LIVE: Qualcomm CEO Cristiano Amon Unveils AI Future at COMPUTEX Taiwan Keynote Event | AI1B

Hello everyone. Very, very happy to be here at Computex. This is a special time, I think, for Qualcomm, a special time for technology, and it's such an honor, I think, to be able to speak with you all here at Computex. So thank you so much for being here. As I said, I think we at Qualcomm are incredibly excited about the future. Is there some Qualcomm folks here in the room? All right. Thank you. Thank you so much.

Before I start this presentation, this presentation is not about Qualcomm. It's not about Qualcomm. It's really, at the end of the day, I think, you know, evolution of technology is really a combination of companies. Not one company is never responsible for everything. It's really about partnership. So first thing I wanted to say a big thank you to all our Taiwan partners. First, all of our suppliers, all of our development partners. A special thank you to TSMC, which has been an incredible partner on this journey. So thank you to our suppliers and our developers. And then also a big thank you to the entire ecosystem and the partnerships, and many of you which are here today. Thank you so much. This is not about Qualcomm. This is all about you and an incredible ecosystem that actually drives everything in technology with a wonderful partnership. Companies that are leaders in technology today, they are leaders because of those partnerships. So, thank you very much and thank you all for having me here speaking to you.

Okay, we are incredibly excited because a lot of the things that we have been talking about, it has really started to become real and it's very easy to see. I have one job today. One job today. And by the way, I won't talk about a data center. So, if you came here to hear about our data center projects, I'll tell a little bit, but we'll talk about this at the end of the month. But I have one job today. I want you to understand how AI is going to evolve for all the devices because it's the new form of computing that is touching every single device. And if you walk out of this keynote today understanding what's going to happen to all the devices at the edge and how much we're excited about it, I have done my job. So, I'm going to do my best to tell you how this future is going to look like. It's very clear to us. We have conviction what's going to happen and we're going to make it very concrete for you. But it's an incredibly exciting time for all of the computing technology and all the devices that we use every day and the ones we're going to use.

So about two years ago, we talked about how AI will change the human-computer interface and as a consequence will change the architecture of all of our personal computing devices, and that is starting to become a reality in 2026. That's why we call it Qualcomm. 2026 is the year of agents, and it's now how AI is really evolving and is going to get to incredible amount of scale. It has evolved from simply answering prompts and working as a tool that augments how we as humans work with our computers into something that can take action on their own. I'm going to give you some examples. I think what you see today in the screen at home, you're going to have agents that are going to basically update everyone on your activity and your schedule. All of the things that you need to do at work will be with you, and think about what are the projects that you need to do, what are the tasks that you need to do, it flags the decision, and it's going to be working with you. And in your personal life, we'll give you information that is relevant to you, all of the updates you need, work with you on a schedule. Everything's going to be prioritized and it's going to be a little different. It's going to feel more autonomous. It's going to feel more personal and it's going to feel like really a companion. And all of this is going to happen proactively across all of your devices.

And for that to be understood, we have to go back to a very basic element of this transition. We have been talking about this, right? Which is the phone today is at the center of your digital life. And I remember when I presented something similar like this when we talked about it at Snapdragon Summit, I actually talked about here at Computex that the phone is the center of a digital life and therefore everything is around the phone, is how we interact with the phone, with the OS and the applications, even the other wearable devices, they are around this phone ecosystem. But now this is different. Those agents become the center of your digital experience. So including the phone and the other devices, they now are around the agent. It's not about an extension of the phone and the digital ecosystem is no longer at the phone itself and the OS and the applications. And what happens, those devices become endpoints for agents. And agents, they are not trapped, they are not attached to a single ecosystem. And everything that connects you with an agent becomes an endpoint for AI. And that's a very important change. Once you understand that change, you understand how the whole mobile industry is going to change. The things that we take for granted today are going to be different tomorrow. And there's one thing I can tell you coming from Qualcomm. I've been in the company for 30 years. One thing is guaranteed: the mobile industry is not static. It changes all the time.

So those endpoints for AI, I am going to talk about what is really the surface area. I'm going to remind you of what are the scale of those devices. There are about 6 billion phones today, 2 billion personal AI devices, which is the evolution of wearables. And that is exactly a small number when you think about those things as endpoints of agents. 2 billion PCs, half a billion connected cars, and people will experience and interact with those agents on these devices. And as I said in the beginning, the role of this conversation today is to help you understand the role of those devices at the edge for this next phase of AI. And we at Qualcomm, we actually told you about those devices running AI before it was popular, but now it's very easy to see. So once you understand the scale, one important message that I want to deliver to you is that today's devices were not designed for those experiences. It will require a different kind of device when agent AI becomes the number one AI function in our daily lives, especially for personal computing as we interact with those devices. All of those devices today, they have been built for actions initiated by the user, not by the agent, and I'm going to break that down for you.

The agents are different. They operate all the time. They carry context forward. They will orchestrate multiple tasks reliably and securely. But they are going to do that on their own. Not necessarily with human intervention. That alone gives you a glimpse how those devices are going to change and how the architecture of those devices are going to change. And when you think about that, the agent acts on intent. As you have now AI taking a role in the man-machine interface, you understand human intentions. We'll proactively understand the text that it needs to react with the human. We'll take a goal, we'll break it into several steps, we'll coordinate across systems, across data sets—your own data set that lives on the device, the data set that is unique to you on your personal graph, and the data set that is going to be in the cloud. It plans, it executes, you verify, you will keep interacting until the task is done. And that by definition changes the hardware with different kind of hardware to be able to do this because it's not just what is initiated by the user. The agent is going to be running on its own. It has also changed the operating system and changing the application. And I think that's the incredible opportunity that we have in computing across this entire surface area of devices which are going to be upgraded for this age of AI.

You need to be able to tightly integrate across all the compute domains. You have strict power and latency constraints. I cannot emphasize enough the importance of power. Think about your phone. If it is challenging to make your phone last all day with you operating it, what happens when you and the agent is operating it? That's an incredible engineering challenge from power and latency. Those devices need to be able to support planning, reasoning, coordination across the system. And for that you can start to understand how the compute is going to evolve. You need a very strong and power-efficient CPU for the orchestration of tasks. I think the industry now understands the importance of CPU for orchestration and where the orchestrator is going to run. You need very power-efficient and high compute density of NPU and GPU to be able to run local models.

One other thing that I'm going to do today is to show clearly how AI is going to be distributed across cloud and device. I almost feel that it's going to be so obvious that we're going to start talking about it. We're going to stop talking about cloud and edge because it's all going to be one system. And for that, you're going to need the computing power in those devices to do that on the NPU and then the GPU. You need to have sensor data. Context is super important. Without context, you cannot have the agent to be useful and for you to be useful and be proactive. All of those things are going to be necessary as we think the next generation of devices for the agent AI, and an upgrade is truly coming.

So, and it's not the same thing for every device. It is not. We know this at Qualcomm. In personal devices, that means all-day intelligence, very focused on sensors and context that matter to you. Incredibly power-efficient because you're going to carry, you are mobile by definition, you're going to carry with you and need to have high-speed connectivity. In cars and robots, that means sustaining processing power on the most demanding condition and also independent of connectivity and cloud when you have to. And in the data center, inference will require incredible amount of scale, and I'll walk you through it when we think about the demand for tokens and power efficiency is also going to matter there. I'm going to tell you something for companies that have the benefit of being working in the mobile industry. The mobile industry for decades, you had a different growth rate of computing power and availability of energy. So you have to come up with solutions that allow you to get all the computing power that you need with the available energy because the battery has to last all day and is limited in its size. That's also going to be true when you think about what's going to happen on the data center. But we'll tell you more about that at the end of the month.

And that in itself is the Qualcomm advantage, because we have been working with all of those devices at the edge. All of it, and especially the new Qualcomm that is incredibly diversified. We now have an ability to build systems. There's sub-2 milliwatts from an earbud with micro-power Wi-Fi that connects to an agent for personal AI audio device, all the way to kilowatts what's going to happen in the data center. And I think that is the incredible opportunity that we see for us and for our partners when all of those devices are going to change for this future of agent AI. And for that you need dedicated computing solutions that scale all the way down to a sub-milliwatt device to a 2,000-kilowatt, and that's an incredible opportunity I think that we have in front of us. Leading performance per watt across smartphones, PC, automotive, compute, and robots, and extending that to the data center. I think that's our mission and that is the incredible opportunity.

The last thing I'm going to tell you as we keep moving in the presentation, it is there is excitement about what's happening especially on the PC side with all of those agentic AI. So as we go through the presentation, I will come back to this and I will come back to the phones because I believe that what's going to happen on the phone is going to be a very good proxy of what could happen on PC and other devices. But hold that thought. So, I'm going to break this down for you across the different categories of devices and I want to show you how those experiences are evolving. I want to make it less abstract and more concrete for you because this is happening right now. So, I want to start with the devices that are really powered by Snapdragon platforms. You will see agents now being deployed and starting to get scale across phones, PC, and this new category of personal AI devices. I would like you to watch a video. Let's run the video.

utilized by a lot of people. If you have a phone, you don't have the ability to buy a computer, connect it to a battery, put it in a backpack, and carry it with you. You're going to carry one device, and those things are going to happen in the phone. So those devices are going to have two personalities, and I'll come back to this. One personality is you going to operate the device as a human. That's how the device of today has been designed to operate. But the agent is going to operate the device as well. And that's how devices are going to be in the future. And that's going to happen on the same device. And the same device doing both, like a phone, is also going to influence how it's going to happen on PC with laptops. And those personal AI devices are going to get scale because they're very natural for AI glasses, for example, as close to your eyes, to your mouth, to your ears. They became very natural and that is this incredible transition we're going to see in compute.

But it's not just unique to personal computing. This is also happening as we think about the transition that we're going to have to the physical world. The physical world is one of the most exciting areas in our industry right now. Cars, robots, industrial systems are going to be also significant endpoints for agent AI. And in those systems, you have other requirements. Latency, safety are measured in milliseconds, in millimeters, every single watt matters. So I want to show you a demo next and I'll walk you through it. For example, how things are going to work in a car. When you think about the car, there will be two layers of intelligence. So in the cockpit, AI will personalize your experience, and that is the same thing that we just talked about. The agent is your agent regardless of the device. When you're behind the wheel, you're going to take that experience with you, it's going to interact with the agent. It doesn't matter if you're going to be in the car, on a personal AI device, on a phone, on your PC, because the agent is at the center, all the devices around it including the car. But also the car has other tasks besides interacting with you. It's interacting with you within the context of the car and on the road. You also have physical AI that is perceiving, it is planning, it is acting using cameras, radar, sensors, and maps for your navigation.

So that's the interesting thing. You have two different layers of intelligence. You have the interaction with you in the cockpit which is personal to you when you are within the context of the car, but you also have the physical AI that drives the car for you. And this is designed to operate as one integrated system. Watch this video of this in action which will understand how the system works as one.

So, this is very interesting because what you can see, the cameras of the car, the radars and all the sensors, they've been on this layer of intelligence to drive the car. Now that also input and context for the agent for your experience, and we're just at the beginning of this. And what we actually see within Qualcomm, this whole transition of the software-defined vehicle is now augmented when the vehicle is evolving to the AI-defined vehicle. Now I talked about personal computing, cars. I want to talk about the next big category of physical AI, which is robotics.

The entire robotics industry right now has been transformed, and robotics is very interesting because it pushes the limit of computing technology and incorporates aspects of what I showed you. Incorporates many aspects and technology of what we see in personal computing for sensing, perception, high integration of sensors. The robot, like your glasses, will see what you see, read what you read, understand the context. At the same time, you have all of those elements of technology that comes from the automotive industry. You need precision, you need industrial grade, you need safety, you need redundancy.

So that is an exciting element.

The computer vision. You can see, I think, the... You can see, I think, autonomous driving, and you're going to get hundreds of millions of those connections in real time. You triangulate all of this data and what you do, it will create a digital twin not only of the neighborhood, of the city, but the whole country. And what you're going to do with this, drone detection is an example. You detect everything that moves and flies. You manage the entire low-altitude aerial economy. You're going to detect on every road, every car, every bicycle, every truck, every pedestrian. You can actually identify those objects and annotate on that image. You bring this whole conversation we just had about computer and visual intelligence as an example into a large digital twin, and that is going to change everything. And what's the connection with this conversation we have of agent AI? This is all real-time context for agents that is going to be running on the endpoints across all of those devices.

This is an exciting, I think, future for everything that is happening on the edge because AI is the new form of computing and is going to be how computers are going to run everywhere. So now I'm going to get to the last part of my presentation. And this is the second part of what I said it was my objective today: to explain to all of you how the devices at the edge is going to change for AI. The second part is trying to get a very, very simple understanding of how cloud and edge are going to work together. It's now very obvious but I want you to get the same conclusion that we got how this world is going to evolve. So agents are now how tokens, how demand for tokens is created, and it's how AI is going to get to scale and is defining both the architecture and the economics of AI. So let's go back and recap.

Today most of the software, the OS, the app store, they have been designed for a human to operate. So we operate and we do things. We start an app or a program or go to the web. We do different things. It's been designed for us. Now with agents, everything changes. The agents are autonomous. They interact with software much faster. They do it across multiple services at once. That's why the device is going to have to be upgraded. And every workflow generates tokens at machine speed, not human speed.

Just as an example, when you think about an application or even the SaaS industry, you think about software for the SaaS industry. Those clients to date, they've been designed for a human to operate the client. Not anymore. You have to design your application and your SaaS for a human and an agent to interact. And the agents are going to interact faster. That is going to fundamentally change the demand for tokens. So as an example, you can see those transitions, how quickly token consumption is growing. So a lot of different estimates, but it currently looks like this. So level one, when we started conversational, you go in and you have a single turn, you prompt, you get a response. That's about, you can talk about 10,000 tokens on a single turn prompt response for level one conversational. Level two, reasoning, then you now have multi-turn, and you can see an order of magnitude increase, about 100,000 tokens per task. And then when you think about level three, agent AI, autonomous, multi-step, multiple tool calls, it's about 1 million tokens per task and is growing. You see the order of magnitude increase, that's about 100x increase in the number of tokens in two generations.

Now let me give you the same data, different perspective. So this year, 2026, it's estimated that the global token demand within 10 seconds is about 31.7 billion tokens. So the projection for 2030 for the same 10 seconds is 1.27 trillion tokens. It's a 40x increase, and that's because agents are going to generate a lot of tokens. So if you think about the latest estimate, and I'm sure that's going to change, but the latest estimate for total token demand globally in 2030 is in the quintillions. I don't even know how to say the number. I'll say it's a four gazillion tokens that is the token demand for 2030. And that is the new currency of AI. Tokens are the currency of AI. And let me now share with you how this is going to work. And there's going to be a lot of people, there's going to argue, there's no need to argue, this is exactly how it's going to work. We have seen this before, resistance is futile. Okay, so let me give you some examples how this is going to work.

So let's talk about coding. Coding workloads are some of the most token-intensive tasks today. Using a real session running Cloud Code, an orchestrator is now intelligently routing the workload, keeping certain tasks on the compute available on the device and sending what's necessary to the cloud. When you apply the distributed agent AI making use of all the compute across this compute continuum, you can see you save about 1.4 million tokens, 60% lower cost for the same result. Now that's coding. Let me give you a completely different example.

So the next demo I'm going to show you is a task: create a Snapdragon web page. There's a lot of specifications on the prompt about this web page needs to be created. On the left, you can see that it runs on the cloud and generates tokens in the cloud. On the right is the same result but you have an intelligent distributed orchestrator that defines this routing of the task about the compute that is available on the device and the compute available on the cloud. Result is you get exactly the same result, 30% fewer tokens, 4x lower cost. And that is the power of distributed AI and this is how this is going to play out.

Now, I like to provide this example. It's interesting because we've been seeing this conversation about cloud and edge and they're both incredibly important. And now you've started to see how this is actually going to play out. It's not an abstract thing. Exactly. The conversation is never about this thing that runs on the cloud can also run on the edge. No, what needs to run on the cloud is going to run on the cloud. What needs to run on the edge is going to run on the edge. It's different. I'd like to do this analogy. Every one of you here has probably over 200 apps organized in folders in your smartphone. And if I ask you anybody to go to the exercise of say for each app what runs on the computing on your smartphone, what runs on the computer on the cloud, it's an exercise in futility. And you know the importance of the cloud because if you put your phone in airplane mode, it's useless. But you need a lot of computing and this is how it's going to work. AI, agent AI is agents running the computer for you. And as the workloads start to shift, the things that you do yourself and the thing that the agents are going to do is going to make use of all the computing that is available and is going to go exactly how it works today. The computing is going to be utilized. Everything will be running AI and it's going to happen across the compute continuum. And I think that's the exciting part for Qualcomm because that's going to create demand for AI computing everywhere and every computing engine becomes interesting and necessary for this agentic future. And you're going to naturally process the workloads where it's more efficient and that is the opportunity because we're uniquely positioned across all of those different devices. And it's never one size fits all. You need to have the right platform for AI for every device. As we talk about it, they are different. They have different purpose. It's about maximum intelligence and maximum efficiency everywhere.

So as I get to the end of the presentation, I have one last thing that I want to tell you. So let's watch the video.

We'll have a lot to share on our investor day in just a few weeks, June 24. I wish I can tell you all about our roadmap, but that is coming. And in summary, I'm just going to tell you we're incredibly excited about the future of technology across every single device. This is going to get transformed. We showed the agents are not coming into the future. They are already here. It's changing a lot of the compute, is going to generate a lot of demand for new classes of devices and computing, and this upgrade cycle can be one of the largest that the industry has seen. And together we're building the products and the technology that will define this new era. Thank you so much for listening to my presentation. Like I said, it's a pleasure and an honor to be here. Hopefully this was helpful for you and I can't wait to build this future together with our partners across the globe and our partners in Taiwan. Thank you very much. Enjoy Computex.