is we want to solve the problem. We're not the solution provider. We're delighted however you work with us.

基于这个问题，我觉得这很有意思。你确实创建了这个平台，百花齐放。但同样，有些花现在想回到技术栈底层，试图与你竞争。谷歌有 TPU，亚马逊有 Inferentia 和 Trainium。每个人都在推出自己的版本，认为‘我能超越英伟达’。尽管他们往往也是大客户。你如何应对？你认为随着时间的推移会发生什么？这些因素在这种视频的格局中扮演什么角色？

To build on this question, because I think it's fascinating. You actually do create this platform. A thousand flowers are blooming. But it's also true that some of those flowers want to now go back down in the stack and try to compete with you a little bit. Google has TPU, Amazon has Inferentia and Trainium. Everybody's sort of spinning up their own version of 'I think I can out-Nvidia Nvidia.' Even though they also tend to be huge customers. How do you navigate that? And what do you think happens over time and where do those things play in the complexion of this kind of video?

首先，我们是唯一的人工智能公司。我们是一家 AI 公司。我们构建基础模型。我们在许多不同领域处于前沿。我们构建每一层、每一个技术栈。我们是世界上唯一一家与世界上所有 AI 公司合作的公司。他们从不向我展示他们在构建什么，而我总是向他们展示我正在构建的东西。

First of all, we're the only AI company. We're an AI company. We build foundation models. We're at the frontier of many different domains. We build every single layer, every single stack. We're the only AI company in the world that works with every AI company in the world. They never show me what they're building, and I always show them exactly what I'm building.

对。

Right.

信心来自这里。第一，我们乐于在最佳技术上竞争。只要我们能够继续快速前进，我相信从英伟达购买仍然是他们能做的最经济的事情之一。我看到了这种难以置信的信心。第二，我们是唯一能够存在于每个云中的架构，这给了我们一些根本性的优势。我们是唯一一种你可以从云中取出，放到本地、汽车、任何地区的架构。

The confidence comes from this. One, we are delighted to compete on what is the best technology. To the extent that we can continue to run fast, I believe that buying from Nvidia is still one of the most economic things they could do. I see this incredible confidence there. Number two, we're the only architecture that could be in every cloud, and that gives us some fundamental advantages. We're the only architecture you could take from a cloud and put into on-prem, in the car, in any region.

没错，在太空中。

That's right, in space.

我们市场的一部分，大约 40% 的业务，大多数人没有意识到这一点。40% 的业务，除非你拥有 CUDA 技术栈，除非你能构建一个完整的 AI 工厂，否则客户不知道拿你怎么办。他们不是要制造芯片，也不是要购买芯片，他们是要构建 AI 基础设施。他们希望你带着完整的技术栈进来，而我们拥有整个技术栈。令人惊讶的是，英伟达正在获得市场份额。看看我们今天的位置，我们正在增加份额。

There's a whole part of our market, about 40% of our business, most people don't realize this. 40% of our business, unless you have the CUDA stack, unless you can build an entire AI factory, the customers don't know what to do with you. They're not trying to build chips, they're not trying to buy chips, they're trying to build AI infrastructure. They want you to come in with a full stack, and we've got the whole stack. Surprisingly, Nvidia's gaining market share. If you look at where we are today, we're gaining share.

实际情况是，这些人尝试了，然后意识到‘天哪，这太难了。’然后他们又回来了。这就是他们份额增长的原因吗？

What happens is these guys try and they realize, 'Oh my god, it's too much.' And then they come back. Is that why their share grows?

我们增加份额有几个原因。第一，我们的速度已经提升。我们帮助人们意识到，关键不是制造芯片，而是构建系统。那个系统真的很难构建。他们与我们的业务在增长。以 AWS 为例，我想他们昨天刚刚宣布，未来几年将购买一百万颗芯片。那是来自 AWS 的大量芯片，而且是在他们已经购买的所有芯片之上。我们很高兴这样做。但第一，过去几年我们份额增长是因为现在 Anthropic 来到了英伟达。Meta SL 也来到了英伟达。开放模型的增长令人难以置信，而且这一切都基于英伟达。我们因模型数量而增长份额。我们还因为所有这些公司都在云之外，并且在区域、企业、行业、边缘增长。如果只是构建 ASIC，那整个增长部分真的很难实现。

We're gaining share for several reasons. One, our velocity has gone. We help people realize it's not about building the chip, it's about building the system. That system's really hard to build. Their business with us is increasing. In the case of AWS, I think they just announced yesterday that they're going to buy a million chips in the next couple years. That's a lot of chips from AWS, and that's on top of all the chips they've already bought. We're delighted to do that. But number one, we're gaining share this last couple years because we now have Anthropic coming to Nvidia. Meta SL is coming to Nvidia. The growth of open models is incredible, and that's all on Nvidia. We're growing in share because of the number of models. We're also growing in share because all of these companies are outside the cloud and they're growing regionally, in enterprise, in industries, at the edge. That entire segment of growth is really hard to do if it's just building an ASIC.

布拉德，与此相关，不深入数字细节，但分析师似乎不相信，对吧？如果你看共识预测，你说算力可能增长一百万倍，对吧？但他们预测你明年增长 30%，后年 20%，而 2029 年，本应是爆发的一年，只有 7%。对吧？所以如果你拿你的 TAM 应用他们的增长数字，那意味着你的份额将暴跌。你在未来的订单簿中看到什么能让这成为现实吗？

Brad, related to that, and not to get in the weeds on the numbers, but analysts don't seem to believe, right? So if you look at the consensus forecast, you said compute could 1 million X, right? And yet they have you growing next year at 30%, the year after that at 20%, and in 2029, which is supposed to be a monster year, at 7%. Right? So if you take your TAM and you apply their growth numbers, it suggests that your share will plummet. Do you see anything in your future order book that would make that correct?

首先，他们根本不了解 AI 的规模和广度。

First of all, they just don't understand the scale and the breadth of AI.

是的。对。

Yes. Yeah.

我认为这是真的。

I think that's true.

大多数人认为 AI 只存在于前五大超大规模云服务商中。

Most people think that AI is in the top five hyperscalers.

没错。

That's right.

还有关于大数定律的正统观念，他们必须回到他们的投行风险委员会，展示一些模型。他们心里不会相信 5 万亿会变成 15 万亿。他们会说，可能到 7 万亿。

There's also an orthodoxy around these law of large numbers where they have to go back to their investment banking risk committee and show some model. They're not going to believe in their minds that 5 trillion goes to 15 trillion. They're like it can go to seven.

或者他们需要一家 10 万亿美元的公司。

Or they need to have a 10 trillion-dollar company.

这都只是自保的把戏，我认为他们……以前发生过，所以你不能说它会发生。

It's all just CYA stuff that I think they're... it happened before, so you can't say it will.

因为你必须重新定义你所做的事情。最近有人评论说，英伟达，黄仁勋，你怎么能在服务器领域比英特尔还大？原因在于整个数据中心的 CPU 市场每年大约 250 亿美元。

Because you have to redefine what it is that you do. There was somebody who made an observation recently that Nvidia, Jensen, how can you be larger than Intel in servers? The reason for that is because the CPU market of the entire data center was about 25 billion dollars a year.

对。

Right.

我们每年做 250 亿美元，正如你们所知，就在我们坐在这里的这段时间里。显然那是个玩笑。

We do 25 billion dollars a year, as you guys know, in a very... in the time that we were sitting here. Obviously that was a joke.

不，但这是……全部在播客里。

No, it's but it's... all in podcast.

不，那不是指引。但无论如何，关键在于你能做多大取决于你制造什么。

No, that was not guidance. But anyhow, the point is how big you can be depends on what it is that you make.

对。

Right.

英伟达不是制造芯片的。第一，制造芯片不再能帮助你解决 AI 基础设施问题。它太复杂了。第三，大多数人认为 AI 狭隘地存在于他们谈论、听到和看到的事物中。

Nvidia is not making chips. Number one, making chips does not help you solve the AI infrastructure problem anymore. It's too complicated. Number three, most people think that AI is narrowly in the things that they talk about and hear and see.

对。

Right.

AI 要大得多。OpenAI 令人难以置信。它们会变得巨大。Anthropic 令人难以置信。它们会变得巨大。但 AI 将比那大得多得多。

AI is much bigger. OpenAI is incredible. They're going to be enormous. Anthropic is incredible. They're going to be enormous. But AI is going to be much, much bigger than that.

是的。

Yeah.

而我们服务于那个领域。

And we address that segment.

跟我们说说太空中的数据中心吧。

Tell us about data centers in space for a second.

我们已经进入太空了。

We're already in space.

外行人应该怎么理解这个业务，跟听到那些大型数据中心在地面建设时相比？

How should the layman think about what that business is versus when you hear about these big data center build-outs that's happening on the ground?

嗯，我们应该先做好地面上的事，因为我们已经在做了。第一，第二，我们应该准备进入太空，显然太空里有很多能源。挑战当然在于冷却，你不能利用传导和对流，所以只能靠辐射。辐射需要非常大的表面积。这并非无法解决，而且太空里空间很大。但尽管如此，成本仍然存在。我们会去探索。我们已经在那里了。我们已经做了抗辐射加固。我们的 CUDA 在全球卫星中运行，它们在做成像、图像处理、AI 成像。这类事情应该在太空里完成，而不是把所有数据传回地球再处理。我们应该直接在太空里做成像。所以有很多事情我们应该在太空里做。与此同时，我们会探索太空数据中心的架构。这需要几年时间。没关系，我有的是时间。

Well, we should definitely work on the ground first because we're already here. And number one, number two, we should prepare to be out in space and obviously there's a lot of energy in space. The challenge of course is that cooling you can't take advantage of conduction and convection. And so you can only use radiation. And radiation requires very large surfaces. And so now that's not an impossible thing to solve and there's a lot of space in space. But nonetheless, the expense is still there. We're going to go explore it. We're already there. We're already radiation hardened. We have CUDA in satellites around the world. They're doing imaging, image processing, AI imaging. And that kind of stuff ought to be done in space instead of sending all the data back here and do imaging down here. We ought to just do imaging out in space. And so there's a lot of things that we ought to do in space. In the meantime, we're going to explore what is the architecture of data centers look like in space. And it'll take years. It's okay. I got plenty of time.

我想深入聊聊医疗保健。我知道你在这方面投入很大。我们都到了一定年纪，开始考虑寿命和健康寿命。我觉得我们看起来都不错，有些人更好一些。我不知道你的秘诀是什么，Jensen。

I wanted to double click on health care. I know you've got a big effort there. We're all of a certain age where we're thinking about lifespan, health span. I mean, we all look great, I think. Some better than others. I don't know what your secret is, Jensen.

这些还不错。

Pretty good these these

我是说，你在吃什么？什么不吃？你得在后台告诉我。我想知道你在休息室里有什么秘诀。

I mean, what are you taking? What's off the menu? You got to talk to me when we're backstage. I want to know in the green room what you got going on.

深蹲、俯卧撑和仰卧起坐。

Squats and push-ups and sit-ups.

完美。好的。但就医疗保健的建设而言，它走向何方？我们取得了哪些进展？我刚刚用 Claude 做了一些分析，比如那些账单代码都在哪里？我们在美国花了双倍的钱，但效果似乎只有一半。似乎 15%到 25%的花费用在了初次就诊上。我想我们都知道，像 ChatGPT 这样的大语言模型在初次就诊时能做得更好、更稳定。那么，要突破所有这些监管，让 AI 对医疗系统产生真正的影响，需要做些什么？

Perfect. Okay. But what you know in terms of the build out in health care, where is that going? And what kind of progress are we making? I was just using Claude to do some analysis and saying like where are all these billing codes? We spend twice as much money in the US. We get seem to get half as much. It seemed like 15 to 25% of the dollars spent were on these first GP visits. And I think we all know like ChatGPT and a large language model does a better job more consistently today at a first visit. So, what has to happen there to kind of break through all that regulation and have AI have a true impact on the health care system.

我们在医疗保健领域涉足几个方面。一个是 AI 物理学，或者说 AI 生物学。利用 AI 来理解、表示、预测生物行为。这在药物发现中非常重要。第二个是 AI 智能体，用于辅助诊断等。OpenEvidence 是一个很好的例子。Hippocratics 也是一个很好的例子。我喜欢与这些公司合作。我真的认为这是一个智能体技术将彻底改变我们与医生互动以及医疗保健方式的领域。我们涉及的第三个部分是物理 AI。第一个是 AI 物理学，用 AI 预测物理。第二个是物理 AI，即理解物理定律属性的 AI，用于机器人手术。那里有大量的活动。每一台仪器，无论是超声波、CT 还是未来医院里我们接触的任何仪器，都将是智能体式的。安全版本的开放爪将内置在每一台仪器中。所以在很多方面，这些仪器将以非常独特的方式与患者、护士和医生互动。

There's several areas that we're involved in in health care. One is AI physics. And that's or AI biology. Using AI to understand, represent, predict biology behavior, biological behavior. And so that's one, that's very important in drug discovery. There's second, which is AI agents, and that's where the assistance in helping diagnosis and things like that. OpenEvidence is a really good example. Hippocratics is really good example. Love working with those companies. I really think that this is an area where agentic technology is going to revolutionize how we interact with doctors and how do we interact for health care. The third part that we're involved in is physical AI. The first one's AI physics, using AI to predict physics. The second one is physical AI, AI that understand the properties of the laws of physics, and that's used for robotic surgery. Huge amounts of activities there. Every single instrument, whether it's ultrasound or CT or whatever instrument we interact with in a hospital in the future will be agentic. Open claw in a safe version will be inside every single instrument. And so in a lot of ways that instrument's going to be interacting with patients and nurses and doctors in a very unique way.

在 AI 武器上投资太多了。如果能有一些投资用于 AI 急救员和护理人员，拯救生命而不是夺走生命，那就太好了。

So much investment in AI weapons. It would be wonderful to see some investment in AI EMTs and paramedics and saving lives, not just taking them.

是的。

Yeah.

我觉得这很自然地引出了机器人技术。你有几十个合作伙伴。我们经历了一个非常奇怪的时期，我不知道该叫它失去的十年还是二十年，波士顿动力、谷歌收购了一堆公司，然后又卖掉或剥离了，人们认为机器人技术还没准备好进入主流。而现在，我们有世界上最伟大的企业家，和你并列。马斯克做得很好，希望我圆回来了。Optimus 令人印象深刻。还有中国的其他公司。这离真正进入我们的生活还有多远？我们可能会看到机器人厨师、机器人护士、机器人管家，这些人形机器人真正在现实世界中工作？以你对那些合作伙伴和保真度的了解，尤其是在中国，他们似乎做得和我们一样好，甚至更好？

Which I think is a great segue into robotics. You've got dozens of partners. We had this very weird I don't know what to call a lost decade or 20 years of Boston Dynamics, Google bought a bunch of companies, they then wound up selling them and spinning them out, where people just thought robotics is just not ready for prime time. And now here we have the world's greatest entrepreneur at this time, tied with you. Musk doing Well, that was a good save, I hope. Optimus, pretty impressive. And then, other companies in China. How close is that to actually being in our lives where we might see a robotic chef, a robotic nurse, a robotic housekeeper, you know, these humanoid factor actually working in the real world? Knowing what you know with those partners and the fidelity, especially in China where they seem to be doing as good a job as we're doing here, or maybe better?

我们基本上开创了这个行业。美国开创了它。可以说我们进入得太早了。我们筋疲力尽了。我们在使能技术出现前大约 5 年就厌倦了。我们只是过早地厌倦了。这是第一点。但现在它来了。问题是，从高性能的存在证明到合理的产品还需要多久？技术从来不会超过两三个周期。所以两三个周期大概就是 3 到 5 年。就这样。3 到 5 年内，机器人将无处不在。我认为中国很强大。原因是他们的微电子、电机、稀土、磁铁，这些都是机器人技术的基础，他们是世界上最好的。所以在很多方面，我们的机器人产业深深依赖他们的生态系统和供应链。他们显然行动非常迅速。我们的机器人产业将不得不大量依赖它。全世界的机器人产业都将不得不大量依赖它。所以，我认为你会看到快速的进展。

We invented the industry largely. America invented it. You could argue we got into it too soon. And we got exhausted. We got tired about 5 years before the enabling technology appeared. We just got tired of it just a little too soon. That's number one. But, it's here now. Now, the question is how much longer from the point of high-functioning existence proof to reasonable products, technology never takes more than a couple two, three cycles. And so, a couple two, three cycles would basically be somewhere around 3 years to 5 years. That's it. 3 years to 5 years we're going to have robots all over the place. I think China is formidable. And the reason for that is because their microelectronics, their motors, their rare earth, their magnets, which is foundational to robotics, they are the world's best. And so, in a lot of ways, our robotics industry relies deeply on their ecosystem and their supply chain. And they are obviously moving very quickly. Our robotics industry will have to rely a lot on it. The world's robotics industry will have to rely a lot on it. And so, I think you're going to see fast movements here.

最终，一对一，埃隆似乎认为我们将为每个人配备一个机器人，70 亿对 70 亿，80 亿对 80 亿。

Ultimately, one for one, Elon seems to think we're going to have one robot for every human, 7 billion for 7 billion, 8 billion for 8 billion.

嗯，我希望更多。是的，我希望更多。首先，会有大量机器人在工厂里全天候工作。会有大量工厂机器人，它们不移动，或者稍微移动。几乎所有东西都将是机器人化的。

Well, I'm hoping more. Yeah, I'm hoping more. Well, first of all, there's a whole bunch of robots that are going to be in factories working around the clock. There's going to be a whole bunch of factory robots that don't move. They move a little bit. Almost everything will be robotic.

世界会是什么样子？

What does the world look like?

抱歉，让我这样说：我认为机器人技术对我来说是解锁每个人经济流动机会的关键之一。

Sorry, let me say I think like this is one of the robotics for me is one of the pieces that I think unlocks economic mobility opportunities for every individual.

就像当年每个人都拥有汽车后，他们可以去从事许多不同的工作。当每个人都拥有机器人时，他们的机器人可以为他们做很多工作。他们可以开一家 Etsy 商店或 Shopify 商店。他们可以用自己的机器人创造任何想要的东西。他们可以做到单靠自己无法做到的事情。我认为，机器人最终将成为地球上比以往任何技术都更能为更多人解锁繁荣的最大推动力。

Everyone now, like when everyone got a car, they could now go and do a lot of different jobs. When everyone gets a robot, their robot could do a lot of work for them. They can stand up an Etsy store or a Shopify store. They can create anything they want with their robot. They could do things that they independently cannot do. I think the robot is going to end up being the greatest unlock for prosperity for more people on Earth than we've ever seen with any technology before.

是的，毫无疑问。我的意思是，从简单的数据来看，目前我们缺少数百万劳动力。

Yeah, no doubt. I mean, just the simple math at the moment is we're millions of people short in labor today.

没错。

Right.

是的。

Yeah.

没错。我们其实非常迫切需要机器人技术。这样，如果这些公司有更多劳动力，它们就能增长更多。我的意思是，我们是第一。你提到的一些事情非常有趣。因为有了机器人，我们将拥有虚拟存在。你知道，我可以进入我家里的机器人，远程操作它。我在出差的时候。

Right. We're actually really desperate in need of robotics. And so that all of these companies could grow more if they had more labor. I mean, we're number one. Some of the things that you mentioned are super fun. I mean, because of robots, we'll have virtual presence. You know, I'll be able to go into the robot of my house and virtually operate it. I'm on a business trip.

对。

Right.

让它在家四处走动。

Tell it to walk around the house.

还有遛狗。

And walk the dog.

对，遛狗。

Yeah, walk the dog.

耙树叶。

Rake the leaves.

对，没错。

Yeah, exactly.

把狗耙起来。

Rake up the dog.

可能没那么夸张，但就是，你知道，四处走走，看看家里发生了什么，和狗聊聊天，和孩子聊聊天。对。

Maybe not quite that, but just, you know, wander around and just see what's going on in the house, you know, chat with the dogs, chat with the kids. Yeah.

对。

Yeah.

那是一个……

That's a...

时间旅行也是，我们将能够以光速旅行，所以显然想先把我们的机器人派过去。

Time travel is also we're going to be able to travel at the speed of light, you know, and so, clearly want to send our robots ahead of us.

对。

Yeah.

不会派我自己去。我会派一个机器人。

Not going to send myself. I'm going to send a robot.

对。

Right.

你知道。

You know.

去看看。

Check it out.

对，对。然后我会上传我的 AI。

Yeah, yeah. And then I'm going to upload my AI.

嗯，这是不可避免的。它解锁了月球和火星作为殖民目标，这给了我们无限的资源。从月球返回实际上移动材料的能量成本为零，因为你可以利用太阳能加速。所以，你可以在月球上建立制造世界所需一切的工厂，而机器人将是实现这一点的关键。

Well, it's inevitable. It unlocks the moon and it unlocks Mars as targets for colonization, which gives us infinite resources. Getting back from the moon is effectively zero energy cost to move material back because you can use solar and accelerate. So, you could have factories that make everything the world needs on the moon, and the robots are going to be the unlock for enabling that.

距离不再重要。

This distance no longer matters.

距离不重要。

Distance doesn't matter.

是的。

Yep.

对。

Yeah.

我们从模型和智能体中获得的收入越多，就越能投资于基础设施建设，这反过来又解锁了模型和智能体更多的能力。Dario 在 Dwarkesh 的播客中最近说：“到 2027-28 年，模型公司和智能体公司将产生数千亿美元的收入。”他预测到 2030 年将达到一万亿美元，对吧？这是非基础设施的 AI 收入。

The more revenue we get out of models and agents, the more we can invest in building the infrastructure, which then unlocks more capabilities on models and agents. Dario on Dwarkesh's podcast recently said, "By 2027-28, we'll have hundreds of billions of dollars of revenue out of the model companies and the agent companies." And he forecasts a trillion dollars by 2030, right? This is non-infrastructure AI revenue.

我认为他非常保守。我相信 Dario 和 Anthropic 会做得比那好得多。

I think he's being very conservative. I believe Dario and Anthropic is going to do way better than that.

哇。

Wow.

好得多。

Way better than that.

哇。

Wow.

对，然后从 300 亿到一万亿。

Yeah, and then 30 billion to a trillion.

是的。原因在于他没有考虑到的一点是，我相信每一家企业软件公司也将成为 Anthropic 代码、Anthropic 令牌的转售商、增值转售商，以及 OpenAI 的增值转售商。

Yep. And the reason for that is the one part that he hasn't considered is that I believe every single enterprise software company will also be a reseller, value-added reseller of Anthropic code, Anthropic's tokens, value-added reseller of OpenAI.

没错。他们将会……

That's right. And they're going to that part of their...

你看，对数级扩张。

Get this, logarithmic expansion.

是的。

Yes.

对。

Yeah.

他们的市场推广将大幅扩展。

Their go-to-market is going to expand tremendously.

在那个世界里，你认为护城河是什么？还剩下什么？我的意思是，有些护城河坦率地说，我认为随着规模扩大，几乎无法逾越。最好的一条没人谈论的可能是 CUDA，这简直是一个不可思议的战略优势。但未来，如果一个模型可以用来创造令人难以置信的东西，那么下一个模型迭代可能会颠覆它。在你看来，对于这些在应用层构建的公司，它们的护城河是什么？它们如何差异化？

What do you think in that world is the moat? What's left over? I mean, you have some moats that are frankly, I think, as this scales, almost insurmountable. The best one that nobody talks about is probably CUDA, which is just like an incredible strategic advantage. But in the future, if a model can be used to create something incredible, then the next spin of a model can be used to maybe disrupt it. Sort of in your mind, what do you think for these companies that are building at that application layer? What's their moat? Like, how do they differentiate themselves?

深度专业化。深度专业化。我相信这些模型——它们将拥有通用模型，这些模型连接到软件公司的智能体系统中。对。其中许多模型是云端模型和专有模型，但许多模型是它们自己训练的专业化子智能体。

Deep specialization. Deep specialization. I believe that these models that they're going to have general models that are connected into the software company's agentic system. Right. Many of those models are cloud models and proprietary models, but many of those models are specialized sub-agents that they've trained on their own.

所以，你对企业家的号召是，看好你的垂直领域。

So, the call to arms for you for entrepreneurs is, look, know your vertical.

没错。

That's right.

比任何人都更深入、更了解它。

Know it as deep and as better than everybody else.

没错。

That's right.

然后等待这些工具，因为它们正在赶上你，现在你可以用你的知识来注入它们。

And then wait for these tools because they're catching up to you and now you can imbue it with your knowledge.

没错。你越早连接你的智能体，越早让你的智能体与客户连接，那个飞轮就会让你的智能体变得……

That's right. And the sooner you connect your agent, the sooner you connect your agent with customers, that flywheel is going to cause your agent to get...

这和我们今天所做的完全相反，因为今天我们构建一个软件，然后问：什么能通用？然后尽可能广泛地销售，再围绕它销售定制化。

It very much is an inversion of what we do today because today we build a piece of software and we say, what generalizes? And then let's try to sell it as broadly as possible and then sell the customization around it.

我们陷入了……事实上，完全正确。我们创建了一个横向平台，但注意有所有这些 GSI 和所有这些专家顾问，他们把你的横向平台专业化成……

And we trap... In fact, exactly right. We create a horizontal, but notice there are all these GSIs and all of these consultants who are specialists who then take your horizontal platform and specialize it into...

没错。而且可以说，定制化是一个大五到六倍的行业。

Exactly. And that's arguably a five or six time bigger industry is the customization.

是的。绝对。

It is. Absolutely.

对，整个……确实如此。

Yeah, the whole... That very much is.

没错。所以，我认为这些平台公司有机会成为那个专家，成为那个垂直领域。

That's right. So, I think that these platform companies have an opportunity to become that specialist, to become that vertical.

对。

Right.

对，领域专家。

Yeah, domain expert.

我只想给你点赞。我想是三年前你说过，你不会因为 AI 而失业，你会因为使用 AI 的人而失业。现在我们在这里，整个对话都围绕着智能体让人变得超人、商业机会扩大和创业精神扩展的概念。你实际上看得很清楚。你改变看法了吗？

I just want to give you your flowers. I think it was three years ago you said you're not going to lose your job to AI, you're going to lose your job to somebody using AI. And here we are, the entire conversation has revolved around this concept of agents making people superhuman and the business opportunity expanding and entrepreneurship expanding. You actually saw it pretty clearly. Have you changed your view?

嗯，我不是悲观主义者。

Well, I'm not a doomer.

我确实……不，你可以容纳两种想法。一种是会有……

I do have... No, you can hold space for two ideas. One is there are going to be...

我们称之为螺旋 J Cal。

That's spiral J Cal we call it.

不，不。你可以看到……

No, no. There you can see...

嗯，那只是因为他和我混得不够多。

Well, that's just because he doesn't hang out with me enough.

嗯，我们聊过一点。

Well, we talk a little bit.

小心你……

Be careful what you...

我们不谈这个。

We don't talk about it.

他会出现在你的早餐桌上，跟着你到处走。

He will show up at your breakfast table and he'll follow you around.

我不是在要求。我只是说说。

I'm not asking for it. I'm just saying.

他会跟着你。

He'll follow you around.

我不是在要求。

I'm not asking for it.

你可以和我还有 Tucker 一起去，我们每年一月在日本滑雪。我们很喜欢。

You can come with me and Tucker we ski in Japan every January. We love it.

我和 Tucker 公路旅行。

Me and Tucker road trip.

好的。

Okay.

工作岗位将会被取代，问题在于这些人是否有勇气和决心去拥抱这些技术？我们将看到人类驾驶完全消失。这是一件好事，能拯救生命，但我们必须认识到，美国有 1500 万人，1000 到 1500 万人从事驾驶工作。所以这将会发生，对吧？

There is going to be job displacement and then the question becomes, do those people have the fortitude, the resolve to then go embrace these technologies? We're going to see 100% of driving go away by humans. That's a beautiful thing and the lives saved, but we have to recognize that's 15 million people in the United States, 10 to 15 million who are employed in that way. And so that is going to happen, yes?

我认为工作会发生变化。例如，现在有很多司机开车。我相信其中许多司机实际上会坐在车里方向盘后面，而汽车自己驾驶。原因在于，记住司机是做什么的。归根结底，这些司机在帮助你，他们是你的助手，帮你拿行李，帮你做很多事情。所以，如果未来的司机成为你的出行助手，帮你做一大堆其他事情，而车自己开，我其实不会感到惊讶。

I think that jobs will change. For example, there are many chauffeurs today who drive the car. I believe that many of those chauffeurs will actually be in the car sitting behind the steering wheel while the car is driving by itself. The reason for that is because remember what a chauffeur does. In the end, these chauffeurs are helping you, they're your assistants, they're helping you with your luggage, they're helping you with a lot of things. So I would be surprised actually if the chauffeurs of the future become your mobility assistant and they are helping you do a whole bunch of other stuff. Yeah, and the car's driving by itself.

飞机上的自动驾驶仪创造了更多的飞行员，并没有让任何飞行员离开驾驶舱，尽管自动驾驶仪 90%的时间在驾驶飞机。

The autopilot in planes created a lot more pilots and didn't take any of the pilots out of the cockpit even though the autopilot is flying the plane 90% of the time.

顺便说一句，当车自己驾驶时，那个司机会在手机上做一堆其他工作，通过做其他事情赚钱。

And by the way, while that car is driving itself, that chauffeur is going to be doing a bunch of other work on his phone and he's going to be making money doing other things.

做其他事情。为你协调一堆事情。

Doing other stuff. Coordinating a bunch of things for you.

整个蛋糕就这么变大了。

It's all the pie just grows in a way.

所以有一点是，是的，每个工作都会被改变。一些工作会被淘汰。然而，我们也知道很多工作会被创造出来。我想对即将毕业、对 AI 感到担忧和焦虑的年轻人说：成为使用 AI 的专家。

So one of the things is that yes, every job will be transformed. Some jobs will be eliminated. However, we also know that many jobs will be created. The one thing that I will say to young people who are coming out of school, who are concerned, who are anxious about AI, be the expert of using AI.

是的。

Yes.

到什么程度？听着，我们都希望员工成为使用 AI 的专家，这并不简单。不简单。所以，知道如何指定，而不是过度规定，给 AI 留出足够的创新和创造空间，同时引导它达到我们想要的结果，所有这些都需要艺术性。

How much? Look, we all want our employees to be expert at using AI, and it's not trivial. Not trivial. And so, knowing how to specify, not to over prescribe, leaving enough room for the AI to innovate and create while we guide it to the outcome we want, all of that requires artistry.

你在斯坦福时有过一个很好的建议，我记得是‘我祝你们痛苦和磨难’。你记得吗？

You had this great advice when you were at Stanford, I think it was, which is, 'I wish you pain and suffering.' Do you remember that?

记得。

Yeah.

太棒了。你对年轻人学习什么有什么建议？如果他们即将高中毕业。因为现在这些孩子非常原生。他们还没有决定是否上大学、学什么。你怎么引导这些孩子？你会告诉他们什么？

Fantastic. What's your advice to young people around what they should be studying? So, if they're about to leave high school. Because now those are the kids that are at this like really native. They haven't made a decision about college, what to study, if at all go to college. How do you guide those kids? What would you tell them?

我仍然相信深科学、深数学、语言技能。你知道，语言现在是 AI 的编程语言。

I still believe that deep science, deep math, language skills. You know, as you know, language is the programming language of AI now.

编程语言。

Programming language.

所以，事实证明，英语专业可能是最成功的。因此，我认为我只会建议，无论你接受什么教育，确保你成为使用 AI 的深度专家。

And so, as it turns out, it could be that the English major could be the most successful. So I think I would just advise whatever education you get, just make sure that you're deeply expert in using AIs.

关于工作，我想说一件事，希望大家都能听到。事实上，在深度学习革命之初，一位我非常尊敬的顶尖计算机科学家说，计算机视觉将完全取代放射科医生。他建议大家都不要进入放射学领域。10 年后，他的预测 100%正确。计算机视觉已经 100%整合到世界上所有的放射学技术和平台中。令人惊讶的结果是，放射科医生的数量实际上增加了，需求也在飙升。原因是每个人的工作都有目的和任务。你的任务是研究扫描图像，但你的目的是帮助医生、帮助病人诊断疾病。所以令人惊讶的是，由于扫描现在可以非常快地进行，他们可以做更多的扫描来改善医疗保健。

One of the things that I wanted to say with respect to jobs, and I want everybody to hear it, is that in fact, at the beginning of the deep learning revolution, one of the finest computer scientists in the world, deeply deeply I deeply deeply respect, said that computer vision will completely eliminate radiologists. And that the one field he advises everybody to not go into is radiology. 10 years later, his prediction was 100% right. Computer vision has been integrated into all of the radiology technologies and radiology platforms in the world, 100%. The surprising outcome is the number of radiologists actually went up and the demand for radiologists is skyrocketing. The reason for that is because everybody's job has a purpose and it's task. The task that you do is studying the scans. But your purpose is helping the doctors, helping the patient diagnose disease. And so what's surprising is because the scans are now being done so quickly, they could do more scans improving health care.

是的。

Yes.

但更快地进行更多扫描可以让患者更快入院、更快治疗，而且事实证明，医院也喜欢赚钱。

But doing more scans more quickly allows patients to be onboarded a lot more quickly, treated a lot more quickly, and as it turns out, because hospitals enjoy making money too.

是的。

Yeah.

对吧。他们做更多的扫描。

Right. They're doing more scans.

读更多的片子。

Reading more.

他们治疗更多的客户和病人。收入在增加，你猜怎么着？他们需要更多。

They're treating more customers and more patients. The revenue is going up and guess what? They need more.

在一个增长更快的国家，生产力提高，更富裕的国家可以在教室里放更多的老师，而不是更少的老师。你只需给每个老师一个针对每个学生的个性化课程。这让他们都变得超人，并带来更多。

And in a country that grows faster, productivity increases, a wealthier country can put more teachers in the classroom, not less teachers in the classroom. You just give every one of those teachers a personalized curriculum for every student in the room. It makes them all bionic and leads to a lot more.

每个学生都会得到 AI 的帮助，但每个学生都需要优秀的老师。

Every single student will be assisted by AI but every single student will need great teachers.

是的。

Yeah.

太棒了。Jensen，祝贺你取得的所有成功，这真是一场非常积极、令人振奋的讨论。我们非常感谢你抽出时间。

Amazing. Jensen, congratulations on all your success and really this is an incredibly positive uplifting discussion. We really appreciate you taking the time for us.

他是我们需要的管理者。

He is the steward we need.

你就是我们需要的管理者。

You are the steward we need.

我认为你需要更直言不讳——我一直在非常直言不讳地谈论积极的一面。我认为悲观论调太多了。

I think you need to be more vocal about I'm being very vocal about the positive side of it. I think there's too much doomerism.

但我也认为，达到这种成功水平需要谦逊，并且要谦虚地说我们只是在做软件，伙计们。是的。我认为人们听到这个其实非常有益。我们以前做过这种事。我们以前创造过类别和行业。我们不需要陷入这种危言耸听的境地。这毫无意义。

But I also think it takes the humility to have this level of success and be humble about we're making software, guys. Yeah. And I think that's actually really healthy for people to hear. We have done this before. We have invented categories and industries before. We don't need to go to this scaremongering place. It does nothing.

而且我们可以选择，对吧？我们有自主权和能动性。我们可以选择如何部署它。

And we get to choose, right? We have autonomy and agency. We get to pick how to deploy this.

我们当然可以。

We sure do.

好了，各位。我们下次 All-In 访谈再见。好的。干得好，兄弟。

Okay, everybody. We'll see you next time on the All-In interview. Okay. Well done, brother.

谢谢，伙计。

Thanks, man.

干得好。

Good job.

谢谢你，先生。太棒了。

Thank you, sir. That was awesome.

好。好。

Good. Good.

你们太棒了。

You guys are awesome.

Jason，谢谢你。

Jason, thank you.

看看这个。看看你们身后的人群。

Look at this. Look at this big crowd behind you guys.

老兄，我觉得他们是冲你来的。

Man, I think they're here for you.

我要全力以赴。

I'm going all in.