Maybe we can start from your newest press release yesterday because you mentioned that you announced one is you are going to invest the 10 billion dollars in advanced packaging companies in Taiwan and another is adopting TSMC's two nanometer process technology. So I think the people want to understand, could you elaborate more about your two announcements?
So, first of all, it's great to be here this morning. Thank you to Commonwealth magazine and for everyone to be here. It's always so exciting to come to Taiwan. I always feel that there's so much energy and you know there's no other place in the world that has so much semiconductors in a small, very, very small area. But yes, it's been a very exciting week for us and we're in this time right now where the semiconductor market and the AI market has never been growing faster. Every month we're surprised by how much growth there is and every time we look at the usage of the technology we're convinced that this is the highest growth time for the market and as a result for AMD our strategy is to lead in high performance and AI computing and to do that we must have a very strong supply chain and we must have very strong partners. So we've been working very closely with all of our partners here in Taiwan. Yesterday we hosted a customer symposium with over 150 of our partners and many are in the room today and our announcements this week are two. The first announcement is around the progress that we've made with TSMC. So TSMC has been a tremendous partner for AMD and I would say is a very key reason that we have been so successful and what we wanted to be was the first to really ramp 2nm HPC technology. So we're happy to announce that our newest CPU chip, Venice, has started its production ramp here in 2 nanometer. So it was very exciting for that. And then the second piece is around the supply chain. So I think all of us in the industry recognize that this is the time we have to invest because we need to have enough capacity for the next one, two, three years but it's a lot of investment that's required. So AMD is partnering with a number of companies here in the Taiwan ecosystem. Because we are leading in advanced packaging and we're leading in high performance computing, we are co-investing to build both the technology and the capacity, about over $10 billion for if you think about all of the advanced packaging substrates, test capacity, rack scale integration and it's a first of all it's a big, big vote of confidence for this is the place where the technology is the best and it's also very, very important for AMD's long-term growth. So, thank you for all of the incredible support here.
So, Dr. I want to follow up because I'm a journalist. Let me clarify it. You say you want to collaborate with the advanced packaging. So, how will you participate? Is it share the cost, research cost or...
Yeah, we do. So if you look at our technology we have been very aggressive in the use of 2.5D technology, CoWoS technology, you know, EFB technology. We're also very aggressive in using 3D technology and for all of those technologies we are very often the first user to use it at volume. So we collaborate with the ecosystem and we ensure that we're making investments upfront to have the capacity available as we ramp over the next several years and as we have seen the demand keeps going up and so we're asking our partners to ramp faster and we should absolutely share in that investment.
I see. Okay. So that's clear right now. So since you mentioned the packaging and also the processing company like TSMC, I want to understand what's your forecast? What opportunity do you see in the Taiwan semiconductor ecosystem in the next five years?
Well, I think what is so special about the Taiwan semiconductor ecosystem is it has every part of the ecosystem starting from the basic materials all the way through the advanced process technology to the back-end technology to the systems, our OEM and our ODM partners and now building to rack scale manufacturing. All of those pieces need to work very, very closely together and the only place in the world that has all of it together is right here in Taiwan. So we all view that this is really the center of the advanced technology that we're bringing out. I think what's different today I would say is in addition to needing the overall let's call it end to end we're also finding the need to have manufacturing capabilities around the world and so many of the partners in the Taiwan ecosystem are now also expanding outside of Taiwan and I think that's a really good thing because semiconductors are so, so important to the world having the geographic capability is very important.
Just as you mentioned that a lot of companies right now start to go global, originally semiconductor industry are all in Taiwan so what's your suggestion to those companies they are going global?
As I said I think it's a very important thing. You know what we're finding is maybe 10 years ago semiconductors were viewed as very special so only very technical people understood or very specific. Now if you look at everywhere, either you're talking about businesses or you're talking about countries everyone sees the semiconductors as essential. It's the most important kind of baseline foundation for computing and as a result I think for our Taiwan colleagues because you have such a special capability being able to participate more in the global ecosystem is actually a really good thing. I know it's not that easy because every country has its different way of doing things but I think it is very worthwhile to participate in that global ecosystem. Certainly I can say in the US you know it's very clear that the US is trying to build more manufacturing capability. That doesn't mean there's any way that the US can build all manufacturing capability. It's impossible. There's just too much requirement. But it does mean that having a very diverse ecosystem I think is a good thing and for those of our partners that are building around the world we're certainly supporting all of that.
But as you mentioned it's quite difficult, challenging for a lot of companies especially in the management skills right so any suggestion to the companies for the management?
Well my view is that I never saw any leadership team or capability more capable than the leadership teams of our Taiwan colleagues. So very, very capable. I like to say where there's that ambition you can find a way. And there's also the desire really by all of us to help bring up that ecosystem. I can tell you for example for TSMC in Arizona many people did not believe that the Arizona FAB could be as good as Taiwan and frankly it was not easy but we're currently running production in TSMC's Arizona Fab and it is excellent and it's excellent because a lot of the practices and procedures were transferred from the learning and I think that's what we do, we learn and help each other.
Great. Actually during the past when we get into the labor room and I asked Lisa what's her feeling about Taiwan supply chain and she gave me a very unique answer. She said the Taiwan partners are frankly competitors. Could you say that again to the audience?
Okay. So, Ian was asking me what do I think is different about the Taiwan ecosystem and I mentioned that last night every year I actually bring my leadership team here together to meet with our most important customers and partners and it's an opportunity for us to talk about our latest priorities and our technologies but the biggest reason is to appreciate and say thank you. So, we had a very nice banquet last night and everybody is so friendly. Everyone loves each other. Even if they're competitors, it doesn't matter.
I asked why. What's your explanation about why people are so friendly to each other? Even enemies.
I didn't say enemies. You said enemies. No, I think the why is actually I really believe this. I think we grew up together. When I say we, I think the entire team when I think about you know I see in the audience you know Emily and Jeff and many we grew up together right we grew up when as I said 10 years ago people didn't care so much about semiconductors you know people don't even understand what is high performance computing or what is rack scale integration but we were working hard at that point and now what I'm really happy to see is all of our friends and colleagues are so, so successful because the time is right. The time right now with the technology, the capability and then I think the spirit of the entire Taiwan ecosystem is such that there's so much success and I think that's a really good thing.
Yeah, it is good things. So let me back to the title of this summit is AI everywhere. So I think in your CES speech you mentioned you did you give a very, very big and bold expectation about you said that we will see 100 times increase in compute capacity and the user growth from 1 billion to 5 billion within five years. So could you explain what's the driving that level of expansions?
So I believe that AI is the most important technology that has occurred over let's call it the last 50 years and we've had a lot of other technologies like when the mobile phone came out we thought it was very special and when cloud computing came out we thought it was very special but AI is actually different than that. The reason AI is different I believe is we have found a technology that everybody can benefit from. Every single person can use it, every single company can use it, every single area can really benefit from AI. And so when I say AI everywhere, I think the vision that we see is that over the next three, five, 10 years, everything that we can do can be made better with AI and you should see AI everywhere in every device that we have in every application that we have and that type of proliferation of course it's good for business no question but I think it's really good for humanity and that's why I think AI everywhere, it's AI everywhere for everyone that I think should be our guiding principle for the technology area.
To create a new trend AI everywhere for everyone. So but I think people are curious about your expectation about the next phase of compute demands. We discussed a lot about cloud. How about edge compute? So cloud versus edge compute and also the training versus inferencing. Could you give us some explanation about that?
Yes. So if you think about two or three years ago when we first saw the beginning of the AI expansion, it was really a few companies that were really spending all of the effort and all of the money around AI for training. So training these large language models you know sort of ChatGPT or Claude or Gemini or any of the Chinese models required a lot of computing power for training. But the truth is nobody really makes money on training. You know training is necessary but it's not where you see the return on investment. Where you see the return on investment is people using the technology or inference. And inference is okay. You can ask this morning, you know, what's the weather this morning? You can ask these very simple questions and you start to see some inference. But that's not really the power. You said it earlier. I think what we're seeing now is that AI is not only entering inference, but it's entering the time when you can do really, really smart things. And you can ask the AI to solve complex problems. And those problems will actually help you significantly change the way you can build products, change the way you can run your companies, change the way we can discover science and all of these types of things. So I think that is a change that we're seeing. It's been very, very fast. That's what's different about the current cycle. Usually technology doesn't change so fast. Usually it changes in maybe a couple of years you see changes. Now we're seeing every couple of months the technology is different and the capability is different and the models are getting so much better. It's still very early in the cycle. It's still not good enough but every few months you're seeing so much progress that that's what's driving the demand.
Okay. So, in the introduction I say you have a forecast about eight GPUs to one GPU in the past. That's very technical for a finance journalist. Thank you. How about when will we see the one GPU to one CPU?
Actually, pretty soon. Pretty soon. I think the key thing if you think about GPUs or CPUs the details are not that important but the higher level point is everyone gets so excited about GPUs. The last few years, GPUs have grown so much. And my fundamental principle is when you're trying to enable AI everywhere, you need every kind of computing. And now you're starting to see that. So you're starting to see CPUs grow. You're starting to see also some ASICs grow. And as I said, the timing is actually pretty fast right now. I would say that we're seeing more and more agentic AI. The way to think about agentic AI, if I can give you a picture in your mind, is you have today you're a wonderful, very capable journalist, but you have to do your own research and maybe your ChatGPT or your AI can help you research Lisa Su. Actually, she did very good research. I think she knew everything I ever said ever 10 years. Even I did not remember who I talked to, but she told me that I said this to this person. But if you had 10 agents, then you can actually you don't you can do all of that at one time.
You can still have 10 agents or you could have when we think about what's the most exciting thing in our world is using AI to build chips faster. So right now for a very advanced chip it may take us two or three years from beginning to end. If I can cut that in half if we can cut that in half that means we can build many more products and satisfy much more of the demand in the market. So I think that's the power that we're looking for.
Thanks. I think next question I will going to ask about the bottleneck because of the supply chain is a shortage of the CPU and a lot of those HBM. What's the biggest bottleneck you think for the AI supply chain right now and how to address this issue?
Well, I think we see bottlenecks everywhere. Wouldn't you say just about everywhere? I say Emily not. But they're being solved quickly. So the thing about the semiconductor supply chain also is this is a very aggressive, very capable supply chain which once you see a bottleneck you find a way to solve it. So yes I think today there's a lot of concern about for example memory tends to be in a bit of a shortage today which is true. But there are ways to solve that as well. You know CPUs we're ramping up capacity very much. So we ask all of our partners to help us. Other bottlenecks include just power as we think about power in the various places you know we need to be able to build data centers and capabilities. I don't think it's any one bottleneck. I think together what we're seeing is that we didn't quite predict the demand to go up this fast and we're all so excited. We want to catch up to that demand and so we're working very hard to solve those bottlenecks and I think we're making great progress. So from our standpoint I think the entire ecosystem I have to say you know big thank you because everybody is making great progress to bring on more capacity and the foundation is very positive which is the world needs more AI computing and we're able to deliver that with the entire ecosystem together.
Since we mentioned a bottleneck, I think that three years ago you promote AI PC in Computex, but I think there's quite a delay a little bit. So I just wonder what's your forecast about AI PC in the future.
I'm actually very optimistic about AI PCs. I think with new technology, it always takes a little bit of time. So the beginning AI PCs you could not see the real application usage that easily. It was only for experts. But I would say today the most interesting thing about AI PCs is we all have our need for AI. But we also care a lot about privacy. We don't want everybody to see our personal information. So you don't want to send it somewhere. And you also want it to be local so you don't always have to connect to the cloud. And frankly the cloud can be very, very expensive if you think about it as well. So I think AI PCs for developers, for personal usage, for local enterprise usage will be a good growth vector over the next few years.
But beyond PC, which edge AI device are you mostly excited about?
Oh, I really think the next big wave is physical AI. So robotics and using all of the capability in the edge is actually also an opportunity whether you're talking about industrial and manufacturing environments or other local environments going forward.
As a journalist I think people always will always try to ask the question and get accurate answers but for you I feel like you always talk about the trends it's really hard to give an accurate timeline about when things... what kind of accurate answer would you like? For example we will see AI PC grows like 20% in next year or next two years something like that.
Let me think my PC head is sitting in the audience I should ask Jack what should we say. I think we will definitely see AI PC growth without a doubt. You know significant double-digit growth in AI PCs. The exact number is not so important. The important thing is are we investing in the right places? And what I will say for sure is nobody should doubt that AI PCs will be important and the more AI computing that we put at the edge the more we can get that AI everywhere vision to become a reality.
Yeah, maybe we can follow up with the social issues because you mentioned eventually we have the AI everywhere and even in the edge. So before that what kind of preparation you think the companies or the society can prepare for that especially for the employer or the company management.
Probably one of the things that I recommend when we think about how to use AI is we actually have to think about overall AI strategy from a high level not just a very low level. In other words some people think oh is AI going to take away people's jobs so should I be worried that on the employee side that your jobs will be less or on the business side oh I can save so much money that's two ways to look at it. I think that's actually the small part of AI. You know the big part of AI is how can I do something very differently and that means that you have to retrain your workforce. So for example at AMD we are actually educating every single person about how to use the AI tools and how to change the way we internally build our products or do our sales, do our marketing. I think every company has to think about how do you educate your employees? How do you ensure that you put the right tools in place? Of course, you know, AI is not always right. Actually, many times it's not right. So, you still have to have the right checks in place because at the end of the day, the company is responsible for what type of information they put out. And so, there's a lot of things that one has to think about.
