AMD CEO talks product pipeline in the 'AI super cycle'

Advanced Micro Devices (AMD) Chair and CEO Lisa Su sits down with Yahoo Finance executive editor Brian Sozzi to talk about the environment for AI chip manufacturers. The sector saw a shaky August as Nvidia (NVDA) and Broadcom (AVGO) both topped earnings estimates, but ultimately failed to wow Wall Street.

"I believe this is a market that will grow to $400 billion by 2027, which is huge. And it is an AI supercycle," Su tells Yahoo Finance at the Goldman Sachs Communacopia & Technology Conference. "We're making big bets now for the next five years. And the key is there's no one-size fits-all in this technology race. It's all about end-to-end AI."

Su expands upon the performance of AMD's latest hardware, calling the chipmaker's MI300X GPU accelerator "the fastest growing product that we've ever seen at AMD" since its launch last year. She cites AMD's "accelerated" AI roadmap as a catalyst for its product pipeline.

"When we look at how fast AI is moving and how fast the models are moving, we need to have the latest and greatest technology every year. So we actually are now on a one year cadence of new products. So later this year, we're going to announce a new product, our MI325. Next year we'll have an MI350 series, in 2026 an MI400 series. So you're going to get to know AMD products really well."

Su also touches on AMD's M&A activity in the last several years as the industry attempts to expand the AI chip infrastructure in the United States.

When asked about the state of the US chip infrastructure, Su says the chip sector is "in a place where we must have geographic diversity" as both public and private investors push more capital into the space.

Catch more Yahoo Finance interviews and coverage from the Goldman Sachs Communacopia & Technology Conference.

For more expert insight and the latest market action, click here to watch this full episode of Market Domination Overtime.

This post was written by Luke Carberry Mogan.

Video transcript

Thats it.

Welcome back to Yahoo Finance's coverage of the Goldman Sachs Communic Copia conference in San Francisco.

Really special guest right now in Yahoo Finance.

That is a MD chair.

And CEO Dr Lisa.

So, Lisa, good to see you.

I was telling my team I've been talking to you for 10 years.

Never seen you in person.

So good to see you.

It's great to see you as well, Brian.

Thanks for having me.

Of course.

Are you kidding me?

So I listened to your webcast.

You were just off your fireside chan.

I think how you ended it is how I want to start this conversation.

Uh, you reminded a lot of investors.

This is a super cycle for a I take us through that a little bit.

And And how do you see MA MD playing in that?

Yeah, so Absolutely.

I mean, we've, you know, really talked about what we think is the potential of, you know, a I in this, you know, accelerator area.

Um, I believe this is a market that will grow to, you know, $400 billion by 2027 which is huge, and it is an A I super cycle.

You know, sometimes people get caught on.

You know what's happening today?

What's happening tomorrow, the next quarter.

What I'd like to say is you have to take a step back and look as a tech company.

It's our job to make bets that are really for 3 to 5 years out.

And, you know, five years ago, we thought a I was going to be super important.

It's much more important today.

It's a much larger cycle today than I would have expected five years ago.

We're making big bets now for the next five years, and the key is there's no one size fits all in this technology race.

It's all about end to end a I.

So whether you're talking about the largest cloud environments or you're talking about what you do with the embedded in edge space or you're even talking about, you know a IP CS, you're going to need the right A. I computing for the right workload.

And that's our job at AD.

I like that you sent a message also to at the end of your presentation.

There's not just one company playing in this super cycle.

Talk to us about that.

That's absolutely true.

Because, you know, again, if you start with, uh, depending on your workload or you know which customer you're gonna need different types of compute.

And you know what we've learned dealing with our largest customers is, um this is about deep partnership of how you optimise hardware and software together.

So look, we're super excited about the progress we've made in a I, um our M I 300 X was just launched less than a year ago.

It's the fastest growing product that we've ever seen at a MD.

So this year, you know, last year was maybe less than $100 million of revenue in a I.

This year it will be greater than 4.5 billion.

So that's pretty nice.

But this is really just the beginning of that cycle, and we're super excited about what we see going forward.

Of course, that chip the new chip very powerful and very important.

What is the backlog look like for next year?

Well, the key for us is we've actually accelerated our a i road map.

So when we look at how fast a I is moving and how fast.

The models are moving.

We need to have the latest and greatest technology every year.

So we actually are now on a one year cadence of new products.

So later this year, we're going to announce a new product, our M. I 325.

Next year, we'll have an M I 350 series in 2026.

An M I 400 series.

So you're going to get to know a MD products Really Well, um, and the key is, uh we believe that, uh, you know, we can have, um a very significant piece of, you know, both, um, you know, training and inference of these, you know, large language models.

What are some of those newer chips that you just talked about?

How will they be different than what you have in the market?

Now?

You know, the key is more performance at lower cost.

I mean, that's the bottom line.

So for the largest data centre operators, you want to be able to train your models as fast as possible.

You want to be able to ask your model questions at, you know, at low cost?

This is all about total cost of ownership.

And the way we get there is we put more computing on a single chip and, you know, if you know, remember Brian.

One of the things that we've really pioneered is this idea that you could actually break up these large, powerful A I capabilities with something called triplets or, you know, really breaking it up into different technologies.

This is a huge advantage for us.

It's one of the reasons that we are, you know, really leading in some of the key workloads today.

And we look forward to where we'll be able to take this in the next generations.

What has surprised you about the build out of a I infrastructure in this country?

You know, what has actually really surprised me is just how fast the technology is moving.

I mean, what we are seeing is we are seeing whether you're a large hyper scaler or you're an enterprise customer or you're the newest A. I start up.

It's all about speed.

We're all learning at an incredibly fast pace, you know, even ourselves at a MD.

I would say that you know, we've learned more and done more in the last 9 to 12 months than in the previous number of years because the market is moving so fast and this is a place where innovation begets more innovation.

And so we're learning a tonne.

On the software side, we're learning a tonne about what is the most efficient way to get.

You know, these these products to market.

We've done a number of acquisitions, which I'm very excited about with all these deals.

Well, look, we're we're a growth company.

And as a growth company, our job is to invest and invest ahead of the curve.

We've done now three or four acquisitions, three in the last, Let's call it year.

And then we just announced a new one.

ZZZ.

Yeah, we've done Zilin was actually back in 2022.

If you can believe that, uh, we did XY links, we did pensando on the networking side.

They were very a I hardware and software centric.

Um, we've done several software acquisitions, mainly so that we could make it easier for our customers to adopt a MD architectures.

And then most recently last month, we announced our intent to acquire GT Systems, which builds out sort of the system infrastructure and the key is all of these pieces are important.

But we've been investing for a number of years, both organically and inorganically, so that we can be that computing partner of choice to those who need high performance.

Compute.

This is, of course, a very exciting time for all things A I but the supply chain is not diversified.

And you talked a little bit about this in your firestar chat here.

I mean, how do you and TS MC great partner?

It sounds like for for a MD.

But how do you How do you diversify from having all of your production in Taiwan?

Well, look, we're always thinking about supply chain resiliency.

I think we all have to, uh, just given, you know, the world's reliance on computing and high performance compute.

Um, you know, TS MC is a great partner for us.

Um, we do want more diversity in the supply chain.

We are actively both, you know, putting our tape outs or our new product designs into TS MC in Arizona as well as you know, looking forward to ramping those and we're going to continue to work with the overall ecosystem.

You know, I'm a big fan of the Chips Act and the work being done so that we could increase the domestic manufacturing, and we're going to continue to work on it as we go forward.

Is it an economic risk for the US to still have a lot of so much production overseas like this?

I think we're in a place where we must have geographic diversity.

And I think the, um you know, the work that has been done over the last few years.

Both, uh, chips act on manufacturing as well as, uh, Brian.

I'm a big fan of the R and D efforts because the work that's being done in the National Semiconductor Technology Centre to build a semiconductor, R and D footprint is extremely strategic.

So those things have to be done.

There's no you know, 11 thing that's going to make this all of a sudden better.

I think we need to keep working on investing, and, you know, this is about public private investment to really strengthen the semiconductor resiliency Before I let you go, what has it been like to just lead this transformation at a MD?

And I mean, how much longer do you want to be doing this.

I mean, you've You've been at a MD for so long, But you've led this amazing transformation, and I'm just curious.

On what?

What else do you have left to do?

Well, look, I am extremely fortunate.

Frankly, all of us in the technology industry are extremely fortunate that we are working on technology that matters.

Like, that's what gets me up every day is, you know, the products and the technology that we're working on, Uh, really does change the world really does impact.

Um, you know, sort of how we're gonna experience.

Um, services really does impact how businesses operate.

Uh, and, um, for that, like, there's no better place to be.

I mean, this is incredibly fun.

Uh, and I actually think you know, as fun as the last 10 years have been when you look at how technology has evolved, the next 10 are going to be much, much more exciting because you can really see we're unlocking sort of with the power of a I and computing.

We're unlocking things that we never believed were possible before really working on some amazing things over there at a MD AD chair and CL Lisa.

So let's, uh, hopefully not wait another 10 years to hang out in person again.

I appreciate it.