AMD CEO Lisa Su Talks New Data Center Chip, Demand

Bloomberg Audio Studios, Podcasts, radio News.

Welcome back to Bloomberg Tech Lisa Sue Helios with m I four to fifty five x AMD's first RAX scale system solution. But inside it AMD's first in the world's first two nanimeter.

Chip of that type.

A lot was made of it when you actually just stood on stage and held it in your hand for the first time.

Why is it significant?

Well, first of.

All, Ed, it's great to be here with you at CS. I think CS is always a great way to kick off the year because you get so much perspective. So it was fun giving the keynote last night. Look, Helios is.

A massive system, you can see it in the.

Background here, and m I four fifty five is just an incredibly powerful chip. And probably the context I would give Ed is, you know, one of the things that we're so clear about is that the demand for AI compute is just continuing to increase. And you know, we have seen that over the last five years. When you think about just how much you know, new capabilities have come on board, we've now seen a real inflection in

And a lot of the timeline for it to be deployed in the real world, and then who will be the principal first user of it.

You'll see it in the second half of twenty six and it will ramp from there. And you know, we have very strong partnerships open AI. Greg Brockman was on stage with us last night talking about all.

The use cases that they see.

We've announced a partnership with the Oracle, many others as well.

So given that it's two h it's in full production now, it's getting ready to are.

We are absolutely getting ready to ship it.

That's at one end of the sort of scale and spectrum. At the other you have m I four forty x, which is for small data centers. I'm trying to simplify it, but it's basically an enterprise product. What was it that you were trying to solve for with that?

Yeah, I think what we're trying to solve for is, you know, the world is a very heterogeneous world.

You have all kinds of use.

Cases for AI from you know, sort of the very biggest cloud data centers that are doing you know, large scale training and inference to enterprise applications as well as supercomputers, and so we actually have a family of chips.

At the highest end, is there.

M I four fifty five for the cloud environment, But we announced last night a m I four forty which is actually using the same basic building blocks, but is now really focused on enterprise applications so that you can go into you know, let's call it current data centers with the new technology.

So we're excited about that as well.

You know, there is enterprises are starting to increase their adoption of AI. In some cases they want their own control of their data centers in terms of on prem environments.

What are they doing with it though?

I mean, you know, we've been so fixated on frontier models with hundreds of billions of parameters and the scale of infrastructure needed for that with MI I F forty, we're talking about something slightly different. I think it's just really interesting if you could explain what the demand is from those enterprises, what they want with the technology.

Well, I think you see many enterprises now using AI all throughout their business processes, whether you're talking about things in their workflow. Even AMD we're using AI through every part of our development process. We're seeing a lot of applications in financial services and healthcare, and these are areas, especially in financial services, that people actually don't want everything

Welcome.

If you're watching us on Bloomberg Television or you're listening on Bloomberg Radio. We're live in Las Vegas and we're with AMD CEO Lisa Sue, and we're talking about the latest generation of accelerators. What makes this generation of AMD accelerators the better option, particularly for on prem and at the edge over what Nvidia is offering right now.

Well, the best way to think about it, ED is we're in this place where AI is at a inflection point. We're seeing AI now in every part of compute. We see it in the largest models, you know, when you're thinking about things like track, GPT and Gemini and Rock. You know, we're also seeing you know, many use cases in UH, you know, new capabilities like you know, video production, entertainment, healthcare, where you're doing drug discovery, all of these various things.

Ecosystem coming together.

So when you put those things in perspective, I think we have a great set of applications that will take advantage of these newest generationships.

You mentioned that Greg Brockman, who's the Opening Eye president, was on stage with you last night, and one of the basic points that he made was there are tools and functions they would love to release and put out into the world, but they are compute constrained. I often ask you to quantify demand, but is there a way to quantify the severity in the lack of compute, you know, the deficit that's out there right now.

Well, let me just give you some numbers to kind of ground what we think the demand environment is looking like. So if you think, you know, today we have about a billion active users and we're ramping that to you know, five billion over the next five years, and we have about let's call it, one hundred zet of flops of.

Compute you know, all around the world.

And that's just a generic number that that aggregates all of that. You know, we think we have to increase compute by another one hundred times as you go over the next you know, four or five years. And I introduced a term last night, the YadA flop. You know, people are like, what is a YadA flop? A YadA flop is actually ten to the twenty fourth in terms of flop.

So that's a one followed by twenty four zeros.

And to give you, you know, just a view of just how much things have really increased. I mean, that's another one hundred times more compute than we have today. So that gives you an idea. Now you think, what are you going to use all that compute for? I mean the truth is, the models that we have today are great. I mean they do amazing things. You know, we talked about a number of use cases. Uh you know, perhaps you know one that's you know, very hits very

But it's good, but it can get so much better. And I mean, I think that's the key point.

You know.

We we like to say that AI is really going to be everywhere, and it's really for everyone, and it's for each one of us to make our businesses more productive, you know, each one of us more productive, you know, going forward. And so we're still in the very early innings of really unlocking the power of AI.

So where we stand is we Okay, there's a compute deficit and software has kind of hit the limits of what current generation.

Compute can offer.

Help us understand the bottlenecks and barriers that to deploying that compute a lot at the moment about memory chips, what else, energy, electricity, what's crossing your desk, Lisa that gives you pause and say this is a problem right now.

Well, our job as a technology industry is to push the bleeding edge.

I mean, that is our job.

And so you know, when we think about like the four fifty five deploying you know two nanimeter and three and nanimeter chips, having the latest generation memory high bendw with memory that is out there, and really deploying these big systems, the important thing is that the entire ecosystem come together and we plan together for this next big inflection in compute.

And that's exactly what we're doing right now.

I think we're working very closely with the entire supply chain to ensure that we have the the resources to expand this compute environment. And yes, you know some of the things that you mentioned are let's call it constrained, but.

Which is most severely so, you know, I don't.

Think that's any one thing.

I think we're all looking at, you know, how do we build faster? You know, certainly power is one of those areas where you know, you see throughout the world, you know, power is being brought online as fast as possible, certainly from a silicon standpoint. You know, we're ramping our production capabilities with our partners. From a memory standpoint, our

We started this conversation talking about helios first, RAC scale architecture and infrastructure from AMD, could you talk about the future and how much of the content you want to own in a server?

You know, we start this story with the.

GPU, Frank k If you look at what Nvidia is doing, they want to increasingly own all of what's inside the box. Is that something that AMD's focused on too.

You know, what's most important for us is to ensure that we have turnkey solutions that are very very easy for our customers to deploy, because when you think about, you know, how do you use all of this AI compute most effectively?

You want it to go into the data.

Center and really be up and running on day one, and for that you have to optimize a full system. But from that standpoint, you know, we are very focused on an open ecosystem. So yes, we designed the CPUs and the GPUs and some of the networking elements, but we also work you know, really with the broad ecosystem of partners with industry standards. It's all about ensuring that we get the best of all worlds when we put our solutions together.

Looking ahead to m I five hundred twenty twenty seven, that has one thousand times the performance of the I three hundred generation. So your last generation of real well deployed gear. Something's coming that's a thousand times better. How did you make it a thousand times better?

It is just incredible engineering at every level. So m I four fifty five is ten times better than the trip that we just launched six months ago them I three fifty five and m I five hundred is another ten x you know. On top of that, we are using the most advanced technology out there. We have a very you know, very clear focus on you know, hardware, software, system code design, and it is you know, clearly the pushing the bleeding edge of capabilities.

What is the status of a m d's ability to sell products into China right now?

So you know, China is an important market for us.

You know, we actually sell a broad range of tips into China, including our you know, our PCs as well as you know other embedded chips.

In the data sets context of course, sorry, in the data.

Center context, we are you know, certainly we see China as an important market. We were We did get some licenses from the US government you know, late last year as it relates to some of our previous generation are m I three oh eight you know chips, and we are in the process of applying for new licenses with ourm I three twenty five chips that were recently allowed to apply for licenses. We haven't gotten those licenses yet, but we continue to view China as an important market.

For the reason I ask about it is in part because a lot of the work that's being done in open source models and bridging the gap between open and closed it is being done in China to some extent. There's been a lot of discussion about the demands being there in China, But could you reflect a little bit on that demand, but also what the Chinese government's attitude is to you taking a later gen neration of tech to the country.

Well, I do think the demand for you know, AI in general and in China is high for all the reasons that we talked about. I think we are in a demand environment where more compute is beneficial across the world. We think, you know, China is an important market for US, and it's very active in having our solutions deployed, so you know, we continue to view it as something that's important.

We're working with the US government as.

Well as our Chinese customers, you know, to find good solutions there.

And there are signs from both governments that the licensed process is moving. Commerce is kind of notorious for things sitting on a desk for quite a long time.

I think we are optimistic that, you know, we'll have an opportunity to get some of those licenses.

Granted you're watching Bloomberg Television, you're listening to Bloomberg Radio, this is Bloomberg Tech, and we live in Las Vegas with the A M D CEO, Lisa Sue. Last question, really in the data center context, is the markets and investors want data and signs that you're taking markets share. What would the metrics be that you'd point to, either that already exist or over the coming twelve months that would evidence that well.

I think M I four fifty five is a clear inflection point in both our technology capability as well as the deep partnerships that we have across the industry. So we're excited about, you know, what we see in front of us. And you know, we've talked about you know, tens of billions of dollars in an AI revenue as we get into twenty twenty seven, and I think these are important metrics, you know, for us as a company when we think about the AI potential.

For all the focus on data centers, some forget that a m D is leader in PC in many respects. The forecasters have very different opinions of what will happen this year. Some see, you know, shrinking market, some see modest growth driven literally by just AIPC. You've been able to tape market share and grow irrespective of what the broader conditions are, but it haven't been great. How have you done that and do you expect that to continue to be the case.

Well, the PC market is a very good market for us. You know, we grew a ton in the PC market in twenty twenty five, and that really came from the strength of our product portfolio.

We bet early in aipcs, so.

It was a clear area where we believe that the technology would generate demand. We also went through a refresh cycle with Windows eleven, and as we go into twenty twenty six, I think we'll want to see how a few quarters play out. I think the general demand for computing is certainly there. There are some supplies chain constraints that you know, we're working through and we want to

You know, we are very very strong in gaming, we're very strong in consumer.

I think we're under represented in enterprise laptops, and we view.

This as a growth area for US.

Is AI PC change that.

Aipcs absolutely help in terms of, you know, just the upgrade cycle coming in. We're excited about some of our work with AI development systems as well. We announced a new AI development system last night that you know, we think will be also very attractive.

Those constraints you were talking about in the PC context are specifically dram or it's broader than that.

It's more around the memory side.

So when you think about memory overall, I think we have so much demand coming from let's call it AI Data center compute that we want to see how it impacts you know, sort of the rest of the memory market out there.

One of the other areas that you discussed with Greg.

Brockman on stage from Open AI was sort of the net or broad economic impact of AI, not just for the companies. I think you were talking more about economy again, very difficulty, so that how does one measure progress in whether AI has or has not had a direct positive economic impact around the world in any given year.

You know, It's true, it's hard to deconvolve all of the things that are happening. But I think from a sense of you know, what we see in the business, and you know, many people want to see direct return on investment for a particular a set of investments, what I would say is that we know that AI.

Is making a difference in the productivity of companies.

We know that I can see that, you know, within a MD in terms of as we deploy AI, you know, we're able to get products to market faster, We're able to you know, significantly improve some of our business processes. So you know, as we go forward over the next several years, I think you're going to see that much broader enterprises. Every CEO that I talked to is talking

You're watching Bloomberg Television, you're listening to Bloomberg Radio. This is Bloomberg Tech and we're live in Las Vegas. We're speaking to A m d's CEO, Lisa Sue. You are an investor in Generative baronyx's also technology partner, and they have unveiled a humanoid robot here at Las Vegas cees. In fact, if the magic of television can happen, and we cut to the why.

Do we have to see it in the background? Right?

You know, this is the first tangible sign I feel we've seen from AM and D on how you intend to play in physical AI. Yes, explain your strategy. It is the next big market, right, Yes.

And I wouldn't say it's the first time, but it's probably one of the areas where we don't highlight as much because there's so much focused on data center and cloud and the opportunities there are you know, very much.

In front of us.

But when we look at physical AI, you know, starting from all of the work we've done in FPGAs and embedded real time you know capability, we have been in this space for a long time. You know, we already power a lot of robotic applications you know out there.

Is the business model to be all things the brain inside of the humanoid robot and inference side that, the underlying software being traded on a trained on a m D accelerators. Just I don't what's to go to the market, I guess is what I'm asking you should.

Expect that our partnerships extend all through all of those levels, so we have the components that can power the humanoid robots, you know, sort of real time local capability, which is a very very important and then we also have the technology behind that in terms of, you know, how to train and inference on these humanoids.

When last we met in person, it was in Washington, d C.

And the President had just outlined a broad strategy for America in AI and it really centered around infrastructure deregulation allowing those building the infrastructure to move faster. That was kind of in the second half of last year. In the months that have followed, have you seen any signs that it worked and anything that you could point to that says, yeah, people are able to build faster maybe to address some of the compute deficits we discussed.

Well, I can say for sure, you know, the President's AI action plan. You know, when we met, and I think this was back in July when it came out, I was very optimistic about having a really forward leaning strategy from you know, sort of the whole view of what does it take for the US to lead an AI. And I think we've made a ton of progress along

Across the controls.

Currently, we are certainly working.

Very closely with uh, you know, the the various parties in the US government to ensure that we have the right balance there. And we also have you know, this notion of how do we invest more here and ensure that in the United States that we are you know, running as fast as possible to bring you AI capacity you know online, to help us in you know, science and you know sort of the broader you know, economic benefits.

Lisa, what what happens in twenty twenty six, what happens in the world of AI, and what do you think will define this year in terms of the progress that your industry hopes to make.

Well, I started our keynote last night with the sentence that you know, you ain't seen nothing yet.

That's really how I feel.

I Mean, we're sitting here in January and it's just amazing how much progress is made, you know, every week and every month when we see how these models are developing, when we see how the use cases are developing, and then when we see the tangible results on businesses and outcomes, I believe that, you know, we saw a good amount of that, you know, come to Fruition in twenty twenty five. We're going to see much more of that in twenty twenty six, So that everyone should understand that, you know,

And I think we're going to see that in twenty twenty six.

Lisa suit AMD CEO AMD with it's in the world's first two animeter chip going into Helos, its first rack scale system solution