高通 CEO 克里斯蒂亚诺·安蒙谈智能体 AI 设备的崛起

我有件特别激动的事要告诉大家。不，不是我周末拿到的宝可梦卡牌——那些很棒。而是我们身处 CNBC 伦敦办公室的全新演播室。我们有炫酷的灯光、新摄像机和很棒的团队。有点热，但很棒。这就是开场。欢迎来到 Tech Download。我是 Arjun Kharpal，CNBC 的高级科技记者，今天我们请来了一位非常有趣的嘉宾：高通 CEO 克里斯蒂亚诺·安蒙。如果你不了解高通，简单介绍一下：他们是全球最大的半导体公司之一。他们设计所谓的系统级芯片——你可能听过骁龙品牌下的 SoC。这些芯片用于智能手机，上面集成了许多不同组件。它们为全球大量智能手机提供动力，也用于笔记本电脑等其他设备。因此，这位 CEO 对消费电子以及 AI 时代该行业的走向有很好的洞察。我们将讨论所有事情，即将到来的设备热潮，他还会透露一些尚未发布的设备。他将谈论 AI 如何改变我们与智能手机这一核心设备的关系，以及 AI 对高通未来设备芯片设计思路的意义。这次对话以及未来节目中你会经常听到的一个词是智能体式 AI。这个想法是，我们从聊天机器人开始，我们提问，它回答，但实际上这些 AI 公司正试图更进一步。他们希望这些 AI 聊天机器人像助手一样，实际上像智能体一样代表我们工作，有时自主执行任务。然后你会听到高通 CEO 介绍一些使 AI 设备成为可能的其他创新。其中之一是 MCP，即模型上下文协议。这实际上是一个开源系统，让 AI 助手能够与其他数据和其它应用连接和通信。如果你想让智能体在另一个应用上执行任务，这很重要。最后你会听到的创新是云与边缘处理。很多时候，当我们使用这些 AI 聊天机器人时，你的数据被发送到云端处理，然后返回你的设备，你得到答案。但许多设备制造商正在思考如何直接在设备上运行这些 AI 模型和 AI 处理，这可能意味着更快、更安全，这也是高通在其芯片上正在做的事情。我很兴奋为你带来与高通 CEO 克里斯蒂亚诺·安蒙的对话。欢迎来到 Tech Download。

I've got something so exciting to tell you all. And no, it wasn't the Pokémon cards that I got this weekend, which were awesome. It's the fact that we are in a brand new studio out of CNBC's London offices. We got the fancy lights, some new cameras, and a great crew. It is a bit hot, but it is wonderful. That's the intro. So, I guess welcome to the Tech Download. I'm Arjun Kharpal, senior tech correspondent here at CNBC, and we're kicking off with a super interesting guest, Cristiano Amon, who is the CEO of Qualcomm. Now, if you don't know Qualcomm, just a quick rundown: they are one of the biggest semiconductor companies in the world. They design what are known as system on chips—SoCs, you may have heard this term under the Snapdragon brand name. These are chips that go into smartphones and have loads of different components on them. They power loads of smartphones in the world and other kinds of devices as well, such as laptops. So the CEO has a great view into consumer electronics and effectively where this industry is going in the age of AI. So we're going to be talking about all the things, the hype of the devices coming up, and he tells us a little bit about some of those devices that are not yet revealed as well. He's going to talk about how AI is changing our relationship with the core device of the smartphone. He's also going to talk about what AI means for the way Qualcomm needs to start thinking about designing the chips that go into our future devices as well. One of the words you're going to hear a lot of in this conversation and other episodes in the future is the word agentic AI. The idea here is that we started off with this chatbot, this kind of input where we'd ask it questions, it gives us a response, but actually these AI companies are trying to take that a step further. They want these AI chatbots to be like assistants, effectively like agents working on behalf of us, at times autonomously carrying out tasks on our behalf. Then you're going to hear a little bit about some of the other innovations that are making AI devices possible from the Qualcomm CEO. One of those is something called MCP, or Model Context Protocol. This is effectively an open-source system for AI assistants to connect and communicate with other data and also other apps. It's important if you think about it—if you want your agent to be able to go and carry out stuff on another app, for example. And then the final innovation you're going to hear about is something called cloud versus edge processing. So a lot of the times when we're using these AI chatbots, effectively your data is being sent to the cloud, it's processed there, and then comes back to your device, and you effectively get your answer. But a lot of device makers are thinking: how can we run those AI models and those AI processes actually on the device? That could mean it's faster and more secure, and that's something that Qualcomm is working on with its chips as well. I'm excited to bring you this conversation with Cristiano Amon, the CEO of Qualcomm. Welcome to the Tech Download.

克里斯蒂亚诺，非常感谢你来到 Tech Download。

Cristiano, thanks so much for joining me on the Tech Download.

很高兴来到这里。Arjun，期待这次对话。

Very happy to be here. Looking forward to this conversation, Arjun.

克里斯蒂亚诺，今天我们要谈论 AI 设备，鉴于 AI 正普及到如此多的设备中，我们未来与个人电子设备的互动方式将如何演变。但我想首先为观众阐明，当我们谈论 AI 设备、AI 智能手机、AI 笔记本电脑时，这到底意味着什么？

So Cristiano, look, we're going to be talking today about AI devices, the future of how we're going to interact with our personal electronics in the future given how AI is proliferating so many of these devices as well. But I think first it's important for our audience just to lay out when we talk about AI devices, an AI smartphone, an AI laptop, what does that exactly mean?

看，这是一个很好的话题，尤其是因为我们正处在 AI 转型旅程的节点上，我们清楚了那些设备会是什么样，你会做什么。但我要回答你的问题：你仍然会有手机——手机不会消失——但手机会变得好得多。手机将拥有智能体为你操作手机。但还会有新类别的可穿戴设备。我会回答你的问题。我想说那些东西到底意味着什么？我们一直在谈论所谓的个人 AI 设备。随着大语言模型和大型视觉模型像我们一样理解世界、理解我们如何交流，我们与那些靠近感官的设备互动是非常合乎逻辑的——更靠近眼睛、嘴巴、耳朵。然后会有我们穿戴的东西。眼镜很自然，但还会有珠宝、吊坠、胸针，因为交互界面类型与过去的 PC 和智能手机非常不同。那将是一类新设备。它们将成为我们移动体验的一部分。

Look, it's a great topic of conversation especially because we are now getting to the point on this AI transition journey that we have clarity on what those devices are going to be and what you're going to do. But I'm going to answer your question by saying you're still going to have your phone—your phone's not going to go anywhere—but your phone's going to get much better. Your phone is going to have agents that are going to operate the phone for you. But then there's going to be new classes of devices that we're going to wear. And I will answer your question. I want to say what do those things really mean? And we've been talking about something called personal AI devices. As large language models and large visual models understand the world the way we understand it, understand how we communicate, it is very logical that we're going to interact with those devices close to our senses—closer to our eyes, close to our mouth, to our ears. And then there's going to be things that we're going to wear. Glasses are very natural, but there's going to be jewelry. There's going to be pendants. There's going to be pins because the type of interface is very different than what used to be in your PC and your smartphone. And that's going to be a new class of devices. They're going to be part of our mobile experience.

你提到了一些关于大语言模型的有趣观点，这引出了是什么改变了，让如今的设备更加个性化。

You mentioned something interesting about large language models and that goes to the point about what's changed to allow devices these days to be more personal.

好的，这是多项技术的结合，我来为你分解。我认为首先是那些大语言模型变得相当好，而且它们也变得非常专业化——不仅是语言模型，还有视觉模型。你开始听到多模态这样的词，模型将接受不同输入——音频、语言、视觉。你开始听到更小的模型、专家混合模型，它们针对特定功能。你开始听到模型蒸馏——有一个大模型经过训练，然后变成更小的模型。你开始听到这些模型连接到不同的应用。你听到 MCP 协议等技术，模型可以上网连接其他应用、其他云服务。然后这一揽子技术的最后一部分是——我相信你听说过 OpenClaw——那些称为编排器的技术，智能体为你操作电脑、设备，查看你的数据，查看模型，为你完成任务。当你把所有这些放在一起，就为我们开始改变移动体验创造了条件，你开始以与智能体互动的方式与移动设备互动。非常不同。没有键盘。你不需要触摸——你只需要看和说。我认为这开始为我们看到这场因 AI 而发生的移动革命创造了前提条件。

Okay, it's a number of technologies and I'm going to break it down for you. I think the first one is those large language models got pretty good, and they also got very specialized—not only language models but also visual models. You started to hear about words like multimodal, where you'll have models that are going to take different inputs—audio, language, vision. You start hearing about smaller models, mixture of experts models, they're very specific for certain functions. You started to hear about distillation of models—there's a big model, they got trained, and then you became a smaller model. And you start to hear about those models being connected to different applications. You hear about technologies such as MCP protocols that a model can go to the internet and connect to other apps, to other cloud services. And then the last part of this basket of technologies is—I'm sure you heard about OpenClaw—those techniques called orchestrators, agents that go operate your computer, your device for you, going to look at your data, look at the models, and get things done for you. When you put all of this together, you created the condition for us to start changing our mobile experience and you start to interact with mobile devices in the way we're going to interact with agents. Very different. There's no keyboard. You don't need to touch—you just need to see and talk. And I think that starts to create the precondition for us to see this mobile revolution that is happening because of AI.

那么，克里斯蒂亚诺，让我们探讨其中一些。

So, Cristiano, let's tackle some of those.

你谈到了模型，因为我们一直在讨论大型语言模型，这些经过训练的巨大模型支撑着许多人使用的聊天机器人。但也有讨论关于更多这些模型变得更小，以便在设备上运行而不是在云端。而这是你在高通工作的重要部分。目前模型的发展对于支撑你一直在谈论的一些智能体式 AI 体验的未来有多重要？

You talked about the models because we've been discussing a lot about large language models, these huge massive models that have been trained that are then being underpinning these chat bots that so many people use. But there is also discussion about more of these models becoming smaller to run on the device rather than in the cloud. And that's a big part of what you're working on at Qualcomm. How important is the development of models right now in terms of underpinning the future of some of the agentic AI experience which you've been talking about?

看，这非常重要，这些模型将继续发展。但我认为我们现在正达到行业成熟点，开始理解这些模型和计算机将如何协同工作。我要稍微挑衅一下，因为我记得大约两年前或一年前有很多讨论，关于这个模型是在云端运行还是在边缘运行。我能在边缘设备上运行这个在云端运行的完全相同模型吗？但我认为那是错误的讨论。例如，Arjun，如果我问你手机上有多少应用，你可能会告诉我大概超过 200 个应用，整理在文件夹里，如果我问你让我们一个一个看，哪个应用在云端运行，哪个部分在设备上运行，那可能是无用的练习，我知道它们都在做两者，因为如果你把手机设为飞行模式，它就变得不太有用。所以我认为 AI 也在发生同样的事情。AI 只是一种新的计算方式，模型现在正以特定方式发展，有些特定模型将在边缘运行，有些特定模型将在云端运行，它们将作为一个整体协同工作，现在实际上有一个行业术语叫做计算连续体。它将像一个集成的系统，我举个例子。我是什么意思？如果我要与未来的个人 AI 设备互动，比如说智能眼镜，我要像现在与你互动一样互动。我需要即时响应。我可能走在街上说，“这个人是谁？”模型会说，“那是 CNBC 的 Arjun。你认识他。你可能想打个招呼。”我不能等待。我不能只是问问题继续走，等待 token 进来。同样，如果我读菜单说这是什么？你想要即时响应。如果我扫描二维码，请为我支付这个。所以正在发生的是计算自然分解为边缘发生什么，云端发生什么。它们将集成。一个没有另一个就无法运作。我认为这将定义我们如何思考 AI 时代消费电子和设备的未来。我告诉你，这与汽车、机器人发生的情况没有不同。实际上，那些是边缘 AI 的绝佳例子。你知道，它们实际上在机器人中运行，在汽车中运行。

Look, it's really important and those models will continue to be developed. But I think we're getting to the point in the industry now that we're starting to reach maturity of understanding how those models and computers are going to work together. I'm going to be a little bit provocative because I remember there was a lot of discussions about maybe two years ago, one year ago, there was a lot of debate about is this model going to run on the cloud or this model going to run on the edge. Can I run this exact model that is running on the cloud on the devices of the edge? But I think that was the wrong discussion. Think about for example Arjun if I ask you how many apps do you have on your phone and you're probably going to tell me probably over 200 apps organized in folders and if I ask you let's go one by one and see which app run on what portion of the app run on the cloud what portion of the app run on the device it's probably a useless exercise and I know they're all doing both because if you put your phone in airplane mode it becomes not very useful. So I think the same thing is happening with AI. AI is just a new form of doing computing and models are now developing in a way that they're specific that are going to be running on the edge. They're models that are specific that going to be running on the cloud and they're going to go work together as one and there's actually an industry term now called compute continue. It's going to be like one integrated system about this and I'll give you an example. What do I mean? If I am going to interact with my future personal AI device, let's say smart glasses and I'm going to interact the way I'm interacting with you right now. I need to have an instant response. I may be walking on the street and I say like, "Who's this person?" And the model is going to say, "That's Arjun from CNBC. You know him. You probably want to say hi." I cannot wait. I cannot just ask the question keep walking and wait for the tokens to come in. So the same thing if I read a menu and I say what does this say? You want an instant response. If I look up QR code, please pay this for me. So what is happening is the compute is naturally breaking down in what's going to happen on the edge, what's going to happen on the cloud. They're going to be integrated. One is not going to function without the other. And I think that's going to kind of define how we think about your future of consumer electronics and devices in the age of AI. And I just going to tell you it's no different than what happened to cars, what happened to robots. Actually, those are great examples of edge AI. You know, they actually run in the robot and they run into the car.

那么，让我们来谈谈 AI 设备这个术语。它到底是什么意思？嗯，我们显然有智能手机，但在过去几年里，当然，所有这些不同类型的聊天机器人出现了爆炸式增长。我们想到 OpenAI 的 Chat GPT，我们想到 Anthropic 的 Claude，我们谈论 Google 的 Gemini，所有这些现在都有。而且它们变得越来越复杂。嗯，现在设备制造商在想，我们如何将所有这些 AI 集成到所有这些不同类型的电子产品中？最终，这对我们未来使用设备的方式意味着什么？这部分非常重要，因为我们将要谈论即将出现的不同类型的 AI 设备，你会听到提到智能眼镜之类的东西。如果你还没见过，你知道，Meta 生产一些智能眼镜，里面装有摄像头。三星、高通和 Google 都在研发智能眼镜。有很多公司关注这个领域。许多人认为这可能是 AI 设备的新前沿。我们在这里还要提到的另一件事是 AI 别针。现在，你可能没见过这类设备，因为它们在 2024 年曾短暂流行过。是一家叫 Humane 的公司将其中一款推向市场。它基本上是一个小方块，你把它附在衣服上。它有一个摄像头，没有屏幕，但你可以和它说话，它给你答案，还会将文本投影到表面，比如你的手或桌子等。它配备了 AI。这个项目实际上失败了。它得到了一些相当差的评价，你可以读到所有关于它当时没有起飞的原因。但它让我们一窥设备公司对 AI 世界中设备未来的思考方式。我们现在开始看到越来越多的公司，非传统设备公司关注消费电子。一个例子是 OpenAI，Chat GPT 的制造商，它聘请了著名的苹果设计师 Jony Ive 来开发一款设备。所以无论何时揭晓，那都将非常有趣。嗯，但它也向你展示了进入消费电子市场的新入口，以及我们可能获得前所未见的新风格设备的方式。

So, let's get into this term AI devices. What exactly it means? Well, we've obviously got the smartphones, but over the past few years, there's been an explosion, of course, of all these different types of chat bots. We're thinking OpenAI's Chat GPT, we're thinking Anthropic's Claude, we're talking about Google's Gemini, all of these now. And they're getting much more sophisticated. Um, and now device makers are thinking, how can we integrate all this AI into all of these different types of electronics? And ultimately, what does it mean for the way we're going to be using devices going forward? This part's really important because we're about to talk about different types of AI devices that are coming up and you're going to hear mention of things like smart glasses. If you haven't seen those, you know, Meta produces some smart glasses which have sort of a camera in them. Samsung, Qualcomm, and Google are all working on smart glasses. There's lots of companies looking at this space. Many thinking this could be a new frontier for AI devices as well. The other thing we're going to mention here is AI pins. Now, you may have not have seen these kind of devices cuz they were popular for a very fleeting moment in 2024. And it was a company called Humane, which bought one of these to market. It was basically a small square thing that you attached to your clothing. It had a camera in it, no screen, but you could talk to it and it give you answers and would also project text onto a surface such as your hand or a table or something. And it was equipped with AI. And the project effectively flopped there. It got some pretty bad reviews and you can read all the reasons why it didn't take off at that time. But it was a glimpse into the way device companies were thinking about the future of devices in a world of AI. And we are now starting to see more and more companies, non-traditional device companies looking at consumer electronics. One example is OpenAI, the maker of Chat GPT, which hired the famous Apple designer Jony Ive to work on a device. So that's going to be very interesting whenever that's revealed. Um, but it also shows you the way you are getting new entrance into this consumer electronic market and the way that we may be getting new styles of devices that we haven't seen before.

现在一个巨大的流行词，我们已经提到过，就是智能体式 AI 这个概念。现在当我听到它被谈论时，它的乌托邦是有效地这些 AI 系统将能够代表我们自主做事，就像最智能的个人助理一样。嗯，当涉及到消费设备上的智能体式体验时，在你看来那是什么样子的？

One word that is a huge buzzword right now, we've mentioned it already, is this idea of agentic AI. Now when I hear it being spoken about the kind of utopia of it is effectively these AI systems are going to be able to autonomously do things on our behalf like the most smart personal assistant out there. Um when it comes to agentic experience on consumer devices what does that look like in your view?

好的，这实际上是我喜欢回答的问题之一，因为你知道，尽管我们一直在改变高通，我们正在成为一家更加多元化的公司，我们仍然是最大的移动技术提供商之一，现在看到这些智能体发生了什么以及手机发生了什么，真是令人着迷。当我们谈论智能体时，我有点确信 2026 年将是智能体之年。我们现在开始看到这规模化，我认为当 OpenClaw 推出时，这些东西在互联网上变得非常流行并病毒式传播，你现在有这个东西安装在你的电脑上，你让智能体为你做事。

Okay, that's actually one of the questions I love answering because you know even though we have been changing the Qualcomm and we're becoming a much more diversified company, we're still one of the biggest providers of mobile technology and it's fascinating now to see what's happened with those agents and what's happened with the mobile phone. When we talk about agents and I'm kind of convinced 2026 going to be the year of agents. We're starting to see this now getting scale and those things became really popular and went viral I think on the internet when OpenClaw was launched and you have now this thing that you install on your computer and you ask the agent to do things for you.

你创建智能体，它进入你的数据，替你做事，替你操作你的机器，上网。然后你看到发生了什么——我记得，人们跑去买电脑。很多人买 Mac mini，都卖断货了。他们把它放进办公室，装上 OpenClaw，让那东西疯狂跑，消耗大量 token，登录他们的 Anthropic 账户和 OpenAI 账户去提交所有 token。这展示了我们未来与电脑交互的一条路径。我们还是会用我们的电脑，但智能体会代表我们使用电脑，替你做事。而在移动环境里，你不可能背个背包、装块汽车电瓶、塞台 Mac mini，再跟手机一起带着走。所以这些事会发生在手机里。我认为这也会定义手机体验，毕竟有几十亿人用手机。所以现在我们清楚了什么是 AI 手机、什么是 AI PC，我来给你拆解一下。

You create agents and it goes to your data, does things on your behalf, operates your machine for you, goes to the internet. And then what you saw happen — I remember, people were going in and buying computers. A lot of people were buying Mac minis; they were sold out. They put it into their office, installed OpenClaw, and let the thing run, consuming tokens like crazy, going to their Anthropic accounts and the OpenAI accounts to submit all the tokens. Now what that did is it showed a path for how we're going to interact with our computers. We are going to use our computers, but the agents are going to use our computers on our behalf and do things for you. Now, in a mobile environment, there's no way you're going to get a backpack, put a car battery, put a Mac mini and carry that alongside your phone. So those things are going to happen in the phone. And I think that's also going to define the phone experience, given that billions of people wear phones. So now we have clarity on what an AI phone is and what an AI PC is, and I'm going to break that down for you.

会发生什么？你的手机将拥有双重人格。手机仍然是你的手机。手机不会消失。你从口袋里掏出手机，打开应用，做那些事。但手机也会拥有那些编排器、那些 claw。从现在到夏天，你会看到每个操作系统厂商都在谈他们的系统内置编排器和 claw。你也能让智能体替你做事。它会代表你操作你的手机，进入你的应用，你给它指令。

What's going to happen? Your phone is going to have dual personas. Your phone will continue to be your phone. The phone's not going to go anywhere. You pull your phone out of your pocket, you go to your apps, and you do those things. But the phone is also going to have those orchestrators, those claws. Between now and summer, you're going to see every OS vendor talking about their OS having an orchestrator and a claw. And you'll be able to get agents to do things for you. It's going to operate your phone on your behalf. It's going to go to your apps. You're going to give them instructions.

这个变化如此深刻，哪怕我暂时跳出手机，只讲讲软件公司发生了什么——整个 SaaS 行业——现在软件公司开发软件，既要供人使用、有人机界面，也要供智能体使用，让智能体进入他们的软件做事。这件事的走向非常迷人。比如，有些事你自己做，但也会由智能体替你做。我见过一些例子：某人有个会议，然后又来一个邀请，它会说：“Cristiano，你有一个和 CNBC 的 Arjun 的播客，这个要不要优先于另一个会议？”我会说：“当然优先。”然后它说：“好，我可以帮你重新安排另一个会议吗？”你说：“可以。”如果需要跟人沟通，智能体会打电话给对方，会发邮件，会根据你的日历想办法重新安排。慢慢地我们会习惯这些新体验，久而久之你的手机会两者兼顾：既由智能体操作，也由你操作。

This change is so profound that even if I step for a moment outside the phone and just tell you what happened with software companies — the whole SaaS industry — now software companies develop their software for a human to utilize, with a human interface, but also for an agent to utilize, to go into their software and do things. And the way this is going to play out is fascinating. For example, you're going to have the things that you do yourself, but it's going to be the agent doing things for you. I saw some examples: somebody has a meeting, and then another invitation comes up, and it says, "Cristiano, you have this podcast with Arjun from CNBC. Would this take priority over this other meeting?" I'm going to say, "Of course it takes priority." Then it says, "Okay, can I reschedule this other meeting for you?" And you say, "Yes." And the agent will call the person if it needs to talk to a person, will send an email, will figure out how to reschedule things for you based on your calendar. And slowly we're going to get used to those new experiences, and over time your phone is going to be doing both. It's going to be operated by the agent and it's going to be operated by you.

我就在等一个能帮我清空收件箱的智能体。这是最重要的事。不过 Cristiano，你看这个路线图，你说 2026 年将是智能体之年。过去两年很多智能手机公司都在谈把 AI 整合进设备，但实际上没有一个是你在描述的这种智能体式体验。你相信这会在 2026 年发生吗？到什么程度？然后展望未来，和你合作的所有设备伙伴，让它变得更先进的路线图是什么？

I'm just waiting for an agent that can clear my inboxes for me. That's the biggest thing. But Cristiano, when you look at the road map for this, you said 2026 is going to be the year of the agents. It feels like the last two years a lot of smartphone companies have been talking about AI integrating into their devices, but actually none of it has been this kind of agentic type of experience you're describing. Do you believe that happens in 2026? And to what level? And then as you look forward with all the device partners you work with, what's the road map for that becoming more advanced?

很有意思。如果你记得从功能机到智能机的转变——那是巨大的转变，手机变成了电脑——当时发生的是，在智能机早期，非常早期，你有的应用都是 OEM 自带的。我确定你买 iPhone 时它自带 iTunes，自带那个可以移动鼓和乐器、随便玩的 app。但最终 App Store 出现了，开发者开始写各种各样的应用。我现在可以告诉你，你今天在手机上体验的绝大多数东西，不是来自操作系统本身，不是来自 OEM，而是来自开发者。它是一个平台，它是一台电脑。智能手机是人类有史以来创造的最大的平台。

It's interesting. If you remember the transition from a feature phone to a smartphone — which is a huge transition, like the phone became a computer — what happened is, in the early days of the smartphone, the very early days, the apps you had were the apps that came from the OEM. I'm sure when you buy your iPhone it comes with iTunes, it comes with that app where you can move drums and instruments and play around. But eventually what happened is the App Store became available and developers started to write all sorts of different apps. And I can tell you right now, the absolute majority of the experience you have on your phone today didn't come from the OS itself, didn't come from the OEM, but actually came from developers. It is a platform, it is a computer. The smartphone is the largest platform ever created by mankind.

现在让我们对照一下 AI 领域发生的事。我们看到的第一个应用是聊天机器人。人们进去问问题，得到答案。然后开始做更高级的事——帮我总结这个。接着很多 OEM 说，我怎么把 AI 用到相机上？怎么把 AI 用到这个那个功能上？再混入一些可用性。有些东西非常有用，有些还很早期。但后来发生了一件事：那些编排器和那些智能体出现了，然后一切开始改变，因为现在 AI 可以变得非常有用，可以替你做事。我认为我们正处在这个拐点上，所以我说 26 年是这些事情开始清晰、开始随智能体规模化的一年。以至于从现在到夏天，你会看到每个操作系统厂商都在谈如何把编排器内置进系统，因为智能体就是新的应用。那些编排器成为新的应用商店。我们会看到新旧在同一设备上共存，因为手机你只带一台，不会带两台。然后我们会看到另一些设备类别，它们纯粹是智能体，没有历史包袱——比如智能眼镜、个人 AI 设备、别针。我觉得这就是我喜欢做这个类比的原因，因为它从 OEM 或操作系统如何添加第一方 AI 能力，走向了无穷的可能性——人们创建自己的智能体，智能体开始代表你操作设备。

Now let's draw a parallel with what we saw in AI. The first application we saw was the chatbot. People go in, they ask a question, they get an answer. Then they started to do more advanced things — summarize this for me. And then a lot of the OEMs said, how can I apply AI to my camera? How can I apply AI to this different functionality? And kind of mix in usability. Some things were very useful, some things were early days. But then something happened: those orchestrators and those agents happened, and then everything started to change, because now AI can become very useful and can do things for you. I think we are at that inflection point, and that's why I said '26 is the year those things start to be clear and start to get scale with those agents. So much so that between now and summer you're going to see every OS vendor talking about how they're going to have those orchestrators built into the OS, because agents are the new apps. Those orchestrators become the new app store. And we're going to see the coexistence of the old and the new in the same device, because in the case of a phone you only carry one device, you don't carry two. And then we're going to see some other classes of devices which are just agents, no legacy — like smart glasses, personal AI devices, and pins. And I think that's the reason I like to draw this parallel, because it went from how an OEM or an OS will add first-party AI capabilities to an endless, I think, set of possibilities as people create their own agents and agents start to operate devices on your behalf.

我们待会儿会聊那些新设备。但基于你刚才的回答，我的第一个问题是：这是否意味着应用已死？

We're going to talk about those new types of devices in a sec. But my first question on the back of that answer is: does this mean that apps are dead?

它们没有死，但应用会改变。绝对会，应用会改变。我要回到两年前说过的话。我们都有银行应用——你有银行应用，你选了你的银行，你就有那个 app。所以你有开发者。想象一下，大多数银行用户会喜欢的 app 该是什么样？用什么颜色？要适配你设备的屏幕。它会有操作说明，你打开 app，它会说，这是我的支票账户，这是我在银行里有多少钱。

They're not dead, but apps are going to change. Absolutely, apps are going to change. I'm going to go back to something I said two years ago. All of us have a banking app — you have a banking app, you have your bank of choice and you have the app. So you have a developer. Imagine how the app is going to be that most of the bank users are going to like. What are the colors? It's going to fit into a screen of your device. It's going to have a manual, and you're going to start the app. It's going to say, this is my checking account, this is how much money I have in the bank.

这是我的最近交易。这是我的信用卡。这是我的投资。而它的实现方式将会是大多数人喜欢的方式。现在，你手机里的那个应用，你手机里的银行应用有你的凭证。所以它可以去银行，作为你的代理进行身份验证，并把信息带给你。现在你可以有一个智能体，你可以把银行凭证给它，而应用体验将会是一样的。我来描述给你听。你会说：“我的支票账户里有多少钱？”它会说：“你有这么多钱。这是你最近的三笔交易。你能帮我画个图吗？”然后如果你说我想把这些信息显示在我的屏幕上，但我不喜欢这个应用的颜色，我想要不同的颜色，它就会为你渲染。所以这就是区别。这就是应用如何演变成智能体的根本区别，它们会变得更强大、更定制化，并且会集成到一切事物中。你可以在餐厅里，看着账单上的二维码。我会说：“帮我付这个。请从我的支票账户扣款，完成后通知我。”这和你自己打开应用走流程是不同的。

This is my last transactions. This is my credit card. This is my investments. And the way it's going to be done is going to be the way that most people are going to like it. Now, that app in your phone, your banking app in your phone has your credentials. So it can go to the bank, authenticate itself as a proxy of you and get that information to you. Now you can have an agent that you can give the credentials to your bank and the app experience is going to be the same. I'm going to describe it to you. You're going to say, "How much do I have in my checking account?" And it's going to say, "This is how much you have. Those are my last three transactions. Can you please plot a graph for me about this?" And then if you said I want this information to be displayed to me on my screen but no I don't like the colors of the app. I want it to be different colors and it will render for you. So this is the difference. This is the fundamental difference in how apps are going to evolve into agents and they're going to become way more powerful, way more bespoke and they're going to be integrated into everything. You can be in a restaurant and you're going to look at a QR code with the bill. I'm going to say, "Pay this for me. Please take it off my checking account, notify me when I'm done." That's going to be different than you open the app and go through the process yourself.

到目前为止，你在本期节目中说过的许多重要事情之一是，你相信智能手机仍然至关重要。你认为它仍将是智能体式 AI 世界的核心吗？你预计我们与智能手机的关系会如何变化？

And among many of the important things you said through this episode so far is that you believe the smartphone is still going to be critical. Do you believe this is still going to be central to the world of Agentic AI and how do you expect our relationship with our smartphone to change?

是也不是。让我稍微退一步，谈谈我们的数字生活，尤其是从消费者的角度，将如何运作。我们今天正进入一个世界，我们数字生活的绝对中心是智能手机。我们最不可分割的设备。一切都在智能手机周围。就像智能手机是太阳，每颗行星都围绕智能手机。以至于如果你看看所谓的智能可穿戴设备，那些智能可穿戴设备的工作是扩展手机的功能给你。举个简单的例子，智能手表。它会给你时间。它会将传感器数据从你发送到手机，然后我们从手机获取通知回传给你。它是如此集成，以至于你看到行业实际上是如何运作的。同一家手机公司实际上是在做很多可穿戴设备及其生态系统。现在未来有一个巨大的变化，而且是非常近的未来，尤其是当智能体开始规模化时。你数字体验的中心是你选择的智能体，或者你选择的多个智能体，现在一切都在智能体周围。手机在智能体周围。新的设备类别，我知道我们马上会谈到，也将在智能体周围。智能体将是理解人类意图并为你做事的那一个。所以我认为，对于你的数字生活，重心发生了转移。这将改变我们设备的关系。然后，我来回答你的问题。手机会消失吗？绝对不会。手机是非常有用的设备。智能手机的规模是巨大的。但我们将以稍微不同的方式使用手机。记得智能手机出现时，有人说笔记本电脑已死。绝对不是。笔记本电脑从未死过。笔记本电脑仍然在这里，仍然强大。但你过去在笔记本电脑上工作的方式改变了，工作负载转移了。我喜欢举这个例子，因为对我们大多数人来说很容易理解。我们开始在笔记本电脑上做电子商务。我们开始在笔记本电脑上做社交媒体。那是你登录 Facebook 账户的地方。现在，大多数电子商务和社交媒体都是在移动设备上完成的。所以，你的手机仍然会在那里，但某些工作负载将转移到新的设备类别，因为智能体与你互动会更好，而且新的设备类别对你也会更好。所以，我认为总结一下，你的手机仍然会在那里。你的手机将会改变。它将拥有更多的计算能力，因为你现在要做手机的事情，但你将运行智能体来为你操作手机。然后你在手机上的体验，其中一些体验将迁移到新设备，这些设备更容易用人类语言和人类视觉与你互动。

Yes and no. So, let me step back a little bit and talk about how our digital life, especially from a consumer standpoint, is going to work. We are coming today in the present time into a world where the absolute center of our digital life is our smartphone. Our most inseparable device. Everything is around the smartphone. Like the smartphone is like the sun. Every planet is around the smartphone. So much so that if you look at what was defined as smart wearables, those smart wearables' job was to extend the functionality of the phone to you. Let's go to a simple thing as a smart watch. It will give you the time. It will send sensor data from you to the phone and we'll get notifications from the phone back to you. And it's so integrated that you see how the industry actually works. The same phone company is actually the one that's doing a lot of the wearables and its ecosystem. There's a big change now in the future and it's a very near future, especially as agents start to get scale. The center of your digital experience is the agent that you choose or the multiple agents that you choose and now everything is around the agent. The phone is around the agent. The new classes of devices, I know we're going to talk about that in a second, are going to be around the agent as well. And the agent will be the one that will understand human intentions and will do things for you. So there is a shift in what the center of gravity is, I think, for your digital life. And that's going to change the relationship of our devices. And with that, I'm going to answer your question. Will the phone disappear? Absolutely not. The phone is an incredibly useful device. The scale of the smartphone is tremendous. But we're going to use our phones a little bit differently. Remember when the smartphone arrived, some people said the laptop is dead. Absolutely not. The laptop was never dead. The laptop's still here and is still strong. But the way you used to work on your laptop changed, workloads shift. I like to provide this example because it's easy for most of us to understand. We started doing e-commerce on a laptop. We started doing social media on a laptop. That's where you go to your Facebook account. Now, most of e-commerce and social media is done on a mobile device. So, your phone's still going to be there, but certain workloads are going to shift to new classes of devices because it's going to be better for the agent to interact with you and there's a new class of devices going to be better for you as well. So, I think in summary, your phone's still going to be there. Your phone is going to change. It's going to have a lot more computing because you're going to now do your phone things, but you're going to run agents that are going to operate the phone for you. And then your experience that you have on the phone, some of those experiences are going to migrate to new devices that are easier to interact with you in human language and human vision.

那么让我们谈谈那些设备，克里斯蒂亚诺，因为你在这次对话中提到了一些，也许我们应该从眼镜开始，因为你知道这是你一直在与谷歌和三星合作的项目，我们知道三星今年也将推出这些产品。你认为眼镜会成为下一个最重要的设备吗？

So let's talk about those devices then Cristiano because you mentioned a few through this conversation and maybe we should start with glasses because you know this is a project you've been working on with Google and Samsung and we know Samsung are going to launch these this year as well. What do you think glasses are going to be the next most important device?

100%。看，数字说明一切。我们现在每年出货量达到数千万副智能眼镜。我们预计在几年内，这可能达到数亿副眼镜，并可能变得和智能手机一样大。智能手机，你知道，每年大约 12 亿部。它成为明显赢家的原因是，当你戴眼镜时，首先人类戴眼镜，我们中有些人戴，对吧？所以你已经有一个规模化的眼镜行业。第二件事是眼镜非常适合安装摄像头，因为当你转头时，你看到什么摄像头就看到什么。所以看到我所看到的。它靠近你的耳朵可以放扬声器。它靠近你的嘴巴可以放麦克风。你将能够读取你所读的内容。所以眼镜变得非常自然，让你与多模态模型互动，我认为，人类语言和视觉。然后你很容易理解某些工作负载，那些与你互动摩擦很低的设备，可以轻松迁移到眼镜上，我们开始规模化，我回到我之前说的，关于可穿戴设备在智能手机为中心的世界中的角色，以及在智能体为中心的世界中可穿戴设备的角色。这很迷人。当第一副智能眼镜推出时，它的工作是拍照并发布到 Instagram 或从 Instagram 生成故事。它基本上是你智能手机相机的延伸，以及听音乐的能力。就这样。但现在有了智能体，每周你都会得到一个新功能。

100%. Look, the numbers speak for themselves. We have now shipping on the order of multiple tens of millions of smart glasses per year. We have line of sight in a couple years this could become into the order of hundreds of millions of glasses and could become as big as smartphones. Smartphones is, for you to know, it's about 1.2 billion phones a year. And the reason that is becoming a clear winner is because when you wear glasses, first humans wear glasses, some of us do, right? And it's so you already have an industry at scale for glasses. The second thing is glasses is very natural for you to put cameras because when you turn your head whatever you're seeing the camera is seeing. So see what I see. It's close to your ears for speakers. It's close to your mouth for microphone. You will be able to read what you read. So glasses become very natural for you to interact with a model with multimodality, I think, human language and vision. And then it's become very easy for you to understand how certain workloads that are very low friction for you to interact with those kind of devices can easily migrate to a glass and we're starting to get scale and I go back to what I said earlier about the role of a wearable in the world where the smartphone is in the center and the role of a wearable where the agent is in the center. It's fascinating. When the first smart glasses was launched, the job was to take a picture and post it on Instagram or generate a story from Instagram. It was basically an extension of the camera of your smartphone and also the ability for you to listen to music. That was it. But now that you have agents, every week you get a new feature.

每周你都会接入另一家云，它的发展方式和手机完全独立。我觉得这正是我们描述它的方式。而且如你所知，我非常看好眼镜。目前早期这一波眼镜主要聚焦于把摄像头放进去，但有很多讨论围绕让眼镜拥有屏幕——增强现实风格、混合现实风格的显示——以及用手部动作和其他各种方式来控制这些眼镜。你认为未来几年智能眼镜会迎来哪些重大创新？

Every week you get an integration with another cloud, and it develops completely independently from how the phone develops. And I think that's exactly how we describe it. And I am very bullish on glasses, as you know. Right now the early wave of glasses has focused on having the camera in there, but there's a lot of discussion around making the glasses have screens — augmented reality style, mixed reality style displays — as well as the ability to control these glasses with movement of the hand and various other things. What are some of the big innovations you think are coming to smart glasses in the coming years?

是的，你刚才说的全部都会发生。首先，会有不同的品类和价位。我认为最基础的一层是模型能够读取并看到图像和音频。但接着我们开始看到很多有趣的技术，可以把信息显示到眼镜里，我相信这会成为主流。所以我认为第一点是：我如何获取信息？我如何获取文字？我如何获取方向？我如何获取叠加的信息？在显示、投影方面有很多非常有趣的技术，我认为这会成为主流。另一件事是你如何运用一些技术和技巧——其中很多是我们为虚拟现实眼镜开发的，比如追踪你的眼睛、追踪你手部周围的不同物体。这会为你创造操作的控制点，尤其是当有图像回显给你的时候。

Yes, everything that you just said. First of all, there are going to be different classes and price points. I think the very basic one is the ability for the model to consume and see images and audio. But then we're starting to see a lot of interesting technology in terms of displaying information for you into the glasses, and I believe that's going to become mainstream. So I think the number one is: how do I get information? How do I get text? How do I get directions? How do I get superimposed information? There's a lot of very interesting technologies on display, on projection, and I think that is going to become mainstream. And the other thing is how you're going to take some techniques and technology — much of it we developed for virtual reality glasses, like tracking your eyes and tracking different things around your hands. And that will create control points for you to operate, especially if you have images that are displayed back to you.

Cristiano，你在这场对话里提到的另一个设备是别针。我记得几年前有一款别针产品出来，引发了一阵病毒式传播，结果有点扑街。是太早了吗？还是人们对这类设备根本没有需求？这类设备还有未来吗？

The other device you mentioned in this conversation, Cristiano, was pins. And I think it was a couple of years ago there was this kind of viral moment with a pin that came out, and it kind of flopped. Was it too early? Was there just not appetite for this kind of device? And is there a future for this kind of device?

是的，我觉得到这个时候答案大家都知道了。它太早了，因为那是在生成式 AI 之前，是在编排器和智能体之前。但现在你有了这些，你就能看到它的价值。但说实话，除了眼镜之外，其他形态还没有被定义。我认为会有大量针对不同形态的尝试。我们一直在和每一家云公司、每一家……目前我们有 40 款这类设备的设计，我告诉你，形态的种类非常非常广。你会看到首饰。你会看到耳机——它们带摄像头。你会看到吊坠。你会看到别针。你会看到手表。你会看到人们随身携带的其他东西。我们会看它如何演变。但原则是：某种你穿戴的东西，某种一直伴随你的东西，某种能看到你周围世界的东西。这样你就有上下文，并且能够接入一个智能体、和智能体对话。就是这样——未来几年内我们肯定会在市场上看到所有这些设备，然后看人们想要什么。

Yes, I think we all know the answer at this point. It was early because it was before GenAI. It was before orchestrators and agents. But now that you have them, you can see the value of this. But to be fair with you, other than glasses, the other form factors are not yet defined. I think there's going to be a lot of experimentation with different form factors. We've been working with every cloud company and every — right now we have 40 designs of those devices, and I'm telling you the types of form factors are very, very broad. You're gonna see jewelry. You're gonna see earbuds — they have a camera. You're gonna see pendants. You're gonna see pins. You're gonna see watches. You're gonna see other things that people carry. And we'll see how this is going to evolve. But the principle is something that you wear, something that is with you all the time, something that can see the world around you. So you have context and have the ability for you to access an agent and talk to the agent. And that's it — it's definitely within the next few years we're going to see all those devices in the market, and we'll see what people want.

Cristiano，我知道这是个公开的秘密——你们在和 OpenAI 合作他们的自有设备，我也知道这方面你能说的不多。但这也凸显了那些此前并未真正涉足消费设备的非传统公司，在 AI 时代是如何思考消费设备的。你提到你们有这么多不同的设计在推进。往后看，你是否预期会有更多通常不在消费电子领域的公司进入这个空间，因为他们在 AI 设备上看到了机会？

And Cristiano, I know it's an open secret — you guys are working with OpenAI on their own device, and I know there's not much you can say about that. But it also underscores the way how non-traditional companies who haven't really been in consumer devices before are thinking about consumer devices in the age of AI. You mentioned you've got all these different designs in the works. Going forward, are you expecting more companies who haven't typically been in consumer electronics to get into this space because of the opportunity they see with AI devices?

百分之百。让我解释一下为什么会这样。我要提一件我之前对 Arjun 说过的事。记得我跟你说过：当智能手机处于你数字体验的正中心时，你更适合采用垂直生态，由提供手机的同一家公司为你提供所有可穿戴设备。但现在，当智能体或智能体集合成为生态的中心时，你可能会采用横向打法。你可能有多个设备，因为它们只是松散连接——因为它们连接的是智能体。这为许多公司进入这个空间创造了机会。事实上，当你看人们买的眼镜，那些是时尚设备。那些是时尚公司变成科技公司；它们不一定是消费电子设备。回到你关于这些公司为什么这么做的问题：在我们经历这个转型的过程中，先是训练——在数据中心里创建模型和 AI。然后是推理，生成 token。但现在你有了智能体，它们为这些 token 创造需求。这些公司意识到，随着智能体的崛起，AI 将被人类使用。不是那些人会跑到数据中心去按铃说‘给我来点 AI’。所以发生的情况是，我们佩戴的所有设备都成为智能体的端点，而那些 AI 公司明白，他们必须赢得这些智能体端点。

100%. And let me explain to you why this is happening. And I will touch on one thing that I said before for Arjun. Remember I told you: when the smartphone is at the very center of your digital experience, it's more conducive for you to have a vertical ecosystem, where the same company that provides the phone provides you all the wearable devices. But now, when the agent or collection of agents is the center of the ecosystem, you may have a horizontal play. You may have multiple devices because they're loosely connected — because they're connected to the agent. That creates opportunity for many companies to enter the space. As a matter of fact, when you look at glasses that humans buy, those are fashion devices. Those are fashion companies becoming technology companies; those are not necessarily consumer electronic devices. And I think, to go back to your question about why those companies are doing this: as we go through this transition, there was training — creation of the models and AI into the data center. Then there was inference to generate tokens. But then now you have agents that generate demand for those tokens. What those companies realize with the rise of agents is that the AI is going to be utilized by humans. It's not that those humans are going to show up in the data center, ring the bell and say, 'Give me some AI.' So what happened is all the devices that we wear become endpoints for agents, and those AI companies understand they have to win those endpoints for agents.

还有最后一块数据——没人谈论它，但几年后当人们开始谈论这件事时，我们会想起这场对话。我们今天之所以有 AI，是因为我们把所有信息都数字化了——我们所有的书、我们写的东西、邮件、互联网上的内容、社交媒体帖子，所有这些信息、网页——都在互联网上，现在都是数字化的，所以你可以训练模型。但一旦我们开始每天使用智能体，我们开始戴着智能眼镜四处走动，看到我们所看到的。那个数据量是巨大的——比我们用来训练模型的数据呈指数级地大。所以那些公司想要获取这些数据，因为这对训练未来的模型、为你获取上下文很重要，因为你的消费体验将是个性化的。所有这些都说明了那些公司为什么对做设备感兴趣。而我们非常兴奋的是，我们做的所有芯片——从耳机里不到 2 毫瓦的芯片，到我们进入数据中心时的 2000 瓦芯片——都与所有这些公司相关。

And there's the last piece of data — the one that nobody talks about, but we're going to remember this conversation a few years from now when people start talking about this. The reason we have AI today is because we digitized all of our information — all of our books and things we write and emails and stuff on the internet, social media posts, all of this information, web pages — it's on the internet, it's digital now, so you can train models. But once we start using agents every day, we start walking around with our smart glasses to see what we see. That amount of data is massive — it's exponentially larger than what we use to train models. So those companies want to have access to the data because it's important to train future models and get context for you, because your consumer experience is going to be bespoke. All of those things point to why those companies are interested in doing devices. And we're very excited that all of the chips that we do — from an earbud of sub-2 milliwatts to a 2000-watt chip as we enter the data center — become relevant to all of those companies.

好。我想先聊一下芯片——不是薯条那种 chips，你知道，取决于你在英国还是美国，这个词的意思很不一样。但我想聊一个叫 SoC、也就是系统级芯片的术语。它实际上是一块硅片，上面集成了很多不同类型的芯片，你会听到 Cristiano 提到其中一些。

Right. I just want to talk about chips for a second — not chips, the type that you eat; you know, depending if you're in the UK, that means something very different to if you're in the US. But I want to talk about a term called SoC, or system on chip. Now, this is effectively a piece of silicon with lots of different types of chips on it, and you'll hear Cristiano mention some of those.

那么，我先快速介绍一下它们是什么。第一个是中央处理器，也就是 CPU。它实际上是设备的大脑，负责所有通用计算任务，比如处理应用和系统。第二个是图形处理器，也就是 GPU，负责为游戏或设备本身渲染图形。还有一个叫神经网络处理器，也就是 NPU，负责处理 AI 工作负载。所有这些都集成到一块硅片中，然后装进你的设备。这实际上是高通设计的，但由其他公司制造。如果你听过播客之前的节目，我们已经多次讨论过芯片供应链。但现在整个半导体供应链存在大量瓶颈，其中之一是内存。现在所有设备都有某种形式的内存，你的笔记本电脑、手机。问题是，世界上只有极少数公司能生产这种内存芯片，而目前所有这些产能都流向 AI 数据中心和 AI 芯片。因此，消费电子设备等产品面临严重的内存短缺。这是目前所有电子公司关注的重大话题，也是我和 Cristiano Amon 讨论的内容。这一切的核心是驱动这些设备的芯片，而高通当然是该领域的领先者之一。这些运行 AI 的新型设备激增，对你思考芯片架构的方式意味着什么？人们希望在一颗芯片上集成更多东西，让它们更强大、更节能。那么，当你考虑所有这些新设备时，这对未来芯片设计意味着什么？

So, I just want to quickly lay out what they are. The first one is the central processing unit or CPU. This is effectively the brains of a device. It does all the general computing tasks like handle apps and the system as well. The second one is a graphics processing unit or GPU. This renders graphics for gaming or the actual device itself. Another one is called the neural processing unit or NPU and this handles the AI workloads and all of these are integrated into one piece of silicon and thrown into your devices. And that's something effectively that Qualcomm designs, that however is manufactured by other companies. And if you've heard previous episodes of the podcast, we've spoken so much about the chip supply chain. But right now there are a load of bottlenecks within the entire semiconductor supply chain. One of those is around memory. Now all your devices have some sort of memory in it. Your laptop, your phone. And what's happened is there are very few companies in the world that can produce this memory chip. And all of that right now is being funneled towards AI data centers and AI chips. And so there's a big shortage of memory for things like consumer electronics devices. That is a big topic right now of focus for all kinds of electronics companies. And it's something that Cristiano Amon and I spoke about. At the heart of all of this is the chips that are going to power these devices and Qualcomm of course one of the leading players in this space. What do all of this proliferation of new types of devices that are running AI mean for the way you're thinking about chip architecture? This desire to pack more and more on a chip and to make them more powerful and more energy efficient. And so as you think about all these new devices, what does it mean for the way chips are going to be designed going forward?

对半导体公司以及像高通这样的公司来说，这是一个极其激动人心的时刻。我们很荣幸——这确实是一种荣幸——能在移动行业培养我们的能力，因为移动行业非常严苛。每一代用户都想要更强大的手机，能做更多事情、玩游戏、拥有强大的 GPU 和 CPU，但它仍然必须能放进口袋，仍然必须有一整天的续航，而且仍然不能发热。我没办法在智能手机里做液冷、装散热器。所以这在我们的半导体工程中形成了一种 DNA：你必须以非常高效的方式集成更多算力。我认为这对像高通这样的公司来说是一个很好的待解决问题。在眼镜上也是如此：我如何在独立的眼镜中集成大量算力并保持电池高效？所以我认为这正在从根本上改变架构。我们正在大幅提升神经处理单元的能力，这些 NPU 将在设备上运行那些较小的模型，与云端的大模型协同工作。这意味着硅片面积和专门用于运行模型的部分将大幅增加，无论是在 GPU 还是 NPU 上。我们需要更快、更节能的 CPU，因为智能体驱动 CPU。而且我们需要以持续明智的方式管理连接性和电池续航。我们整个路线图现在都在升级过程中。整个路线图，因为我相信我们今天拥有的设备没有一款为这个未来做好了准备。你将不得不升级你的手机，升级你的 PC，你将拥有眼镜和吊坠等新类别的设备。你的汽车也将升级，你将能够在车里与智能体对话，我认为这将发生在我们交互的一切事物上。这是一种新的计算形式、一种新的软件形式，需要新的硅架构。

It's an incredibly exciting time for semiconductor companies and for semiconductor companies like Qualcomm. We had the privilege and it was a privilege to be developing our capability in the mobile industry because the mobile industry is very unforgiving. Every generation users want a more powerful phone that does more things, plays games, has big GPUs, big CPUs, but it still has to fit in your pocket, still has to have an all day battery life, and it still cannot get hot. I don't have the ability to do liquid cooling and put a radiator into a smartphone. So that creates a DNA in our semiconductor engineering that you have to pack more compute in a very efficient way and I think that is a great problem for a company like Qualcomm to solve. It's also true on glasses: how can I put a lot of compute and be very efficient with battery in a glass independent? So I think that is fundamentally changing architecture. We're increasing dramatically the ability to have neuroprocessing units that are going to run those smaller models into those devices working alongside the big models in the cloud. That means the area of the silicon and the amount dedicated to running models increase dramatically. That's both on the GPU and in the neuroprocessing unit. We need a faster and more power efficient CPU because agents drive the CPU. And we need to do this in a way that continues to be smart about how we manage connectivity and the battery life. Our entire road map is in a process of upgrade right now. An entire road map because I believe none of the devices we have today are prepared for this future. You're going to have to upgrade your phone. You're going to upgrade your PC. You're going to have those new classes of devices like glasses and pendants. Your car is going to get upgraded. You're going to be able to talk to agents on the car and I think that's going to happen across everything that we interact with. It's a new form of computing, a new form of software and requires a new architecture of silicon.

那么 CPU 会有多重要？到目前为止这将会是一场大辩论，人们谈论 CPU 的复兴，不仅是在数据中心，甚至在我们考虑这些消费设备时也是如此。

And how important will the CPU be? This is going to be a big debate so far and people talking about CPU having a renaissance not just in the data center now but even as we think about these consumer devices.

超级重要，完全正确。回到我开始谈智能体时给你的例子：你有像 GPU 甚至 NPU 那样的大规模并行处理，但那些编排器实际上处理非常复杂的任务，它们检查你的数据、检查模型、上网，而那是 CPU 在编排所有这些。那些编排器运行在 CPU 上。所以现在所有船都水涨船高，因为当我看到高通时，我们的 CPU 变得非常重要，而且我们在手机和 PC 上拥有行业领先的 CPU，很快我们将在 6 月 24 日宣布用于数据中心的 CPU。CPU 将非常重要。用于推理的 NPU 和高效推理将非常重要，GPU 也将继续重要。所以这就是为什么我认为现在是进入半导体行业的好时机。

Super important, it's 100% correct. And go back to this example I gave it to you about when we started the conversation about agents: you have parallel processing at scale like a GPU or even an NPU, but those orchestrators, they actually handle very complex tasks, they go check your data, check the model, go to the internet, and that's a CPU that orchestrates all of this. Those orchestrators live on the CPU. So now all boats are rising because like when I see a Qualcomm, our CPU became very important and the fact that we have the leading CPU in the industry in phones and PCs and soon we're going to announce our CPU for the data center on June 24. CPU is going to be very important. NPUs for inference and power efficient inference is going to be very important and also GPUs will continue to be important. So this is why I think it's great timing to be in the semiconductor business right now.

半导体行业正看到兴趣、创新和需求的巨大爆发，随之而来的是供应链上的许多瓶颈，无论是内存的获取，还是晶圆厂对这些芯片的实际制造。关于半导体行业历史上的一些周期性是否正在改变，存在很大争论。你现在如何应对所有这些不同的挑战，特别是关于内存和产能限制方面？

And the semiconductor industry is seeing a huge explosion of interest, of innovation and demand, and with that of course has brought a number of bottlenecks around the supply chain, whether that's in access to memory, whether it's in the actual manufacturing of these chips from the fabs as well. There's a big debate over whether some of the cyclicality that we've historically seen in the semiconductor industry is changing. How are you navigating all of these different challenges right now, particularly as it pertains to things like memory and capacity constraints?

嗯，首先，在回答这个问题之前，如果你有内存，我会从你那里买。好吧。所以我认为目前行业不幸地正在应对类似我们在疫情期间不得不应对的情况。我们的需求显著超过供应，内存已成为瓶颈。我认为产能不是为这个 AI 现实而建的，但我乐观地认为行业已经找到了应对方法，产能将会建成。我认为关于周期性问题，我觉得这有点不同。我认为全世界的计算基础设施已经升级。整个计算基础设施——AI 基本上是一种不同的运行软件的方式。你需要不同的、更强大的计算机。

Well, first of all, before I answer the question, if you have some memory, I'll buy it from you. Okay. So I think the industry right now is unfortunately dealing with something similar that we had to deal with during the pandemic. We had significantly more demand than supply and memory has become a bottleneck. I think capacity was not built for this AI reality, but I'm optimistic that the industry has found ways to deal with those things and capacity will be built. I think the question about cyclical process, I feel this is a little bit different. I think the whole computing infrastructure of the world has been upgraded. The whole computing infrastructure—AI is basically a different way to run software. You need different and more powerful computers.

你生成的是 token——这是 AI 世界的新货币。正如我们在对话中提到的，现有的算力基础设施还不够用，需要经历一个升级的过程。我们在数据中心开始看到的需求会持续下去，而随着智能体站稳脚跟、开始规模化，其他设备上也会出现同样的需求。我希望我们有更多内存。我想我们一直在与业界合作来提升产能，半导体制造方面也是如此。但这是个好问题，而且我很乐观。有人说这个问题要到 2028 年才能解决，我对此更乐观一些。我认为在 27 年的某个时候，至少在下半年，我们应该会看到更多产能变得可用，也会有更高效利用内存的不同架构。我认为这为高通这样的公司创造了机会——我们在边缘设备的功耗和内存使用上一向非常节省——去在数据中心做一些颠覆性的事情。这正是我们想在 6 月底公布路线图时展示的。

What you generate is tokens — that's the new currency of the AI world. And as I said during our conversation, the existing compute infrastructure is not capable and needs to go through a process of upgrade. The same demand that we're starting to see in the data center will continue, but as agents get a hold and start to get scale, it's also going to happen with the rest of devices. I wish we had more memory. I think we've been working with the industry to get more capacity, and that is also true on semiconductor manufacturing. But it's a good problem to have, and I feel optimistic. Some people talk about this only getting resolved by 2028; I feel more optimistic about that. I think sometime in '27, or at least the second half, we should see more capacity becoming available, different architectures that are going to make use of memory more efficiently. And I think that creates opportunities for a company like Qualcomm, which has been very economical in the use of power and the use of memory building devices at the edge, to do something disruptive in the data center. And that's what we want to show when we unveil our road map at the end of June.

伴随着半导体领域这么多的创新和关注，也带来了相当有意思的竞争——我觉得更多是在数据中心这一侧。你怎么看竞争在消费电子这一侧会如何演变？因为我们还没看到消费电子领域出现大量新玩家。我知道苹果一段时间以来一直专注于为自家产品做自己的芯片。Pixel 里有 Google 的 Tensor。三星当然是你们的合作伙伴，但也用自家的 Exynos 芯片。你是否预期，随着新设备、新形态进入市场，你可能会看到这个领域竞争加剧、出现新玩家？

And with all this innovation and interest in semiconductors, it's brought an interesting amount of competition — I think more on the data center front. How do you think competition is going to play out in terms of the consumer electronics side? Because we haven't seen a huge amount of new players in terms of consumer electronics. I know Apple's been laser focused for some time on its own chips for its products as well. We've got Google Tensor in the Pixel. Samsung, of course, is a partner of yours, but also uses its own Exynos chips. Are you expecting, as new devices come to market, new form factors come to market, that you may see increased competition and new players in this space?

这是个非常棒的问题。你看，我无法预测未来，但我可以告诉你一件事——一件我能预测的事。我们在移动行业从 2G 到 3G 到 4G 到 5G 所看到的：玩家在变，设备在变。我知道我们这次对话里没聊到，但我们正在做 6G，它实际上是面向 AI 时代和那些新类别设备的下一代无线技术。所以我们现在开始看到的情况——我们在聊个人 AI 设备时也提到了这一点——我们现在看到的是，对移动行业感兴趣的玩家正在改变。现在所有 AI 公司都在关注移动设备、关注终端，而且它们在造设备。所以我开始看到的一件事——今天我已经看到了——就是开始看到智能手机 OEM 和 AI 公司之间的结合，智能手机 OEM 和模型之间的结合。而且我确实认为，随着你手机里的工作负载开始由智能体和模型驱动，这个格局会变得非常不同，玩家也会非常不同。所以我认为这个行业会发生变化。我无法预测赢家和输家，但我可以预测变化，而且我认为这很可能会改变智能手机 OEM 与智能体之间的关系。这一点会变。

This is an awesome question. Look, I cannot predict the future, but I can tell you one thing — one thing I can predict. What we have seen in the mobile industry from 2G to 3G to 4G to 5G: the players change, the devices change. And I know we didn't talk about it in this conversation, but we're working on 6G, which is actually the next generation of wireless for the age of AI and those new classes of devices. So what we're starting to see happening right now — and we touched on this point when we had this conversation about personal AI devices — what we're seeing right now is that the players interested in the mobile industry are changing. Now all AI companies are looking at mobile devices, at endpoints, and they're building devices. So one thing I've started to see — I already see it today — is starting to see associations between smartphone OEMs and AI companies, associations between smartphone OEMs and models. And I actually think this landscape is going to be very different as the workloads in your phone start to be driven by agents and models, and the players are going to be very different. So I think this industry is going to see change. I cannot predict the winners and the losers, but I can predict change, and I think it's likely going to change the relationship between smartphone OEMs and agents. That's going to change.

而在芯片这一侧，你看，我们一直都有竞争。我一直都有竞争，而且我认为竞争对所有那些行业都是好事，而且好得惊人。这也是我认为数据中心也会看到竞争的原因。我们一直遵循的规则是——我认为高通独特的一点是，玩家来了又走了。我们在移动行业看到很多公司从英雄变成零。我们留下来了。我们留下来，是因为如果你有正确的技术，如果你有领先的技术，你总会有立足之地。而今天的成功并不保证明天的成功。所以我们必须继续重塑芯片、持续创新，我认为这就是移动行业和消费设备在 AI 转型中摆在我们面前的事情。

And on the silicon side, look, we always had competition. I always had competition, and I think competition has been good and incredible for all those industries. That's why I think the data center is going to see competition too. And the rule that we have followed — I think one thing that is unique about Qualcomm is that as players came, they went away. We saw in the mobile industry many companies go from hero to zero. We stay. We stay because if you have the right technology, if you have leading technology, there's always room for you. And success today is no guarantee of success tomorrow. So we have to continue reinventing silicon and innovating, and I think that's what we see in front of us with this transition of the mobile industry and consumer devices with AI.

Cristiano，和你聊天我总是学到很多。也让人有很多思考，我很期待看到你在这期节目里预告的一些东西在未来几个月到一年左右陆续推出。非常感谢你来参加。

Cristiano, I always learn so much speaking to you. It's so much food for thought as well, and I'm excited to see some of these things you've teased out on this episode coming out over the next few months and year or so. So thanks so much for joining me.

我的荣幸。和你聊天总是很愉快，Arjun，期待很快再见。

My pleasure. Always a pleasure talking to you, Arjun, and I look forward to seeing you soon.

和 Cristiano 的对话里真有不少有意思的内容。不过其实，我很想听听你们的想法。我很想听听你们对这期节目、或者对 Cristiano 谈到的任何内容的看法。你们可以直接联系我。我在 X、LinkedIn、TikTok、Instagram 上都有账号——搜我的名字 Arjun Carpool 就行。也可以发邮件到 thetechdownload@cnbc.com。如果你喜欢这期节目，就订阅、点个赞、留个评论，随你喜欢。我们下期 Tech Download 再见。

Some really interesting stuff in that conversation with Cristiano. But actually, I'd love to hear from you. I'd love to hear your thoughts about the episode or anything that Cristiano was speaking about there. You can get in touch directly with me. I'm on X, LinkedIn, TikTok, Instagram — just search my name, Arjun Carpool. You can also email us at thetechdownload@cnbc.com. If you did like this episode, subscribe, drop us a like, drop us a comment, whatever you like. And we'll catch you next time on the Tech Download.