高通 CEO 克里斯蒂亚诺·安蒙谈个人 AI 设备与移动智能体的未来

本周的《Bold Names》节目，我们对话 Cristiano Amon。他是高通公司的 CEO。这家科技巨头之所以闻名，是因为它为许多人离不开的设备——智能手机——提供核心动力。

This week on Bold Names, we're talking to Cristiano Amon. He's the CEO of Qualcomm. The tech giant's claim to fame is that it powers the device many of us can't live without, the smartphone.

当你触摸你的智能手机时，你接触到的就是一项高通技术。

When you touch your smartphone, you have a piece of Qualcomm technology.

现在，Cristiano 正展望一个由我们与人工智能的关系所主导的未来。那么，那会是什么样子？接下来揭晓。Cristiano，感谢你参加《Bold Names》。非常高兴你能来。你的公司为许多最受欢迎的数字设备提供核心部件。我想如果回到 100 年前，这就像是你把发动机卖给亨利·福特本人。除此之外，你的公司还与开发人工智能设备的领先科技公司合作。呃，这真正是我们与 AI 互动方式的物理体现。呃，你们的芯片用在 Meta 的智能眼镜里。你拥有的这个位置让你能看到行业趋势的走向。所以我很好奇，当涉及到 AI 设备时，近未来会是什么样子？你对今年、明年有什么预期？世界正在如何变化？

Now, Cristiano is looking to a future dominated by our relationship with artificial intelligence. So, what does that look like? That's next. Cristiano, thanks for joining Bold Names. So excited to have you here. Your company provides the guts to many of the most popular digital devices out there. I think if we were 100 years ago, it would be like you're selling the engines to Henry Ford himself. On top of that, your company is working with the leading tech companies that are developing artificial intelligent devices. Uh the really the physical manifestation of how we're going to interact with AI. Uh your your chips are in meta, smart glasses. This perch you have allows you to see the trends where things are going in the industry. And so I'm curious, what does the near future look like when it comes to AI devices? What do you expect this year, next year? How's the world changing?

好吧，这是个复杂的问题，Tim。所以我们得好好拆解一下。呃，你知道，我们是做芯片的公司，人类使用的芯片。而 AI 正在改变一切。它将从根本上改变计算机和整个计算领域。你知道，我喜欢这样想：AI 只是你进行软件、架构和计算以及我们与计算机互动的一种不同方式。我认为这可能是理解那些设备将会发生什么的基础。让我从一个非常基本的概念开始，我想我们已经讨论这个话题好几年了。实际上，我们在它流行之前就开始谈论了。我想我们应该把这个术语注册商标。AI 是新的用户界面，这是未来的趋势。呃，计算机现在能理解我们了。所以我们不必学习如何与计算机设备互动。我们说的话、我们看到的东西，所有这些都将被计算机理解。这将改变计算机、操作系统、应用程序。你

All right, that's a loaded question, Tim. So, we're going to have to unpack that. Uh, you know, we're the company doing the chips that the humans use. And AI is changing everything. It is going to fundamentally change computers and computing in general. You know, I like to think about it. AI is just a different way that you're going to be doing software and architecture and computing and how we interact with computers. And I think that's probably the foundation to understand what's going to happen with those devices. And let me just start with a very basic concept that I think we've been talking about for a couple years. Actually, we started talking about this before it was popular. I think we should have trademarked the terminology. AI is the new UI going forward. Uh the computer now understands us. So we don't have to learn how to interact with the computer device. What we say, what we see, all of those things are going to be understood by the computer. That's going to change the computer, the OS, the application. You

你将改变与这个数字世界互动的方式。对。现在，如果你想与 AI 互动，你打开智能手机，进入你喜欢的任何应用，它连接到云端并给你发送答案。这就是现在的世界。但你现在看到的是截然不同的东西，这就是你所说的。

You're going to change how you interact with this digital world. Right. Right now, you go onto your smartphone if you want to interact with AI, you go to the whatever app you like and it goes off to the cloud and sends you an answer. That's kind of the world. But you're seeing it something much different now is what you're talking about.

是的。让我们从移动设备开始，因为它可能是最大的消费电子产品。它实际上是人类有史以来创造的最大的平台。

Yes. And let's just start with mobile devices because it's probably the largest consumer electronic product. It's actually the largest platform ever created by mankind.

是的。这对你来说是一个巨大的驱动力。巨大的收入。你最大的收入驱动力，对吧？移动电话，对吧？智能手机。

Yeah. It's a big driver for you. Big revenue. Your biggest revenue driver, right? Mobile phones, right? Smart phones.

是的。我们正在努力多元化，但我们热爱移动业务。我们会一直热爱移动业务，但让我们继续这个思路。现在你开始看到的是新型移动设备的出现。我们称之为个人 AI 设备。其中一个例子是眼镜。我想你提到了 Meta 眼镜。所以眼镜非常自然，因为首先我想你和我一直在使用眼镜。我认为人类使用眼镜。他们不会戴着大头盔到处走，但他们会使用这样的眼镜。它贴近我们的感官。它贴近我们的眼睛、嘴巴、耳朵。当你转头时，无论你在看什么，如果上面有摄像头，模型就在看。我们开始看到这些新型的计算机与人类互动，与计算机纯智能体式应用的互动，因为如果你现在有一个智能体，它正在与一个大语言模型或多个大语言模型合作，突然之间这个智能体理解了你的意图并连接到互联网，那么它就可以做任何事情。它取代了应用，它将取代一些工作负载，让我给你一些例子。我要让这个对你来说非常实际。

Yeah. We are working hard to diversify but we love mobile. We're always going to love mobile but let's keep going down this thought process. So now one of the things you're starting to see is the emergence of new type of mobile devices. We call them personal AI devices. One example of that is the glasses. I think you mentioned about the meta glass. So glasses is very natural because first of all I think you and I have been using glasses. I think humans use glasses. They're not going to walk around with big helmets, but they will be using glasses like this. It's close to our senses. It's close to our eyes, to our mouth, to our ears. When you turn your head, whatever you're looking at, if you have a camera on it, the model is looking. And we're starting to see those new type of computer-human interactions with computer pure agentic applications because if you have an agent now that is working with a large language model or multiple large language models and all of a sudden the agent understands your intentions and is connected to the internet then it can do anything. It replaces the app and it's going to replace some of the workloads and let me give you some examples. I'm going to make this very practical for you.

好的，请讲。

Yeah, please.

所以，举个例子，我喜欢用银行业务这个例子，因为我们都有银行应用。所以，今天你拿起手机，点击图标。这是 Tim 的支票账户。这是 Tim 的余额。我确信那里有很多钱。他所有的

So, let's say for example, I like to provide this example of banking because all of us have a banking app. So, you go today, you think about this banking app. You pick up your phone, you click on the icon. Here's Tim's checking account. Here's Tim's balance. I'm sure there's a ton of money over there. His all his

哦，绝对。Bold Names 付得很好。是的。

Oh, absolutely. Bold Names pays very well. Yeah.

是的。他最近的几笔交易是什么，对吧？然后你走向这个东西。你会有你的智能眼镜。你会看到你想买的东西。你只需问模型：“我想买这个。我可能想用我支票账户里的钱。我的支票账户里有多少钱？”模型会有你的凭证，就像你把凭证给了应用一样，它会说：“你有这么多钱。最近五笔交易是什么？那张支票到账了吗？”说：“是的，到了。你能帮我买这个吗？”你会收到一张账单。你会看着账单说：“帮我付这个。完成后请通知我。”这将彻底改变我们对银行业的看法。这将无处不在。就像帮我叫 Uber，在这家餐厅预订。它还将改变我们互动的方式。例如，我们开始看到 Meta 在眼镜上做的一些很好的用例。你可以在 WhatsApp 通话中。你开始广播那个通话。嗯，你如何看待社交媒体。可能是这样的：我和我的朋友在散步，我说：“哦，那边那个人是谁？让我看看。哦，是 Tim。你要我跟他打个招呼吗？”所以，这就是移动设备将如何改变。我们只是转型的开始，而且只是在那些个人 AI 设备上。

Yes. Who was his last few transactions, right? and you're going to this thing. You're going to have your smart glasses. And you're going to see something you want to buy. And you just ask the model, "I would like to buy this. I probably want to use the money I have in my checking account. How much do I have in my checking account?" And the model will have your credentials like you gave your credentials to the app and we'll say, "This is how much you have. What was the last five transactions? Did that check got deposit?" Say, "Yes, it did. Can you buy this for me?" You're going to get a bill. You're going to look at the bill, say, "Pay this for me. Please notify me when I'm done." It's going to completely change how we think about banking. And this is going to go everywhere. It's going to be like call me an Uber, make a reservation in this restaurant. And it's going to also change how we interact. For example, we're starting to see some great use cases with Meta is doing on the glasses. You can be on a WhatsApp call. You start broadcasting that call. Um how you think about social media. It could be it's like I'm walking with my buddy and I say, "Oh, who's this person over there? Let me check. Oh, it's Tim. You want me to say hi to him?" So, that's how mobile devices are going to change. We're just at the beginning of the transformation and just on those personal AI devices.

是的。我想我很好奇。你看好眼镜，因为真正好看的人往往戴眼镜。我想你和我都同意这一点，对吧？非常时尚。

Yeah. I guess I'm curious. You're bullish on glasses because really good-looking people tend to wear glasses. I think you and I agree with that, right? Real stylish.

100% 同意。

100% agree with that.

但你认为还有其他功能，其他形态。除了眼镜，你还看好什么？

But you think there are other function other form factors out there. What are you bullish on other than glasses?

是的。所以我们看到几种形态。我们看到眼镜。我们看到耳塞。我们看到珠宝。我认为人类一直佩戴珠宝，而且

Yes. So, we've seen a couple form factors. We see glasses. We see earbuds. We've seen jewelry. I think humans have been wearing jewelry and

钢笔或类似的东西在翻领上

pens or something like that on the lapel

别针吊坠。我们看到手表。我认为这很令人兴奋。那些东西将与你的手机一起工作。你的手机将扮演双重角色。它将继续做手机的事情，我们将在其中一些应用中做大量的处理并拥有 AI 的前端。这和我们之前看到的没什么不同。

pin pendants. We've seen watches. I think this is exciting. And those things are going to work alongside your phone. Your phone is going to have like a dual role. It's going to keep doing phone things and we'll be doing a lot of the processing and having a front end of AI in some of those applications. And it's kind of no different than what we've seen before.

Tim，我记得在智能手机的早期，我们都习惯在笔记本电脑上做电子商务。

Tim, I remember in the early days of the smartphone, we all used to do e-commerce on our laptop.

现在不一样了。目前世界上大多数人都在手机上做电子商务，比如在美国，人们在手机上亚马逊购物。这就是未来会发生的事情。你会拥有那些新设备，然后自然而然地开始在新设备上做事，移动生态系统就是这样演变的。

Not anymore. Most of the world right now does e-commerce, buys things on Amazon, for example, in the United States on their phone. This is what's going to happen. You're going to have those new devices and you're going to start naturally doing things on those new devices and that's how the mobile ecosystem is going to evolve.

你似乎仍然看好手机。它不会很快消失。我认为科技界有些人将 AI 及其带来的个性化，以及不同种类应用和新生态系统出现的新能力，视为取代一些智能手机制造商的机会。

You seem to still remain bullish on the phone. It's not going away anytime soon. I think some in tech see AI and the kind of personalization that it allows and the new ability for different kinds of apps and new ecosystems to emerge as an opportunity to displace some of the smartphone makers.

是的。让我详细说明一下，因为如果说有什么的话，我认为我们在移动行业有一些经验，几年的经验，我们见过——我不愿这么说——但我们见过一长串从英雄到零的公司，我们见过很多这样的例子，尤其是因为你有重大的技术转型，所以公司可以改变。我认为参与者可以改变。我认为真正捕获价值的公司可以改变，不一定。我认为设备可以改变。让我给你举一些实际的例子。

Right. Let me elaborate on that because if anything I think we have a little bit of experience with the mobile industry, a few years of it, and we have seen—I hate to say it that way—but we have seen a big list of companies that went from hero to zero, like we've seen many of those, and especially because you have big technology transitions so companies can change. I think the players can change. I think the companies that actually capture value can change, not necessarily. I think the device can change. Let me walk you to some practical examples of that.

当笔记本电脑被发明时，所有人都说台式机已死。台式机没死。没死。

When the laptop was invented, everybody said the desktop is dead. Desktop's not dead. Not dead.

我想我写过一些这样的故事。但它变了。它变成了工作站。它变成了游戏 PC。你可以在酒店业看到很多台式机。你看它变得不同了，专用于特定用例。然后当我们开始用智能手机时，很多人说笔记本电脑已死。它没死。所以我认为手机不会消失。但你用手机做的一些事情，你不会再用手机做。你会在其他地方做。这里给你一个提示：谁有能力理解人类意图，谁就有机会成为赢家并捕获最大价值。例如，如果未来你手机上的助手变得非常有用，因为它是你选择交互的智能体，也许会有多个智能体。一旦智能体理解了 Tim 想要什么，无论是去某个地方，告诉我我的日历，打电话给某人，支付某物。当智能体理解交易的价值在于理解人类意图，然后智能体可以自由地去互联网、去不同的应用做一切事情。所以那个信息，那就是生态系统价值所在。不是操作系统，不是应用商店。它会改变。我认为我们已经看到了无线行业的几代人，那些事情是如何改变的。所以我认为我们会再次看到变化。我们不知道它们在哪里，但这并不意味着手机会消失。

I wrote a few of those stories, I think. But it changed. It became like it became a workstation. It became a gaming PC. You can see a lot of desktops in hospitality. You see it became different, dedicated to a particular use case. Then when we started with the smartphone a lot of people said the laptop is dead. It's not dead. So I think the phone is not going to go anywhere. But some of the things you do with your phone you're not going to do with your phone. You're going to do somewhere else. And here's a hint for you. Whoever has the ability to understand the human intention has the opportunity to be a winner and capture the most value. For example, if in the future the assistant on your phone becomes incredibly useful because it's the agent that you chose to interact with and maybe there's going to be multiple agents. Once the agent understands what Tim wants, whether it's going to go somewhere, tell me about my calendar, call somebody, pay something. When the agent understands the value on that transaction is in the understanding of the human intention and then the agent is free to go to the internet, to go to different apps and do everything. So that information, that's where the value of the ecosystem is going to be. It's not the OS, it's not the app store. It's going to change. And I think we've seen those generations of the wireless industry how those things have changed. So I think we're going to see change again. We don't know where they are, but it just doesn't mean the phone is going to disappear.

所以你在谈论一个世界，这些设备收集用户的各种个人信息，然后他们可以利用这些信息帮助用户导航生活。对。我认为有人担心所有数据进入云端可能成为监控国家或隐私问题。你对此有何看法？

So you're talking about a world where these devices are picking up all sorts of personal information about the user and then they can use that information to help them navigate their life. Right. I think there's some concern that all that data going into the cloud could be surveillance state or it could be a privacy issue. What's your thought on that?

嗯，这就是为什么有像高通这样的公司。看，我们做的很多事情实际上是我们所谓的边缘计算。

Well, that's the reason there are companies like Qualcomm. Look, a lot of the things that we do is really what we call on the edge.

让我们稍微解释一下。你说的边缘是什么意思？边缘计算，对吧？

Let's unpack that a little bit. What do you mean by edge? Edge computing, right?

是的。边缘现在是一个术语，用于数据中心之外的一切。

Yes. Edge now is a term that has been used for everything that is outside the data center.

所以它发生在你的设备上。

So it happens on your device.

是的。没错。它是你手机、PC、汽车里的处理器。那么发生了什么？这实际上是一个很好的机会来谈论现在 AI 领域正在进行的有趣讨论。经常有关于在哪里进行处理的争论。你是在云端进行处理？还是在边缘进行处理，也就是在你的设备上？哪个更好？哪个会赢？我认为问这个问题的方式是错的，我回到今天的智能手机。今天的智能手机有很多处理能力。每次我们买新手机，我们想要更快的处理器。它是最云连接的设备。如果你打开飞行模式，你会对你的设备非常沮丧。你不会谈论它。设备上的东西是在云端完成的。我认为这就是 AI 的未来。AI 将会——有些事情你会在设备上做。有些事情会在云端做，它 simply 会工作。但你之前说的是一个非常重要的问题。当你考虑智能体时，上下文非常重要。你知道，这是 Tim。我知道 Tim。我知道 Tim 现在在哪里。所以，当他问我一个问题时，我必须拥有那个上下文。否则，我对 Tim 不太有用。就像，特别是如果这是一个智能体，他知道你在工作还是在家，你需要它就像你在和另一个人说话。所以，那个上下文非常重要。所以我们正在开发芯片，例如我们所谓的环境 AI、感知 AI。那些东西在你的设备上运行，在你的设备上提供上下文分析，并通知你设备上的智能体，并且真正由用户控制他是否想把这些东西发送到云端。但我认为我们将在不久的将来看到这种情况发生。你将拥有一个在设备上完成的处理器，其意图实际上是具有给你那个上下文的能力。

Yes. Exactly. It's the processor in your phone, in your PC, in your car. So what's happening? This is actually a good opportunity to talk about an interesting discussion that is happening now with AI. There's often this debate about where are you going to do the processing. Are you going to do the processing in the cloud? Are you going to do the processing on the edge, which means on your device? Which one is better? Which one's going to win? And I think that's the wrong way to ask the question and I go back to our smartphones today. Our smartphone today has a lot of processing power. Every time we buy a new phone, we want a faster processor. It is the most cloud-connected device. If you put on airplane mode, you're going to be very frustrated with your device. And you don't talk about it. What's on the device was done on the cloud. I think that's what the future's going to be with AI. AI is going to be—there's certain things you're going to do on device. There's certain things going to do on the cloud and it's simply going to got to work. But what you said before is a very important question. When you think about agents, context is really important. You know, this is Tim. I know about Tim. I know where Tim is right now. So, when he asks me a question, I have to have that context. Otherwise, I'm not very useful for Tim. Like, especially if this is an agent, he knows when you're at work or you're home, you need it's like you're talking with another human. So, that context is very important. So we're developing chips for example for what we call ambient AI, perception AI. Those things run on your device, provide an analysis of the context on your device and inform the agent on your device and is really going to be the control of the user whether he wanted to send those things to the cloud or not. But I think we're going to see that happening in the near future. You're going to have processor that's done a device with the intent of actually having the ability to give you that context.

当我想到高通时，我想到你们制造的那些非常高效的芯片，这就是为什么它们在智能手机上如此需要或如此受欢迎，但因为它们没有很多电池，你知道，你希望它能持续一整天。这就是为什么我有点惊讶，我们越来越多地听到谈论，你谈论你在数据中心的作用，对吧？世界上对 AI 的如此多关注是资金流入这些数据中心。这就是 AI 的训练方式。这就是它的运作方式。大量资金流向英伟达购买他们的芯片。你认为你在数据中心领域竞争的位置在哪里？

When I think of Qualcomm, I think of those chips that you make that are so efficient, which is why they're so needed or so desired for smartphones, but because they don't have a lot of batteries, you know, you want it to be able to last all day. That's why I'm a little surprised that increasingly we're hearing talk and you talk about your role in data centers, right? So much attention of AI out there in the world is money going into these data centers. That's how AI is trained. It's how it operates. Lots of money going to Nvidia for their chips. Where do you see your spot there to compete in the data center realm?

是的。你知道，如果你实际上在 Google 搜索框中输入，在 ChatGPT 或 Gemini 上问这个问题，你输入搜索数据中心和能源，你会发现很多关于数据中心巨大增长预测的新闻。

Yes. You know, if you actually go type in on a Google search box on the ChatGPT I think or Gemini ask the question, you put it in on the search data center and energy and you're going to find a lot of news about that there's a massive projection of growth of data center.

那些数据中心消耗大量能源，对吧？它们只是需要大量电力。

They suck up so much energy those data centers, right? They just require just a lot of electricity.

Elon Musk 想把数据中心搬到外太空，因为那里有太阳能，或者你听到其他这些人说要用核反应堆来让数据中心持续运转。所以没错，它们耗电量巨大，对吧？

Elon Musk wants to take data centers to outer space because of solar power up there, or you hear all these other folks talking about having nuclear reactors to keep their data centers humming. So yeah, they suck up a lot of juice, right?

所以这正是你会看到大量讨论的地方。人们基本上会说，这是我们对数据中心算力需求的预测，这是能源的可用量，两者之间存在脱节，对吧？

So that's where you're going to see a lot. People are basically going to say this is the projection of what we need in computing with a data center, this is the availability of energy, there's a disconnect, right?

对吧？所以这是相关的。

Right? So that's connected.

所以这正是 Qualcomm 可能成为一个有意思的参与者的地方，因为我们的基因来自电池供电的设备。我们必须设计功耗非常高效的芯片。我知道我们花了很多时间谈论手机。我想从工程角度举这个例子。手机极具挑战性，因为你有能跑光线追踪游戏的 GPU，你还有同时运行数十亿参数 AI 的神经处理单元。然后你还要以千兆级的数据速率向基站传输。所有这些都在发生，而设备必须能放进口袋，不能发烫。不能发烫。你会用那个设备贴着脸。而且电池必须撑一整天。算力多大都不重要。作为工程师，我们拿到的预算是：电池就这么大，必须撑一整天。这就迫使我们去开发非常高效的算力。我们之所以成功，是因为我们拥有低功耗下的这种计算密度，那为什么不把它带到数据中心去呢。所以你看，世界上没有足够的电力供所有这些数据中心使用，这个限制因素正是你切入、找到自己赛道并参与竞争的地方。

So that's where Qualcomm could be an interesting player there, because our DNA comes from a battery-powered device. We have to design chips that are very efficient in power. And I know we spend a lot of time talking about phones. I like to provide this example from an engineering standpoint. The phone is super challenging because you have GPUs that can do ray tracing games. You have neuroprocessing units doing multi-billion-parameter AI at the same time. Now you blast gigabit of data speed to the base station. All of those things happen, and the device has to fit in your pocket, cannot get hot. Cannot get hot. You're going to touch your face with that device. And the battery has to last all day. It doesn't matter how much computing power. As engineers, we get a budget: the battery is going to be this size and has to last all day. So that forces the development of very power-efficient computing. We become successful because we have this compute density with low power, and why not take that to the data centers. So you see, the limiting factor that there's not enough electricity out in the world for all these data centers is really the place for you to come in and find your own swim lane and to compete.

百分之百。这很有意思，因为我们看到，大量注意力都集中到了这些数据中心上，对吧？对 AI 算力的需求也在造成智能手机内存短缺，对吧？而且你警告过这会拖累你们的业务。那边到底发生了什么？

100%. Which is interesting because we are seeing, I mean, just so much attention going onto these data centers, right? The demand for AI computing power is also creating a shortage of memory for smartphones, right? And you've warned that's going to hinder your business. What's happening there?

这很有意思，因为我们是少数几家开始谈论这件事的公司之一，但考虑到规模，手机有点像煤矿里的金丝雀。但情况是这样的：数据中心正在大规模建设，而目前现有的方案用的是这种叫 HBM 的内存，也就是面向数据中心的高带宽内存。所以今天很多内存公司都优先把工厂产能用于为数据中心生产 HBM。世界上原本有这么多内存产能。随着其中很大一部分流向数据中心，剩下的才可供消费电子使用。而消费电子设备里的内存有两种：一种是 DRAM，就是你买手机和 PC 时，它有多少内存，8 GB、16 GB；另一种是存储，比如闪存或 NAND。我们看到的是，可供消费电子设备使用的内存量现在比去年还少。所以我在上次财报电话会上说的就是：在情况缓解之前，内存将决定市场的规模。我认为我们已经看到这影响了手机。有意思的是，因为公司刚经历了 40 年来创纪录的一个季度，我们创下了公司营收纪录。移动需求很高，人们想买新手机，但这将取决于内存的供应量。手机不能幸免，PC 不能幸免，游戏主机也不能幸免。所有消费电子产品都要在这场数据中心建设潮中争夺内存。

It's interesting because we're kind of one of the few companies to start talking about this, but given the scale, phone is a little bit of the canary in the coal mine. But here's what's happening: there's this massive buildout on data center, and the existing solutions today are using this memory called HBM, which is high bandwidth memory for data centers. So a lot of the memory companies today have prioritized the buildout of using their factory capacities to build HBM for the data center. There was this much memory capacity in the world. As a lot of those go to the data center, whatever was left became available for consumer electronics. And you see memory in consumer electronic devices in two ways: you have your DRAM, that's when you buy a phone and a PC, how much memory is it going to have, 8 gigabyte, 16 gigabyte, and then you have your storage, which is like flash or NAND. And what we saw is the amount of memory that was available to consumer electronic devices is now less than was available last year. So I think what I just said in the last earnings call: until the situation resolves, memory will define the size of the market. And I think we've seen that impacted in phones. And what's kind of interesting is because we just had a record quarter in the 40 years of the company, we had the record revenue quarter for the company. Revenue record. Mobile demand is high, people want to buy new phones, but it's just going to be defined by the availability of memory. Phones not immune to this. PCs not immune to this. Gaming consoles are not immune to this. Everything that is consumer electronic is going to be fighting for memory with this data center buildout.

外面确实很狂热，对吧？你觉得围绕 AI 的兴奋和热潮，与 25 年前互联网时代上一轮大繁荣然后泡沫破裂，有什么相似之处吗？

It's frothy out there, right? Do you see any parallels between what's going on around the excitement, the boom around AI, with the kind of the last big boom and then bust of the dot era 25 years ago?

所以我认为两者之间的第一个比较是：如果你回到 2000 年，看看当时人们如何想象互联网会是什么样，再看看现在的 2026 年，26 年之后，我们大概可以毫无疑问地说，今天的互联网远比 2000 年时人们以为的要大得多。他们在 2000 年时怎么想都不够大胆，今天比那还要大。那么，它是在一年内全部发生的吗？不，不是。它经历了一个过程才走到那一步。所以我认为 AI 也是一样。正如我之前说的，AI 算是这种新型计算，它将改变一切。从长远看，它可能比人们今天想象的还要大。我们多快能到那一步，我们会不会保持那个速度？会加速吗？会暂停吗？这真的很难预测。做这种比较时还有一件有意思的事，当年——我知道这会暴露我的年龄，对吧，我做这行已经很久了——

So I think the first comparison between the two: if you go back to the year 2000 and you look at how people imagined at the time what the internet would be, and you look right now in 2026, 26 years later, we can probably say without any doubt the internet today is way bigger than people thought the internet would be in 2000. Whatever they thought in 2000, they're not thinking big enough; it's bigger today. Now, did it happen all in one year? No, it did not. It went through a process to get to that. So I think it's the same with AI. As I said before, AI is kind of this new form of computing, it's going to change everything. Probably in the long run bigger than what people think it is today. How fast we get there and are we going to keep going at that speed? Are we going to accelerate? Are we going to take a pause? It's really hard to predict. The other thing that is interesting when you make those comparisons back then—and I know this is going to show my age, right, I've been doing this for a while—

你在公司是从什么时候开始的，1995 年？你一开始是工程师。

You've been at the company since what, 1995? You started as an engineer.

1995 年。我 1995 年以工程师身份加入，实际上在公司已经 30 年了。到 2026 年就 31 年了。但如果你回到那个时候，社交领域的赢家——当时并不清楚。赢家不是 MySpace，是 Facebook，是 Instagram。赢得地图领域的不是 MapQuest，是 Google Maps。所以我们就看会发生什么吧。我认为我们还处在非常早期的阶段。现在确实，AI 是一项革命性技术，它可能只会越来越大。现在让我给你一个不同的视角。

1995. I joined as an engineer in 1995, been at the company for 30 years actually. It's going to be 31 in 2026. But if you go back at that time, the winners of social—it wasn't very clear. It wasn't MySpace, it was Facebook. It was Instagram. It wasn't MapQuest that won the map space, it was Google Maps. So we'll see what happens. I think we're in the very early days. Now it is true that AI is a revolutionary technology and it's probably only going to get bigger. Now let me give you a different perspective.

好。

Yeah.

你有这么多钱投入到 AI 的创造中，但最终人们想靠它赚钱。你必须让所有投资有回报，而那就是你从训练转向我们所说的推理的时候。推理——没错——一旦你已经开发出 AI，你开始提问，AI 开始回答你的问题或为你做事并生成 token，那就叫推理。你这时才把 AI 投入生产。训练，是你创造了一个 AI；推理，是你把它投入生产。那才是你赚钱的地方。所以当你开始用推理来运营算力时，衡量指标与你用于训练的那些非常非常不同：运营这套算力要花我多少钱，我的总拥有成本是多少，我用多少电，这台计算机多少钱，架构内存是多少。而且我们说过——我想那是在它流行之前——推理未来的架构很可能是后 GPU 架构，数据中心将被拆解成不同的功能，你会努力找到最高效的计算机来做那件事，因为现在你做的是运营算力的生意。我认为那大概会是下一个周期。

You have all this money going to creation of AI, but eventually people want to make money on it. You're going to have to have a return on all the investment, and that's when you move from training to what we call inferencing. Inferencing—exactly—once you already develop the AI and you start asking questions and AI starts answering your questions or doing things for you and generating tokens, that's called inferencing. You just now put the AI into production. Training, you created an AI; inference, you put it into production. That's where you're going to get paid. So when you start operating the computing with inference, the metrics are very, very different than the ones you use for training: how much money does it cost me to operate this compute, what is my total cost of ownership, how much power do I use, how much does this computer cost, how much is the architecture memory. And we said—I think before it was popular—that the architecture for the future of inference is probably going to be the post-GPU architecture, where the data center is going to get disaggregated into different functions and you're going to try to find the most efficient computer to do that, because now you're in the business of operating that compute. And I think that's going to be probably the next cycle.

我们即将进入数据中心推理的周期，而其中各参与者的机会可能会有些不同。

We're about to enter the cycle of inferencing in the data center, and the opportunities for the players may be a little bit different.

是的。我是说，我觉得你点到了做创业者那种有趣又令人兴奋的地方。硅谷令人兴奋之处在于，一旦出现新想法，大量资金就会涌向它。这些钱用得可能并不高效。它是为了增长，为了试验，为了更快地做大。但到了某个节点，你必须转向，必须工业化，必须变现。我想你说的就是这个意思。

Yeah. I mean, I think you're getting at this kind of interesting, exciting thing about being an entrepreneur. The exciting thing about Silicon Valley is when there's a new idea, a lot of money goes to it. It's probably not used really efficiently. It's for growing. It's for experimenting, to try to get big faster. But at a certain point you've got to pivot and you've got to industrialize. You've got to monetize. I think that's what you're getting at.

那是个机会。

That's an opportunity.

我确实想知道，当你回看 2000 年互联网泡沫时期，那里有没有什么今天可以借鉴的东西，用来思考如何在繁荣时代成为赢家？听起来在你看来，关键就是尽快想清楚如何变现。

I do wonder, when you look at the bubble of 2000 around the dot-com era, is there anything there that you can use today to think about how to be a winner in a boom era? And it sounds like to you it's figuring out how to monetize as quickly as possible.

没错。你看，毫无疑问，2000 年建起来的一切最终都被充分利用了。我认为这次也会如此。没有什么会被浪费。你建了算力，你就会用起来；你建了宽带连接，你就会用起来；你建了那些网络，你就会用起来。但当你开始思考不同的商业模式、人们将如何使用 AI 时，最终你会需要最具竞争力的运营者，对吧？这些东西会变得重要。现在大家还都在忙着做产品，这些还不重要，但接下来，一个人的产品要跟另一个人的产品竞争，靠的是什么？靠的是能力和运营成本。我认为有些公司会比另一些公司准备得更充分。我们在互联网的历史上见过这一点。

That is correct. Look, there's no question that everything that got built in 2000 got fully utilized. And I think that's going to be the case. Nothing's going to go to waste. I think you build computing, you're going to use it; you build broadband connectivity, you're going to use it; you build those networks, you're going to use it. But as you start to think about the different business models, how people are going to use AI, eventually you're going to have the most competitive operator, right? And those things will matter. They don't matter right now if everybody's trying to create a product, but then what is going to be somebody's product competing against somebody else's product? It's going to be the capability and the cost to operate. And I think some companies are going to be better prepared than others to deal with that. We saw that in the history of the internet.

好的，Cristiano，感谢你今天参与我们的节目。这是一次很棒的对话。

Well, Cristiano, thank you for joining us today. This has been a great conversation.

是我的荣幸。也感谢你抽出时间。

My pleasure. Thank you for your time as well.