WEBVTT - AMD CEO Lisa Su Talks New Data Center Chip, Demand 0:00:02.480 --> 0:00:07.000 Bloomberg Audio Studios, Podcasts, radio News. 0:00:07.800 --> 0:00:13.239 Welcome back to Bloomberg Tech Lisa Sue Helios with m 0:00:13.320 --> 0:00:17.759 I four to fifty five x AMD's first RAX scale 0:00:17.800 --> 0:00:21.600 system solution. But inside it AMD's first in the world's 0:00:21.600 --> 0:00:23.759 first two nanimeter. 0:00:23.800 --> 0:00:25.000 Chip of that type. 0:00:25.560 --> 0:00:27.440 A lot was made of it when you actually just 0:00:27.440 --> 0:00:29.319 stood on stage and held it in your hand for 0:00:29.360 --> 0:00:30.000 the first time. 0:00:30.320 --> 0:00:31.400 Why is it significant? 0:00:31.520 --> 0:00:32.000 Well, first of. 0:00:31.960 --> 0:00:34.360 All, Ed, it's great to be here with you at CS. 0:00:34.440 --> 0:00:36.360 I think CS is always a great way to kick 0:00:36.360 --> 0:00:38.800 off the year because you get so much perspective. So 0:00:38.920 --> 0:00:42.360 it was fun giving the keynote last night. Look, Helios is. 0:00:42.280 --> 0:00:44.760 A massive system, you can see it in the. 0:00:44.680 --> 0:00:48.080 Background here, and m I four fifty five is just 0:00:48.400 --> 0:00:52.479 an incredibly powerful chip. And probably the context I would 0:00:52.680 --> 0:00:55.360 give Ed is, you know, one of the things that 0:00:55.840 --> 0:00:58.920 we're so clear about is that the demand for AI 0:00:59.000 --> 0:01:02.280 compute is just continuing to increase. And you know, we 0:01:02.360 --> 0:01:04.280 have seen that over the last five years. When you 0:01:04.319 --> 0:01:07.760 think about just how much you know, new capabilities have 0:01:07.840 --> 0:01:11.240 come on board, we've now seen a real inflection in 0:01:11.280 --> 0:01:13.800 the number of people who are using AI. So if 0:01:13.840 --> 0:01:16.440 you think today they're probably more than a billion active 0:01:16.520 --> 0:01:19.960 users using AI and we expect that to scale to 0:01:20.160 --> 0:01:22.640 over five billion users over the next five years. So 0:01:22.680 --> 0:01:25.160 for all of that, you need compute and lots and 0:01:25.200 --> 0:01:28.080 lots of compute. And from that standpoint, you know, m 0:01:28.120 --> 0:01:33.000 I four fifty five is a significant leap forward in 0:01:33.080 --> 0:01:36.160 terms of technology capability, made up of two and three 0:01:36.240 --> 0:01:40.440 nanometer trips, three hundred and twenty billion transistors, just a 0:01:40.480 --> 0:01:41.840 lot of performance. 0:01:41.360 --> 0:01:43.039 And a lot of the timeline for it to be 0:01:43.120 --> 0:01:45.440 deployed in the real world, and then who will be 0:01:45.480 --> 0:01:46.920 the principal first user of it. 0:01:47.080 --> 0:01:48.800 You'll see it in the second half of twenty six 0:01:49.040 --> 0:01:51.280 and it will ramp from there. And you know, we 0:01:51.400 --> 0:01:55.400 have very strong partnerships open AI. Greg Brockman was on 0:01:55.440 --> 0:01:58.880 stage with us last night talking about all. 0:01:58.840 --> 0:02:00.520 The use cases that they see. 0:02:00.920 --> 0:02:04.360 We've announced a partnership with the Oracle, many others as well. 0:02:04.480 --> 0:02:07.920 So given that it's two h it's in full production now, 0:02:07.960 --> 0:02:09.600 it's getting ready to are. 0:02:09.680 --> 0:02:11.920 We are absolutely getting ready to ship it. 0:02:12.360 --> 0:02:15.760 That's at one end of the sort of scale and spectrum. 0:02:16.080 --> 0:02:18.760 At the other you have m I four forty x, 0:02:18.840 --> 0:02:22.160 which is for small data centers. I'm trying to simplify it, 0:02:22.160 --> 0:02:26.320 but it's basically an enterprise product. What was it that 0:02:26.360 --> 0:02:28.400 you were trying to solve for with that? 0:02:28.840 --> 0:02:31.240 Yeah, I think what we're trying to solve for is, 0:02:31.520 --> 0:02:34.080 you know, the world is a very heterogeneous world. 0:02:34.760 --> 0:02:36.640 You have all kinds of use. 0:02:36.480 --> 0:02:39.200 Cases for AI from you know, sort of the very 0:02:39.240 --> 0:02:42.240 biggest cloud data centers that are doing you know, large 0:02:42.240 --> 0:02:47.440 scale training and inference to enterprise applications as well as supercomputers, 0:02:47.720 --> 0:02:49.639 and so we actually have a family of chips. 0:02:49.919 --> 0:02:51.359 At the highest end, is there. 0:02:51.400 --> 0:02:53.960 M I four fifty five for the cloud environment, But 0:02:54.040 --> 0:02:57.960 we announced last night a m I four forty which 0:02:58.000 --> 0:03:01.560 is actually using the same basic building blocks, but is 0:03:01.639 --> 0:03:05.240 now really focused on enterprise applications so that you can 0:03:05.280 --> 0:03:07.840 go into you know, let's call it current data centers 0:03:07.919 --> 0:03:09.080 with the new technology. 0:03:09.320 --> 0:03:10.880 So we're excited about that as well. 0:03:11.440 --> 0:03:14.400 You know, there is enterprises are starting to increase their 0:03:14.440 --> 0:03:17.480 adoption of AI. In some cases they want their own 0:03:17.800 --> 0:03:20.840 control of their data centers in terms of on prem environments. 0:03:21.000 --> 0:03:22.200 What are they doing with it though? 0:03:22.320 --> 0:03:25.960 I mean, you know, we've been so fixated on frontier 0:03:26.000 --> 0:03:29.200 models with hundreds of billions of parameters and the scale 0:03:29.200 --> 0:03:32.200 of infrastructure needed for that with MI I F forty, 0:03:32.480 --> 0:03:34.519 we're talking about something slightly different. I think it's just 0:03:34.560 --> 0:03:36.960 really interesting if you could explain what the demand is 0:03:37.480 --> 0:03:40.280 from those enterprises, what they want with the technology. 0:03:40.400 --> 0:03:43.960 Well, I think you see many enterprises now using AI 0:03:44.680 --> 0:03:49.960 all throughout their business processes, whether you're talking about things 0:03:49.960 --> 0:03:54.560 in their workflow. Even AMD we're using AI through every 0:03:54.640 --> 0:03:57.840 part of our development process. We're seeing a lot of 0:03:57.840 --> 0:04:02.280 applications in financial services and healthcare, and these are areas, 0:04:02.360 --> 0:04:06.000 especially in financial services, that people actually don't want everything 0:04:06.080 --> 0:04:08.200 necessarily in the cloud. They'd like to be able to 0:04:08.240 --> 0:04:13.280 have their own on prem deployment or private cloud deployments. 0:04:13.560 --> 0:04:15.440 And in this case, you don't want to have to 0:04:15.480 --> 0:04:18.320 build a brand new data center for every new generation 0:04:18.400 --> 0:04:21.080 of chip. M four point forty allows us to use 0:04:21.120 --> 0:04:24.680 some of those existing data centers in upgrade with the 0:04:24.680 --> 0:04:25.680 new capabilities. 0:04:26.160 --> 0:04:26.600 Welcome. 0:04:26.760 --> 0:04:29.360 If you're watching us on Bloomberg Television or you're listening 0:04:29.400 --> 0:04:32.240 on Bloomberg Radio. We're live in Las Vegas and we're 0:04:32.240 --> 0:04:36.280 with AMD CEO Lisa Sue, and we're talking about the 0:04:36.360 --> 0:04:42.000 latest generation of accelerators. What makes this generation of AMD 0:04:42.120 --> 0:04:46.440 accelerators the better option, particularly for on prem and at 0:04:46.480 --> 0:04:49.200 the edge over what Nvidia is offering right now. 0:04:49.360 --> 0:04:52.200 Well, the best way to think about it, ED is 0:04:52.839 --> 0:04:57.880 we're in this place where AI is at a inflection point. 0:04:57.960 --> 0:05:01.360 We're seeing AI now in every part of compute. We 0:05:01.400 --> 0:05:04.520 see it in the largest models, you know, when you're 0:05:04.560 --> 0:05:07.279 thinking about things like track, GPT and Gemini and Rock. 0:05:07.640 --> 0:05:10.240 You know, we're also seeing you know, many use cases 0:05:10.440 --> 0:05:17.000 in UH, you know, new capabilities like you know, video production, entertainment, healthcare, 0:05:17.160 --> 0:05:20.320 where you're doing drug discovery, all of these various things. 0:05:20.480 --> 0:05:23.479 You know, our claim to fame is really you know, 0:05:23.680 --> 0:05:28.080 outstanding performance at you know, very advantage total cost of ownership. 0:05:28.360 --> 0:05:30.280 And the other thing that you know we believe very 0:05:30.320 --> 0:05:34.560 strongly in is an open ecosystem and deep partnerships, you know, 0:05:34.640 --> 0:05:37.400 with our UH you know, with our overall. 0:05:37.240 --> 0:05:38.560 Ecosystem coming together. 0:05:38.839 --> 0:05:41.560 So when you put those things in perspective, I think 0:05:41.600 --> 0:05:46.640 we have a great set of applications that will take 0:05:46.640 --> 0:05:48.600 advantage of these newest generationships. 0:05:48.960 --> 0:05:51.920 You mentioned that Greg Brockman, who's the Opening Eye president, 0:05:51.960 --> 0:05:53.960 was on stage with you last night, and one of 0:05:53.800 --> 0:05:56.400 the basic points that he made was there are tools 0:05:56.440 --> 0:05:59.600 and functions they would love to release and put out 0:05:59.640 --> 0:06:02.960 into the world, but they are compute constrained. I often 0:06:03.000 --> 0:06:05.960 ask you to quantify demand, but is there a way 0:06:06.000 --> 0:06:10.000 to quantify the severity in the lack of compute, you know, 0:06:10.080 --> 0:06:12.040 the deficit that's out there right now. 0:06:12.120 --> 0:06:14.240 Well, let me just give you some numbers to kind 0:06:14.279 --> 0:06:18.320 of ground what we think the demand environment is looking like. 0:06:18.400 --> 0:06:21.719 So if you think, you know, today we have about 0:06:21.720 --> 0:06:24.719 a billion active users and we're ramping that to you know, 0:06:24.800 --> 0:06:28.120 five billion over the next five years, and we have 0:06:28.200 --> 0:06:30.560 about let's call it, one hundred zet of flops of. 0:06:30.560 --> 0:06:32.360 Compute you know, all around the world. 0:06:32.440 --> 0:06:36.640 And that's just a generic number that that aggregates all 0:06:36.640 --> 0:06:38.679 of that. You know, we think we have to increase 0:06:38.720 --> 0:06:43.680 compute by another one hundred times as you go over 0:06:43.720 --> 0:06:46.520 the next you know, four or five years. And I 0:06:46.560 --> 0:06:50.440 introduced a term last night, the YadA flop. You know, 0:06:50.440 --> 0:06:52.800 people are like, what is a YadA flop? A YadA 0:06:52.800 --> 0:06:56.800 flop is actually ten to the twenty fourth in terms 0:06:56.800 --> 0:06:57.159 of flop. 0:06:57.200 --> 0:06:59.320 So that's a one followed by twenty four zeros. 0:06:59.800 --> 0:07:02.880 And to give you, you know, just a view of 0:07:02.920 --> 0:07:05.720 just how much things have really increased. I mean, that's 0:07:05.760 --> 0:07:09.440 another one hundred times more compute than we have today. 0:07:09.760 --> 0:07:11.840 So that gives you an idea. Now you think, what 0:07:11.880 --> 0:07:13.680 are you going to use all that compute for? I 0:07:13.720 --> 0:07:16.560 mean the truth is, the models that we have today 0:07:16.640 --> 0:07:19.640 are great. I mean they do amazing things. You know, 0:07:19.680 --> 0:07:22.240 we talked about a number of use cases. Uh you know, 0:07:22.280 --> 0:07:24.600 perhaps you know one that's you know, very hits very 0:07:24.600 --> 0:07:27.480 close to home is is writing software. Like you know, 0:07:27.520 --> 0:07:30.960 people are using uh, the AI tools right now to 0:07:31.160 --> 0:07:34.520 significantly enhance the productivity of software developers. 0:07:34.880 --> 0:07:37.880 But it's good, but it can get so much better. 0:07:38.000 --> 0:07:39.760 And I mean, I think that's the key point. 0:07:40.040 --> 0:07:40.240 You know. 0:07:40.240 --> 0:07:43.880 We we like to say that AI is really going 0:07:43.920 --> 0:07:46.640 to be everywhere, and it's really for everyone, and it's 0:07:46.640 --> 0:07:50.520 for each one of us to make our businesses more productive, 0:07:50.720 --> 0:07:53.600 you know, each one of us more productive, you know, 0:07:53.680 --> 0:07:56.080 going forward. And so we're still in the very early 0:07:56.120 --> 0:07:58.680 innings of really unlocking the power of AI. 0:07:58.840 --> 0:08:01.440 So where we stand is we Okay, there's a compute 0:08:01.440 --> 0:08:04.280 deficit and software has kind of hit the limits of 0:08:04.280 --> 0:08:05.600 what current generation. 0:08:05.320 --> 0:08:06.760 Compute can offer. 0:08:07.240 --> 0:08:10.920 Help us understand the bottlenecks and barriers that to deploying 0:08:11.000 --> 0:08:13.920 that compute a lot at the moment about memory chips, 0:08:14.640 --> 0:08:19.880 what else, energy, electricity, what's crossing your desk, Lisa that 0:08:20.840 --> 0:08:24.360 gives you pause and say this is a problem right now. 0:08:24.560 --> 0:08:28.440 Well, our job as a technology industry is to push 0:08:28.480 --> 0:08:29.240 the bleeding edge. 0:08:29.320 --> 0:08:30.280 I mean, that is our job. 0:08:30.600 --> 0:08:32.760 And so you know, when we think about like the 0:08:33.040 --> 0:08:35.640 four fifty five deploying you know two nanimeter and three 0:08:35.640 --> 0:08:39.439 and nanimeter chips, having the latest generation memory high bendw 0:08:39.440 --> 0:08:43.120 with memory that is out there, and really deploying these 0:08:43.120 --> 0:08:47.120 big systems, the important thing is that the entire ecosystem 0:08:47.200 --> 0:08:50.760 come together and we plan together for this next big 0:08:50.800 --> 0:08:52.280 inflection in compute. 0:08:52.400 --> 0:08:54.840 And that's exactly what we're doing right now. 0:08:55.000 --> 0:08:57.520 I think we're working very closely with the entire supply 0:08:57.600 --> 0:09:01.960 chain to ensure that we have the the resources to 0:09:02.080 --> 0:09:06.480 expand this compute environment. And yes, you know some of 0:09:06.520 --> 0:09:10.480 the things that you mentioned are let's call it constrained, but. 0:09:10.480 --> 0:09:13.480 Which is most severely so, you know, I don't. 0:09:13.200 --> 0:09:14.640 Think that's any one thing. 0:09:14.760 --> 0:09:16.600 I think we're all looking at, you know, how do 0:09:16.640 --> 0:09:20.559 we build faster? You know, certainly power is one of 0:09:20.640 --> 0:09:23.000 those areas where you know, you see throughout the world, 0:09:23.640 --> 0:09:26.920 you know, power is being brought online as fast as possible, 0:09:27.520 --> 0:09:30.360 certainly from a silicon standpoint. You know, we're ramping our 0:09:30.400 --> 0:09:33.920 production capabilities with our partners. From a memory standpoint, our 0:09:33.960 --> 0:09:37.559 partners ramping as well. So it's not any one thing, 0:09:37.640 --> 0:09:39.480 I think, it's all of these things have to go 0:09:40.280 --> 0:09:43.120 sort of in tandem, and that's why partnership is just 0:09:43.240 --> 0:09:45.439 so important in this business. 0:09:45.760 --> 0:09:49.560 We started this conversation talking about helios first, RAC scale 0:09:50.520 --> 0:09:54.840 architecture and infrastructure from AMD, could you talk about the 0:09:54.880 --> 0:09:57.200 future and how much of the content you want to 0:09:57.240 --> 0:09:58.319 own in a server? 0:09:58.880 --> 0:10:02.240 You know, we start this story with the. 0:10:02.200 --> 0:10:05.400 GPU, Frank k If you look at what Nvidia is doing, 0:10:05.480 --> 0:10:08.480 they want to increasingly own all of what's inside the box. 0:10:08.960 --> 0:10:10.920 Is that something that AMD's focused on too. 0:10:11.160 --> 0:10:14.480 You know, what's most important for us is to ensure 0:10:14.600 --> 0:10:18.480 that we have turnkey solutions that are very very easy 0:10:18.520 --> 0:10:21.040 for our customers to deploy, because when you think about, 0:10:21.320 --> 0:10:23.839 you know, how do you use all of this AI 0:10:23.920 --> 0:10:25.320 compute most effectively? 0:10:25.720 --> 0:10:27.680 You want it to go into the data. 0:10:27.480 --> 0:10:30.080 Center and really be up and running on day one, 0:10:30.160 --> 0:10:32.520 and for that you have to optimize a full system. 0:10:32.880 --> 0:10:35.800 But from that standpoint, you know, we are very focused 0:10:35.840 --> 0:10:39.280 on an open ecosystem. So yes, we designed the CPUs 0:10:39.320 --> 0:10:41.720 and the GPUs and some of the networking elements, but 0:10:41.760 --> 0:10:44.800 we also work you know, really with the broad ecosystem 0:10:44.840 --> 0:10:49.360 of partners with industry standards. It's all about ensuring that 0:10:49.720 --> 0:10:51.880 we get the best of all worlds when we put 0:10:52.240 --> 0:10:53.280 our solutions together. 0:10:53.960 --> 0:10:57.560 Looking ahead to m I five hundred twenty twenty seven, 0:10:58.600 --> 0:11:02.560 that has one thousand times the performance of the I 0:11:02.559 --> 0:11:05.719 three hundred generation. So your last generation of real well 0:11:05.760 --> 0:11:10.160 deployed gear. Something's coming that's a thousand times better. How 0:11:10.200 --> 0:11:12.079 did you make it a thousand times better? 0:11:12.440 --> 0:11:16.920 It is just incredible engineering at every level. So m 0:11:16.960 --> 0:11:20.400 I four fifty five is ten times better than the 0:11:20.480 --> 0:11:23.400 trip that we just launched six months ago them I 0:11:23.440 --> 0:11:26.679 three fifty five and m I five hundred is another 0:11:26.880 --> 0:11:29.439 ten x you know. On top of that, we are 0:11:29.520 --> 0:11:32.480 using the most advanced technology out there. We have a 0:11:33.120 --> 0:11:37.800 very you know, very clear focus on you know, hardware, software, 0:11:38.520 --> 0:11:42.640 system code design, and it is you know, clearly the 0:11:42.840 --> 0:11:44.800 pushing the bleeding edge of capabilities. 0:11:45.520 --> 0:11:48.840 What is the status of a m d's ability to 0:11:48.960 --> 0:11:50.960 sell products into China right now? 0:11:51.920 --> 0:11:54.240 So you know, China is an important market for us. 0:11:54.679 --> 0:11:56.600 You know, we actually sell a broad range of tips 0:11:56.600 --> 0:12:00.080 into China, including our you know, our PCs as well 0:12:00.120 --> 0:12:03.160 as you know other embedded chips. 0:12:02.559 --> 0:12:05.600 In the data sets context of course, sorry, in the data. 0:12:05.400 --> 0:12:08.520 Center context, we are you know, certainly we see China 0:12:08.559 --> 0:12:11.560 as an important market. We were We did get some 0:12:11.640 --> 0:12:14.840 licenses from the US government you know, late last year 0:12:15.080 --> 0:12:17.920 as it relates to some of our previous generation are 0:12:18.080 --> 0:12:20.720 m I three oh eight you know chips, and we 0:12:20.760 --> 0:12:24.000 are in the process of applying for new licenses with 0:12:24.200 --> 0:12:30.440 ourm I three twenty five chips that were recently allowed 0:12:30.559 --> 0:12:33.600 to apply for licenses. We haven't gotten those licenses yet, 0:12:33.760 --> 0:12:36.520 but we continue to view China as an important market. 0:12:36.559 --> 0:12:38.560 For the reason I ask about it is in part 0:12:38.640 --> 0:12:41.200 because a lot of the work that's being done in 0:12:41.240 --> 0:12:44.240 open source models and bridging the gap between open and 0:12:44.280 --> 0:12:46.800 closed it is being done in China to some extent. 0:12:47.880 --> 0:12:49.800 There's been a lot of discussion about the demands being 0:12:49.840 --> 0:12:52.120 there in China, But could you reflect a little bit 0:12:52.120 --> 0:12:55.520 on that demand, but also what the Chinese government's attitude 0:12:55.600 --> 0:13:00.680 is to you taking a later gen neration of tech 0:13:00.720 --> 0:13:01.319 to the country. 0:13:01.679 --> 0:13:04.520 Well, I do think the demand for you know, AI 0:13:04.600 --> 0:13:07.600 in general and in China is high for all the 0:13:07.640 --> 0:13:09.480 reasons that we talked about. I think we are in 0:13:09.520 --> 0:13:14.880 a demand environment where more compute is beneficial across the world. 0:13:15.320 --> 0:13:17.600 We think, you know, China is an important market for US, 0:13:17.720 --> 0:13:20.839 and it's very active in having our solutions deployed, so 0:13:21.080 --> 0:13:23.520 you know, we continue to view it as something that's important. 0:13:23.840 --> 0:13:25.520 We're working with the US government as. 0:13:25.440 --> 0:13:28.240 Well as our Chinese customers, you know, to find good 0:13:28.240 --> 0:13:29.280 solutions there. 0:13:29.200 --> 0:13:31.840 And there are signs from both governments that the licensed 0:13:31.840 --> 0:13:34.800 process is moving. Commerce is kind of notorious for things 0:13:34.840 --> 0:13:36.480 sitting on a desk for quite a long time. 0:13:36.559 --> 0:13:39.360 I think we are optimistic that, you know, we'll have 0:13:39.400 --> 0:13:41.920 an opportunity to get some of those licenses. 0:13:41.920 --> 0:13:45.840 Granted you're watching Bloomberg Television, you're listening to Bloomberg Radio, 0:13:45.960 --> 0:13:48.320 this is Bloomberg Tech, and we live in Las Vegas 0:13:48.679 --> 0:13:53.160 with the A M D CEO, Lisa Sue. Last question, 0:13:53.200 --> 0:13:56.080 really in the data center context, is the markets and 0:13:56.120 --> 0:14:00.360 investors want data and signs that you're taking markets share. 0:14:00.880 --> 0:14:03.800 What would the metrics be that you'd point to, either 0:14:03.840 --> 0:14:06.720 that already exist or over the coming twelve months that 0:14:06.760 --> 0:14:08.120 would evidence that well. 0:14:08.200 --> 0:14:12.280 I think M I four fifty five is a clear 0:14:12.360 --> 0:14:16.440 inflection point in both our technology capability as well as 0:14:16.440 --> 0:14:19.680 the deep partnerships that we have across the industry. So 0:14:19.720 --> 0:14:21.960 we're excited about, you know, what we see in front 0:14:22.000 --> 0:14:25.120 of us. And you know, we've talked about you know, 0:14:25.200 --> 0:14:29.360 tens of billions of dollars in an AI revenue as 0:14:29.400 --> 0:14:31.960 we get into twenty twenty seven, and I think these 0:14:31.960 --> 0:14:34.520 are important metrics, you know, for us as a company 0:14:35.120 --> 0:14:37.240 when we think about the AI potential. 0:14:38.360 --> 0:14:41.160 For all the focus on data centers, some forget that 0:14:41.360 --> 0:14:44.920 a m D is leader in PC in many respects. 0:14:46.600 --> 0:14:50.680 The forecasters have very different opinions of what will happen 0:14:50.720 --> 0:14:54.640 this year. Some see, you know, shrinking market, some see 0:14:54.920 --> 0:14:59.560 modest growth driven literally by just AIPC. You've been able 0:14:59.600 --> 0:15:03.120 to tape market share and grow irrespective of what the 0:15:03.120 --> 0:15:06.400 broader conditions are, but it haven't been great. How have 0:15:06.520 --> 0:15:09.640 you done that and do you expect that to continue 0:15:09.640 --> 0:15:10.360 to be the case. 0:15:10.840 --> 0:15:14.560 Well, the PC market is a very good market for us. 0:15:14.640 --> 0:15:16.640 You know, we grew a ton in the PC market 0:15:16.800 --> 0:15:20.400 in twenty twenty five, and that really came from the 0:15:20.400 --> 0:15:22.000 strength of our product portfolio. 0:15:22.400 --> 0:15:24.560 We bet early in aipcs, so. 0:15:24.560 --> 0:15:27.560 It was a clear area where we believe that the 0:15:27.560 --> 0:15:31.200 technology would generate demand. We also went through a refresh 0:15:31.280 --> 0:15:34.280 cycle with Windows eleven, and as we go into twenty 0:15:34.320 --> 0:15:37.080 twenty six, I think we'll want to see how a 0:15:37.120 --> 0:15:41.360 few quarters play out. I think the general demand for 0:15:41.720 --> 0:15:45.560 computing is certainly there. There are some supplies chain constraints 0:15:45.600 --> 0:15:47.360 that you know, we're working through and we want to 0:15:47.520 --> 0:15:50.280 you know, watch going forward. But you know, our case 0:15:50.400 --> 0:15:54.440 is one where we are still underrepresented in parts of 0:15:54.440 --> 0:15:55.000 the market. 0:15:55.160 --> 0:15:57.760 You know, we are very very strong in gaming, we're 0:15:57.840 --> 0:15:58.880 very strong in consumer. 0:15:59.120 --> 0:16:03.080 I think we're under represented in enterprise laptops, and we view. 0:16:02.920 --> 0:16:04.400 This as a growth area for US. 0:16:04.560 --> 0:16:06.000 Is AI PC change that. 0:16:06.160 --> 0:16:11.600 Aipcs absolutely help in terms of, you know, just the 0:16:11.720 --> 0:16:15.760 upgrade cycle coming in. We're excited about some of our 0:16:15.800 --> 0:16:19.160 work with AI development systems as well. We announced a 0:16:19.200 --> 0:16:22.280 new AI development system last night that you know, we 0:16:22.280 --> 0:16:24.080 think will be also very attractive. 0:16:24.400 --> 0:16:27.240 Those constraints you were talking about in the PC context 0:16:27.560 --> 0:16:29.960 are specifically dram or it's broader than that. 0:16:30.360 --> 0:16:31.720 It's more around the memory side. 0:16:31.760 --> 0:16:34.720 So when you think about memory overall, I think we 0:16:34.840 --> 0:16:38.280 have so much demand coming from let's call it AI 0:16:38.400 --> 0:16:41.480 Data center compute that we want to see how it 0:16:41.560 --> 0:16:43.480 impacts you know, sort of the rest of the memory 0:16:43.520 --> 0:16:44.720 market out there. 0:16:45.200 --> 0:16:48.800 One of the other areas that you discussed with Greg. 0:16:48.600 --> 0:16:52.040 Brockman on stage from Open AI was sort of the 0:16:52.120 --> 0:16:56.000 net or broad economic impact of AI, not just for 0:16:56.840 --> 0:17:01.760 the companies. I think you were talking more about economy again, 0:17:02.040 --> 0:17:06.119 very difficulty, so that how does one measure progress in 0:17:06.200 --> 0:17:10.000 whether AI has or has not had a direct positive 0:17:10.000 --> 0:17:12.400 economic impact around the world in any given year. 0:17:13.000 --> 0:17:15.720 You know, It's true, it's hard to deconvolve all of 0:17:15.800 --> 0:17:18.320 the things that are happening. But I think from a 0:17:18.440 --> 0:17:21.520 sense of you know, what we see in the business, 0:17:22.240 --> 0:17:25.080 and you know, many people want to see direct return 0:17:25.119 --> 0:17:29.199 on investment for a particular a set of investments, what 0:17:29.280 --> 0:17:33.159 I would say is that we know that AI. 0:17:33.080 --> 0:17:35.160 Is making a difference in the productivity of companies. 0:17:35.200 --> 0:17:37.280 We know that I can see that, you know, within 0:17:37.320 --> 0:17:40.639 a MD in terms of as we deploy AI, you know, 0:17:40.680 --> 0:17:43.440 we're able to get products to market faster, We're able 0:17:43.440 --> 0:17:46.800 to you know, significantly improve some of our business processes. 0:17:46.880 --> 0:17:49.280 So you know, as we go forward over the next 0:17:49.640 --> 0:17:51.360 several years, I think you're going to see that much 0:17:51.400 --> 0:17:55.400 broader enterprises. Every CEO that I talked to is talking 0:17:55.440 --> 0:17:58.320 about AI. It is front and center in terms of 0:17:58.359 --> 0:18:00.199 how to build a better company, how to build the 0:18:00.240 --> 0:18:03.720 better portfolio, and so you know, I think what you 0:18:03.760 --> 0:18:06.680 know Greg was talking about is when you aggregate all 0:18:06.720 --> 0:18:11.919 of that, AI has to impact you know the world 0:18:12.040 --> 0:18:15.600 at a GDP level, and we'll see that over the 0:18:15.600 --> 0:18:16.280 next few years. 0:18:16.880 --> 0:18:20.200 You're watching Bloomberg Television, you're listening to Bloomberg Radio. This 0:18:20.240 --> 0:18:22.760 is Bloomberg Tech and we're live in Las Vegas. We're 0:18:22.760 --> 0:18:26.640 speaking to A m d's CEO, Lisa Sue. You are 0:18:26.720 --> 0:18:32.000 an investor in Generative baronyx's also technology partner, and they 0:18:32.000 --> 0:18:35.320 have unveiled a humanoid robot here at Las Vegas cees. 0:18:35.440 --> 0:18:37.760 In fact, if the magic of television can happen, and 0:18:37.800 --> 0:18:38.440 we cut to the why. 0:18:38.400 --> 0:18:40.320 Do we have to see it in the background? Right? 0:18:40.880 --> 0:18:46.440 You know, this is the first tangible sign I feel 0:18:46.440 --> 0:18:48.560 we've seen from AM and D on how you intend 0:18:48.560 --> 0:18:52.840 to play in physical AI. Yes, explain your strategy. It 0:18:52.960 --> 0:18:55.360 is the next big market, right, Yes. 0:18:55.160 --> 0:18:58.040 And I wouldn't say it's the first time, but it's 0:18:58.080 --> 0:19:01.439 probably one of the areas where we don't highlight as 0:19:01.520 --> 0:19:04.000 much because there's so much focused on data center and 0:19:04.040 --> 0:19:07.760 cloud and the opportunities there are you know, very much. 0:19:07.600 --> 0:19:08.240 In front of us. 0:19:08.359 --> 0:19:10.520 But when we look at physical AI, you know, starting 0:19:10.600 --> 0:19:14.359 from all of the work we've done in FPGAs and 0:19:14.440 --> 0:19:17.560 embedded real time you know capability, we have been in 0:19:17.600 --> 0:19:19.680 this space for a long time. You know, we already 0:19:19.720 --> 0:19:23.280 power a lot of robotic applications you know out there. 0:19:23.760 --> 0:19:27.880 But I think as we go into the humanoid capability, 0:19:27.920 --> 0:19:30.719 and you know, we're excited about our partnership with you know, 0:19:31.040 --> 0:19:34.840 g Bionics and the work with on June one, I 0:19:34.840 --> 0:19:37.240 think that takes us to another level in terms of 0:19:37.359 --> 0:19:40.560 capability and intelligence and what we're trying to do so. 0:19:40.840 --> 0:19:43.280 Is the business model to be all things the brain 0:19:43.400 --> 0:19:46.600 inside of the humanoid robot and inference side that, the 0:19:46.680 --> 0:19:49.200 underlying software being traded on a trained on a m 0:19:49.280 --> 0:19:51.840 D accelerators. Just I don't what's to go to the market, 0:19:51.840 --> 0:19:53.800 I guess is what I'm asking you should. 0:19:53.560 --> 0:19:57.840 Expect that our partnerships extend all through all of those levels, 0:19:57.840 --> 0:20:02.320 so we have the components that can power the humanoid robots, 0:20:02.359 --> 0:20:06.040 you know, sort of real time local capability, which is 0:20:06.080 --> 0:20:08.119 a very very important and then we also have the 0:20:08.520 --> 0:20:11.040 technology behind that in terms of, you know, how to 0:20:11.080 --> 0:20:14.880 train and inference on these humanoids. 0:20:15.320 --> 0:20:18.199 When last we met in person, it was in Washington, 0:20:18.280 --> 0:20:18.480 d C. 0:20:19.160 --> 0:20:24.040 And the President had just outlined a broad strategy for 0:20:24.200 --> 0:20:29.760 America in AI and it really centered around infrastructure deregulation 0:20:30.119 --> 0:20:34.280 allowing those building the infrastructure to move faster. That was 0:20:34.359 --> 0:20:36.600