AMD CEO Lisa Su on New AI Chips, Demand and Computing

Lisa Su
2 followers
AMD CEO and President Lisa Su discusses the demand for the company's AI products with Ed Ludlow on "Bloomberg The Close." AMD suffered its biggest stock decline in more than a month after the company unveiled new AI chips but didn’t provide hoped-for information on customers or financial performance.
#lisasu; #amd
From robo-taxis to AI chips, Bloomberg's Ed Ludlow is standing by with Advanced Micro Devices CEO
Lisa Su, who joins him live from AMD's Advancing AI event in California, where they just unveiled
their latest generation of chips. Ed? That's right, Scarlett, and Lisa Su celebrating 10 years
as CEO of AMD. Lisa, MI325X, everyone I speak to in the world of technology,
accepts and sees that AMD is the number two rival behind NVIDIA. How does the latest accelerator
change that competitive landscape in that market? Well, first of all, Ed, it's great to be here with
you today. Thank you for having us. It was a big day for us. We talked about so many new technologies.
We launched a number of new products. We launched our new fifth-gen EPYC server product called
Turin. We launched MI325 and talked about our future roadmap and MI350 and beyond. We talked
about AI PCs and networking, so it's been a great day. As it relates to, you know, MI325 and what
we're doing in the data center accelerator market, I mean, this is such an exciting market right now.
There's so much demand overall for new AI capability and, you know, we are doing extremely
well. We had a number of customers on stage with us today, too, including Microsoft, Meta, Oracle,
a number of new, you know, AI companies, as well as all of the major OEMs, and everyone is excited
about MI325 and the roadmap ahead. And yet, shares closed down 4%. The disappointment is not in the
long term, right? You updated the forecast for the market, $500 billion in 2028. But what can
you tell me, or how can you quantify the near-term demand of those new customers that you just outlined?
Well, you know, the best thing to say, Ed, is this is an incredibly fast-growing market. I mean, if you
think about, you know, what we've done just over the last 12 months or so, we've seen MI300 demand
increase from when we started the year to in our most, you know, recent earnings, we actually said
we would be over $4.5 billion in this important AI accelerator market. You know, we view this as
a multi-year opportunity for us. And what's really important, you know, what our customers say is,
you know, we've shown that we have excellent hardware, but even more importantly, we've made
tremendous progress in the overall software roadmap. And what that allows is that allows
customers to get to market, you know, as fast as possible. So we're seeing, you know, with MI325,
we expect the new chips to be anywhere from 20% to 40% higher performance on key inferencing
workloads than the competition. And that's a great place to be.
Lisa, describe the demand for your AI products in just a single word.
Amazing.
Give me more than that. How is this different to a year ago? You remember us sitting in San Jose,
right? We were sitting together, and it was the start of a very different time in the investment
cycle. How is it different now?
Well, I think, so in a single word, the best thing to say is it's an accelerating market. And,
you know, what we see is so many new applications, so many new opportunities to use AI.
We're at the very beginning of the cycle. And one of the things that I've been very,
very encouraged at is, you know, when we started maybe 12 months ago, you would see that,
you know, there would be a lot of people trying things, but it would actually take them a while
to cut over to their production environment. That's actually much faster now. And what that
means is, you know, there's more demand for AI capability, but we've also significantly improved
our tools and overall software infrastructure. You know, we call our software suite something
called Rockum, and it has just made tremendous progress over the last 10, 11 months that really
allows customers to adopt our solutions, you know, much, much faster. And that's the key for
innovation is to try to get as much in as fast as possible. And CPU as well. If you're an investor
on Wall Street, or you are a retail investor, or you're a technologist, or dare I say it,
a technology journalist, everyone is so over indexed to GPU. I really want to know if you see
a sort of revision back towards emphasis on CPU, with all of the data center investment
and build out that is to come. I think the most important thing to think about, Ed, when you think
about data center infrastructure, is there is no one size fits all in computing. You know, you need
great CPUs, you need great GPUs, you need to be able to bring those together in systems, you need
to have all these different types of computes, you need to have, frankly, AI PCs is another category
that we find to be a very, you know, interesting and exciting category. So when you put all those
things together, first of all, we're super excited about our new generation of Epic servers. This is
our fifth generation we called Turin. We've seen tremendous performance. In many cases, our
performance over the competition has increased substantially with this new generation. And what
CIOs are telling me is they're looking to modernize their data centers. You know, the idea of, you know,
there's a lot of data center infrastructure out there that is, you know, three, four, or five
years old. You know, those guys are not as performant. They take up a lot of space. And
frankly, they cost a lot from an operational standpoint. And so we're able to substantially
modernize the data center by adopting our new generation of Epic CPUs. We're seeing anywhere
from a seven to one ratio, you can replace seven legacy servers with a single new, you know, Turin
fifth generation Epic server. And that just results in tremendous, you know, paybacks for
enterprises. Lisa, this is a question from our Bloomberg Technology audience direct.
Would you consider using Intel for Foundry? Well, look, when we look at the manufacturing landscape
for semiconductors, it's really important to have, you know, a strong manufacturing set of partners.
We're extremely happy with the work that we do with TSMC. They're a great partner. It is also
super important for us to have geographic diversity. I think what the CHIPS Act has done has been
excellent in terms of trying to bring more manufacturing back to the US. We're excited
about the work that's being done in Arizona. And we're very active in bringing some of our chips
into the new factories of TSMC in Arizona. And we're always looking at, you know, the
manufacturing landscape. And we'll always think about how we can have the most resilient supply
chain. Lisa, for the last month, Bloomberg's Ian King and I have been using AI PCs with
processors based on ARM architecture. I feel like the gap with x86 is closing. When does AMD
wade into that arena? Well, I need to send you an AI PC, Ed, that's for sure. So look, AI PCs are
an exciting new category. It really will change the way the PC form factor behaves, especially
for enterprises. When you think about, you know, all of the business productivity, all the
collaboration we do, all the security that you need, AI PCs really fits that. You know, we just
announced that we'll have co-pilot plus enterprise PCs with our new Strix, you know, rising AI
capability. We had Microsoft on stage to talk about some of the new experiences there. You'll
see plenty of AI PCs from AMD this year, and that even increases as we go into 2025. So over,
you know, 100 plus enterprise platforms with best-in-class AI capability. Lisa,
let's end the conversation looking to the future. You celebrated 10 years. A year ago,
you set out aggressive targets for your company and the industry. You sat next to me and outlined
them. Where do you go next? Well, I would say, first of all, you know, this is an incredibly
exciting time to be in the industry. When I think about what we have in front of us,
the promise of AI, you know, the fact that high performance computing is, you know, at the center
of everything in the world, and we're able to really use the technology for good. You know,
that's super exciting for me. And then we have this humongous opportunity when you think about
AI overall, and AMD is the only one in the industry that can really bring together all
these end-to-end AI pieces. So it's a super exciting time, and I'm really looking forward.
I'd like to say we're just getting started.
