Lisa Su, Chair and CEO, AMD | Behind the Tech with Microsoft's CTO Kevin Scott

Lisa Su
2 followers
No offense, software is very interesting, but at the time hardware was much more sexy to me
and I had the opportunity to see how you could, you know, build chips and build very, very, you
know, they weren't the most advanced chips in the world, but to me it was like amazing. It was
amazing that you could build, you know, some transistors on something the size of a coin,
you could look at it in the microscope, you could see, you could measure it on a test system and,
you know, that's how I got into hardware and that's how I got into semiconductors actually.
Hi everyone. Welcome to Behind the Tech. I'm your host, Kevin Scott,
Chief Technology Officer for Microsoft. In this podcast, we're going to get behind the tech.
We'll talk with some of the people who have made our modern tech world possible
and understand what motivated them to create what they did. So join me to maybe learn a little bit
about the history of computing and get a few behind the scenes insights into what's happening today.
Stick around.
Hello and welcome to Behind the Tech. I'm co-host Christina Warren, Senior Developer Advocate at
GitHub.
And I'm Kevin Scott.
And today we are bringing you a conversation with someone who is such an important part
of the actual materials that make up our technology and just an incredibly cool person.
Yeah. Lisa is just sort of inspirational, I think, in general, you know, given her career
arc and like the position that she's in now of, you know, leadership in the technology industry.
But I always like talking to semiconductor people. Like my bias as a computer scientist
is low-level system stuff. So I am just a super big fan of Lisa and her company.
And everything that they're doing. Yeah. Like I even have a AMD like custom built machines
sitting under my desk right now that I've been very proud of owning for the past three years.
Yeah, it's funny when during the pandemic, when everyone was building gaming PCs, I was
amongst them and I built an AMD system and it was my first AMD system that I'd built
in a really long time. But I also helped build a Threadripper system for a friend who claimed
that they needed more compute than I think they actually did. But it was it was really fun for me
anyway. And so I'm right there with you. I don't have the same,
you know, passion for low-level languages and system calls like you do, but I'm a huge fan
of Lisa Su and an AMD and I'm really looking forward to this interview.
Dr. Lisa Su is chair and chief executive officer at AMD, where she's led the company's
transformation into a high-performance and adaptive computing leader. She's passionate
about working closely with partners to deliver the next generation of computing and AI solutions
that can solve the world's most important challenges. In 2018, she was elected to the
National Academy of Engineering. In 2021, she was recognized by the IEEE with its highest
semiconductor honor, the Robert N. Noyce Medal, and was appointed by President Biden to the
President's Council of Advisors on Science and Technology. She also serves on the board of
directors for the Semiconductor Industry Association. Lisa, welcome to Behind the Tech.
Thank you so much for joining me today. Oh, it's great to be here with you, Kevin.
Thanks for having me. Yeah, so we always start with folks from the beginning. So I'm just curious
how you got interested in the first place in science and technology. Like, did it start when
you were a kid with your parents? How did it begin? Yeah, absolutely. So Kevin, I was born
in Taiwan. I actually grew up in New York and my dad was a mathematician. He was actually a
statistician. And so we would sit at the dining room table and he would make me do
multiplication tables with him so that, you know, I had to be decent in math.
But I, you know, look, I've always really enjoyed, you know, sort of just seeing how things work
and building things. So, you know, one of my earliest engineering memories was, you know,
my brother and I were playing with his remote controlled car and it was kind of going down
the hallway and then it just suddenly stopped. And it was like, why did it stop? And so, you
know, I opened it up and saw there was a loose wire. And if I put the wire back in, it started
working again. And that was just super interesting that, you know, that's how things work. So
how old were you? I was, I don't know, I might've been like 10 or something like that. And my
brother's like, was younger, my younger brother. And yeah, we just kind of got curious about how
things work. Yeah. I think those are, it's really interesting. Those serendipitous moments when
you're young, where you, you go from like looking at something is like this inscrutable, almost
magical thing that you don't understand to being able to like have some kind of agency over it.
Like you peer into how the thing works and like you understand something like, I think they are
just super interesting moments. Yes. No, absolutely. Cause you're just so proud of
thinking, wow, like I learned something there and, and you know, that's, that's kind of carries with
you. Yeah. And so you were, you were a double E like how, you know, as you're going through high
school, did you know that you wanted to be an electrical engineer before you went to college?
Or did you decide once you got into school? You know, I, I don't think I could say I knew
Kevin. I again, in high school, I was, I have to admit, I was kind of a nerd. So, you know,
I did things like, you know, the math team and you know, projects like that. But I happened to get
into MIT and that was my first choice. And so I went to undergrad at MIT and at MIT, like everyone's
an engineer and you know, much of the, the the starting class is either electrical engineering
or computer science. And and so I think that that helped me kind of go in that direction. You know,
sort of the big, big decision wasn't, are you, are you going to be an engineer? It was, are you
going to be a hardware person or a software person? It was, are you going to be an electrical
engineer? Are you going to be a computer scientist? And that perhaps was what at least all the, all
the kids around me were thinking about, but there was no question I was going to be an engineer. So
that, that was, that was clear. And how did you, how did you gravitate towards being a hardware
person? Because like, I, I sort of had the same thing where I did you know, when I was young,
like a little bit of electronics and electrical engineering stuff and a little bit of software
stuff. And I like clearly gravitated towards software. I used to tell myself it's because
like software is inherently, inherently better. And like, it's faster. And, and yeah, I mean,
like that's the, that's the lie you tell yourself when you're, you know, a teenager, like in your
twenties. But like other people like have similar beliefs in the, in the other direction. Like,
I'm just sort of curious, like, how did you know that like, you wanted to be a hardware person?
Well, you know, I had, I had two experiences. I had, you know, experience in software where I was
like, you know, helping you know, sort of a lab program. One of the things that's nice about MIT
is that they really encourage sort of undergraduate research. So, you know, in addition to the, to the
coursework, they encourage you to get, you know, sort of side jobs. And I don't know, I must've got
paid $5 an hour or something like that, but it allowed you to work in labs with people. And so
I had two experiences working in labs. I had one that was more of a software thing, which I was,
you know, helping program, you know, whatever, some, some professors' experiments. And then I
had one that was a hardware thing, which was in semiconductors. And I got to do basic grunt work
for the semiconductor graduate students, but I was building something. So, you know, I was putting
wafers into a big reactive ionetcher. I was running the experiment. I could look at it in
the microscope. And, and that's how I fell in love with hardware. I mean, no offense, software
is very interesting, but at the time hardware was much more sexy to me. And I had the opportunity
to see how you could, you know, build chips and build very, they weren't the most advanced chips
in the world, but to me, it was like amazing. It was amazing that you could build, you know,
some transistors on something, the size of a coin, you could look at it in the microscope.
You could see, you could measure it on a test system. And, you know, that's,
that's how I got into hardware. And that's how I got into semiconductors, actually.
Yeah, look, I think the thing that all of us realize at some point is if you are in computing,
it's actually both pieces that are very important. And like, we'll get back to that in a minute when
we start talking about the work that we've been doing together, like Microsoft and AMD for so
many years. But like, I want to first like poke on this, this practical experience that you have,
because one of the themes that we have talking to computer scientists and engineers is this notion
that the abstraction layers that we have built up over the decades, sometimes obscure some of
the low level things. Like, I wonder, like, I'm not, I'm not a double E. And so I don't even know
whether you can get that experience that you had anymore in a material science program or a double
E program at a university where you get that visceral feel for what it's like to actually make
a semiconductor. Yeah, I think it's so important, Kevin. I mean, I'm a big believer. I mean,
there are people who are, let's call it, you know, very book smart. And then there's, you know,
sort of like people who are just pragmatic and have had a chance to experience different things.
And I'm definitely the latter. You know, I think I learned through experiences and I think those
experiences are so important, right? I remember, you know, sort of the first classes I took as an
undergrad and it was things like, you know, building your own little computer. And yes,
you have to, you have to build the circuit, but you also have to program it. Or as I said, you know,
building your first, you know, semiconductor device and just seeing how you go through each
of those steps. I won't say that everybody loves those experiences, but I certainly do. And not
that, you know, school is supposed to be job training. It's not supposed to be job training,
but it's supposed to help us think through what do we like to do in life? And it's kind of stuck
with me with the notion of it's so important to see the results of what you're doing. And,
you know, I love the fact that I can build products that I can touch and feel and, you know,
walk into Best Buy and see those products or walk into your data center and see those products. So
that's what I enjoy. Yeah. Well, and there is just this fun thing, I think, about
hardware in general, because even when you're writing a program, you don't get the same sense
as you're building it. Even if you're sort of assembling it piece by piece, you don't get the
same sense that you get when you're like building your own PC, like where you buy a motherboard and
a case and a power supply and a CPU. And like you just sort of are assembling this thing physically
and you at the end get to like have this working physical artifact. Yeah. So, Kevin, am I encouraging
you to come over to the hardware side or? No, look, I've been doing both for a very long time.
Like I'm thinking right now, actually, so the machine that I'm on right now taping this has
like an AMD, like 32 core ThreatRipper CPU and I'm getting ready, which I, you know,
very custom configured and I'm getting ready to do another build. And like I'm probably going to,
you know, hand assemble this next one because I haven't built my own PC in a long time.
Oh, Kevin, that's very exciting. I must send you our latest ThreatRipper because we just
kind of got the next generation out there. It's pretty cool. But look, I completely agree with
you. I mean, the opportunity to kind of build tech and touch tech, I think is super cool.
And it's great for getting students into STEM that I know is something that's very important
to you and to me as well. Yeah. So you you're majoring in electrical engineering at MIT.
And then what's next? You go to grad school, you go straight into the workforce.
Like, how are you making that decision? Because, like, that's a really interesting time
in your career. Yeah, it's you know, I was what they call a lifer at MIT. So I did my
undergraduate master's and PhD at MIT. It wasn't a super easy decision. I think you're right,
Kevin, because, you know, at the time, all my friends were graduating, they were getting jobs,
you know, they were they were moving to cool places. But I felt like I wasn't done learning.
And there was still more to learn. And I really appreciate sort of my PhD advisor.
His name was Dimitri Antoniotis. He was one of the people who built sort of the very early
simulation capabilities for for for semiconductors. Like, I felt like there was more to learn. And so,
you know, I decided to get a PhD and my focus was semiconductor devices. So I was building,
you know, a quarter micron devices, which at the time was very, very advanced. You know,
today we're talking about two nanometer. But hey, back then it was it was it was, you know,
sort of state of the art, you know, type things. And we were thinking about, you know, how do we
how do we push the envelope on and scaling? Because, you know, even then people were talking
about whether Moore's law was ending. It obviously didn't end. So, yeah, so I worked on something
called silicon on insulator devices. And it was it was a lot of learning. But it was also like,
like fun to be able to think that you're doing something that is, you know, sort of state of
the art research as as as part of your your studies. I'd love to get your perspective on
what you think the value of your PhD was, because I think a lot of people think that most of the
value is that contribution that they're making to the state of the art. But like, I kind of think that
the value is in the sort of discipline of getting a very complicated thing done and synthesizing,
like, incredibly complicated information. And, yeah, just learning how to be a researcher or
like someone doing something incredibly sophisticated around a whole bunch of other
people who have the same like high ambition level for the things that they're doing. So I'd love to
get your perspective on like what the value of that degree was for you.
Yeah, absolutely. So now you have to understand my perspective. Like I was so impatient, Kevin, as
as a student. So as soon as I started the PhD program, I was like, I need to finish as soon as
possible. So I was I was a little bit impatient. But I think the value of a PhD for me and as I
advise other people is it is not job training. It is it is not whatever project you do.
Some are great projects. Some are more esoteric projects. But it's an opportunity to really for
me, it was how to learn to think about solving very difficult problems. And if you think about
it's a problem that nobody else has solved out there. So you can't go to a book and say,
hey, this is how to solve it. You actually have to think through, how do I solve this? And how
do I do how do I contribute something to the industry or to academia where you don't?
The answer isn't clear. And the process of doing that over, you know, for me, it was four years,
three, four or five years really gives you the confidence that you can contribute sort of at
the highest level in a certain field. And that's that's what I take away from it. I just think it's
it's it teaches you how to think. And, you know, what we, you know, what we do, what you do, what
I do, you know, all day long is, you know, how do we solve very, you know, interesting, tough
problems. And, you know, sort of that time, you know, gave me sort of the confidence that, you
know, you could do that. And even at the time, it was also very much about a team. I loved working
with the other grad students and we worked together on how to solve some of these problems.
Yeah. Yeah. And I think the thing that you said about like they're very tough problems
and they're also ones that no one else has solved before. So like you can't go consult someone to
say like, hey, how did you do exactly this or skip ahead to the answer? Like there's no skipping
ahead. Like you just got to sort it out. Yes. So after after MIT, what was your first job?
You know, my first job actually was at Texas Instruments in in Dallas, and I wasn't there
super long. I was there about a little less than a year. I was kind of homesick. You know, I live
by the way, I live in Texas now, but at the time it was so far from home for me. So I ended up
spending the majority of the early part of my career at IBM in New York. And I joined what was
T.J. Watson Research Center and then the IBM Microelectronics team. And so were you working on
the RISC processors that they were building there? I was. I was. I was always you know, I spent my
time in their process technology area as we were continuing to look at the next generation
technologies. But yeah, the first processor I've been working on processors like forever. I mean,
like, you know, it's been, I don't know, now 30 years. The first processor that I worked on at
IBM was actually a power PC processor that that went into, you know, PCs. It actually went into
PCs as well as into some of the larger server systems at IBM. Yeah. I remember like I went to
a science and technology governor's school and like which had us do a bunch of internships. And
I interned at a place that had that had a power PC, like one of the very, very first ones. And
I just remember like what an amazing, amazing thing that was. It was super, super fun. I mean,
it was again, the idea of, you know, sort of RISC processing. And, you know, for us, it was how do
we get the performance and power and all that in the right place? So, yeah, it's sort of an
interesting, interesting thing. I like I don't I don't know whether you want to talk about this or
not, but, you know, we had this big revolution, like I'm guessing early in both of our careers
where like there was all of this innovation on instruction set architectures. And, you know,
like there was power PC and PA RISC and MIPS and the the the deck part, the like 21164, 264 series.
And, you know, what ended up happening just in terms of like where the bulk of the compute in
the world went was to processors running this x86 instruction set, which you build them,
you know, Intel builds them. And now like we've got this sort of interesting world where you've
got ARM in a bunch of places, you know, that are RISC instruction set machines. And then you've
got these GPUs, which are like an increasingly large volume of compute. And so like it's sort
of interesting, like a lot of the things happening now feel to me like the things that were happening
30 years ago, 20 years ago. And like, I don't know if you reflected any on that.
Yeah, you know, it's interesting. I hadn't actually bridged the parallel between the two.
I think what has been the case, and I'm actually curious your thoughts on this,
is the instruction set, certainly back in those days when there were so many different ones and
sort of the consolidation, I don't think it's been so much about the instruction set. I mean,
people ask me all the time about ARM versus x86. And I'm like, look, it's not about ARM versus x86.
They're both, you know, great instruction sets. It's really about sort of the applications and
the ecosystem and what you're trying to run on top of it. And, you know, there's a lot of
reasons that you want to be at scale. So, you know, if I think about what happened,
you know, 20 years ago, it's just that you had too many instruction sets and many of them were
not at scale. And so it just wasn't, you know, it didn't scale. And now you look today and what
we're doing going forward, the workloads are changing, right? I mean, that's what's made the
GPU so, so important now as we think through AI, as the workloads change, you need, you know,
somewhat different compute. And, you know, that's where the choices get made.
Yeah. Yeah. Look, I think I'm like one of the weird people who cares about instruction sets
because I started my career writing like a lot of assembly language code in grad school. I was
a compiler and computer architecture person. And, you know, so having written a software decoder
for x86. So, like, I think you care if you're implementing something at the lowest level of
the stack, but like that's such a tiny little minority of the development activity. And like
everybody else, like what you want is like low power, like high performance and cheap.
And, you know, it's like those three things. And to your point, like, I think scale drives that.
And would you agree, Kevin, that as you think about where things are right now,
the relative benefit of programming at the lowest, lowest level is perhaps less just because there's
so much computing power then. And that's why people are moving up the stack for speed and
agility and all that stuff. What do you think about the relative tradeoffs of the lowest level
programming versus. I think it's going to be the same thing. I think there's going to be this very
small minority of folks. And I think it's going to get to be a smaller and smaller minority over
time who need to deal with the lowest level details of the compute stack because, you know,
they're trying to do something that's right on the edge of possible or like they have to bring
every last bit of performance out of the lowest level things. And I think eventually most of that
gets abstracted away for most developers. And that's what you want. It's what always happened.
That makes I completely agree. I mean, that's certainly everything that we see is
and the technology is is changing so fast that, you know, the basic foundation of the
compute is getting so fast that, you know, you can sort of make up for whatever, let's call it
loss of abstraction, you know, pretty easily.
Yeah. The thing, you know, I do have this concern, though, like I sort of I mentioned this a few
minutes ago. Even though that's true, you still need this population of engineers who are excited
about those low-level details, like you need to hire them, like, you know, we need to hire them
for like folks building low-level system software. And so, you know, I do worry when
you know, when I look at what some kids are learning in computer science curricula now,
like your abstraction level that you're operating at is so high, like whether or not you like really
deeply understand like the full stack that you're operating on top of and then have the opportunity
even get interested in being one of those systems people who's poking around at the low levels
because that doesn't go away at all. Yeah, actually, Kevin, I think you're totally right.
I think as we as we think about, you know, sort of, you know, how do you have the best
sort of engineers or, you know, sort of the brightest people, you want them to have that
breadth of experience. I mean, it's really important. I mean, you can still specialize
but the breadth of knowing, you know, really, you know, how computers work and, you know,
what do you need to do to enable that I think is super important. You know, one of the things
that happens in, you know, for a lot of the folks that I talk to is how do we get enough people
interested in hardware because, you know, software is a sexy place. And I try to say, look, you know,
everyone's different in their interests but there is so much that can still be done around
optimization of hardware and driving that. More now than in the past 20 years. Like,
I think this is the most exciting time in hardware that we've seen in a few decades.
I completely agree. Completely agree.
So I go into work every day and I'm just completely surprised at the things that I'm
working on. So, you know, when I was in college and grad school, I did, you know, internship at
National Center for Supercomputing Applications and wrote a whole bunch of stuff for the thinking
machines, CM5 supercomputer. You know, I had an internship at Silicon Graphics right after they
bought Cray Research and, like, was working on the Origin 2000, like, this big cash-coherent
NUMA machine that they built that was super innovative. And then I left grad school and,
like, precisely none of that mattered for 20 years. And now all of it, it, like, matters again.
You really do have to think about some of these, like, old high-performance computing principles
to write some of the software that we're building today. And computer architecture matters again.
Like, it's awesome.
Yes, it is. I mean, certainly for some of the stuff that you're doing and Microsoft's doing,
you are absolutely pushing the envelope on everything that we think about as it relates
to hardware and systems.
So let's go back to your career progression. So you were at IBM, and then at some point,
like, was the next step joining AMD? Like, you've been there for a while now.
Yeah. So I was at IBM for 12, 13 years, you know, did a lot of stuff around our
semiconductor R&D and, you know, sort of the next-generation processor technologies.
And then, you know, I'm a semiconductor person, right? That's who I am through and through. And
from an opportunity standpoint, you know, sort of getting to be able to influence at a
sort of larger scale. So I actually went to Freescale Semiconductor first.
Oh, interesting.
Yeah. Yeah. I was at Freescale for five years. I was actually, I had your title. I was CTO
at Freescale as a company was thinking how to reshape their portfolio. And then I ran their
networking and multimedia business for a couple of years. And, you know, by that time, I'd already
moved to Austin. So now I'm officially, you know, a Texas person. And then I had the opportunity to
come join AMD. And so I joined AMD about 12 years ago, 12 years ago, something like that.
And it's been a great ride. You know, we've gone through, you know, a set of things as we
reshaped the company. But like I said, I haven't been far from processors
anywhere near my career. Somehow processors find me or I find them.
So when in your career did you decide that you wanted to lead teams of people,
that leadership was a thing that you either enjoyed or believe was necessary to
get done what needed to be done?
Yeah, it was probably in my early years at IBM. You know, one of the things that I
had to kind of decide is and someone asked me, I think my manager asked me,
hey, do you want to be an IBM fellow or do you want to be an IBM vice president?
And at the time I was like, huh, that's an interesting question. I actually thought that
sort of what was most interesting to me, yes, I liked working on my own research and,
you know, I had some good ideas. But what was much more fun was seeing teams get together
and do things that, you know, frankly, we didn't think were possible. And, you know,
the early memories of, hey, you're on a project, you know, you know, you have to ship something,
you know, to a customer at some given time and like nothing works. That's what I enjoyed the
most. Like I enjoyed thinking about how do I pull this together? How do I pull teams together? So
the answer to that question was, I don't think I'm smart enough to be a IBM fellow. So I guess
I'm going to try to be an IBM vice president. And I had the opportunity to lead, you know,
sort of small teams that became medium teams that became larger teams. But that's actually
really what I enjoy the most, Kevin. The technology is super fun, but it is even more,
you know, sort of rewarding to see teams come together and do something that, you know, truly
is groundbreaking. And that's what I've always enjoyed in my career. And, you know, I want to
dig into like what the career progression was or like has been at AMD. But, you know, I think the
whole arc of your career is extraordinary. You know, you're the daughter of immigrants to the
United States and you go from, you know, like first generation, you know, American to chair
and CEO of one of the most important semiconductor companies in the world. And like your path
through that was just being technically excellent, like getting a STEM degree and, you know,
leveraging technical excellence to like get into this like fantastic leadership position. And like
that's an inspiration for a lot of people and like probably particularly for both immigrants and
young women who are thinking about their path. And so I like I wonder, how do you think about
you? Probably you're probably going to blush me. You're making me feel like uncomfortable,
Kevin. Yeah. But but so look, whether you like it or not, you are a role model. And so I wonder,
you know, how do you think about like that job? So in addition to being CEO,
like you are a role model for for folks who aspire to be like you?
Yeah. Well, thank you for that, Kevin. I think well, I think a few things. I mean,
first of all, I was as as as good as one is, one also needs to be sort of in the right place
at the right time. And I think I've been somewhat fortunate in the sense that I've found sort of
the right place at the right time. I mean, you know, AMD, when I joined, you know, many people
asked me, well, why would you join AMD at that point in time? And actually, I never thought
about why wouldn't I join AMD? I mean, look, in the United States, you know, how many companies
are building high performance processors, right? They're just not that many that are doing that.
And I thought this was a place that I could help. I was passionate about what we were doing. Like,
for me, like I never said, like, I have to be a CEO. That wasn't my my thing. What I what I said
in my mind is, it was really important for me to work on something that I thought was important.
Like I wanted to be, you know, I love semiconductors. I want to be in an industry and a
place where I can make an impact on the industry. And AMD has been a great platform for that because
I do think that, you know, high performance computing and, you know, this technology is so
foundational for what we have to do. But, you know, to your point about, you know, being a role
model or helping like a lot of people help me get to where I am. You know, I think the mentors
that helped me the most were the ones that told me when I screwed up, frankly, because, you know,
everyone can tell you how great you are, but it's much, much better when someone tells you when
you've made a mistake. And so I've appreciated that. And I think my my my job or I hope what I
can do is is also, you know, help others feel like, hey, it's it's it's absolutely possible
for you to, you know, sort of follow your your aspirations and dreams. You'll make some mistakes
along the way, but that's OK. And so, you know, I get an opportunity to meet a lot of
women who are earlier in their career and much of what I encourage them to is is actually be ambitious
and, you know, feel like you can actually tell somebody what you want to do, because there are
plenty of people who want to help. But, you know, sometimes people feel, you know, shy or like, oh,
I can't say that. I guess you can. You know, yes, you can. You can absolutely do incredible
things and people will be glad to help you along the way. Of course you have to be good. Of course
you have to work hard. All those things are true. But I think it's just good to encourage people
that, yeah, you can do some some amazing things. Yeah, I I could not agree more with that advice,
like telling people that you have permission to be ambitious and you should advocate for your
ambitions is so, so important. It's crazy to me how many people don't do that or like don't have
a clear sense in their head of what their ambition actually is. You have just in your time at AMD
like seen a pretty incredible like set of developments in the in the semiconductor
industry. So like, you know, we've continued to make just amazing improvements in process technology.
You know, like I know since I've I've been like a professional computer scientist, which is a
very long time now, people have been talking about the end of Moore's law. And yet we,
you know, we continue to I mean, like Dennard scaling is done right. But we we have been able
to figure out how to like economically get more and more and more compute over time. And we have,
you know, figured out how to put that compute to use for crazy things like the whole mobile
ecosystem, like what you can do today with PCs, like what you can do with high performance
computing for scientific workloads and like now all of this stuff. So, you know, I wonder,
you know, over your 12 years, like what you think are the most interesting trends?
Yeah, no, that you're absolutely right. You know, Moore's law. I mean, we've talking we've been
talking about Moore's law either slowing down or or ending for the longest time. It it it really
hasn't. It has slowed down, but it has also, you know, allowed us to kind of think differently
about how we put chips together. By the way, Bob Dennard was like, you know, one of my heroes as
as a as a young as a young engineer. And I think what we've learned through this process is,
you know, I often say and to my teams and to others that it's so important to
make the right bets on technology because it does take so long for it to really play out.
So, you know, the fact that Moore's law has slowed down has meant that there are different ways to put
together chips, like one of the probably the most important decisions that we've made at AMD.
And this was back in the, I would say, 2014 2015 time frame is we decided that if Moore's law
slowing down, then the better way to put together chips is to to really break them up in these
things called chiplets. And and that was like a really important decision. Like I remember when
we made that decision and I was like, like it was almost a bet the company decision, frankly,
because we were trying to get a very, very competitive roadmap out there. And
the thought process is this is the future. This is the future of how do you put chips together?
We have to figure out how to put them, make them smaller because they yield better. They're much
more cost efficient, but the interconnect between them is so important. And so how do you make that
happen? How do you ensure that from a programming standpoint, the it doesn't affect software too
much? And now we've seen that that idea now on steroids, right? We just launched our newest AI
chip. And by the way, Kevin, thank you so much for for being with us at that. It's like chiplets on
steroids, right? It's like 12 chips stacked on top of sideways up and down and all that stuff.
If you had asked me 20 years ago as a semiconductor student or as a semiconductor
sort of engineer, I'm like, this stuff is never going to work. You're never going to work like
it's too complicated. It requires too much precision for it to work at a level of like
150 billion transistors. But that's the beauty of our industry. It's like, you know what,
we found a way to make it work. Now, I think as we look forward, there are, that's what we have
to be looking for is that there are inflection points in technology which enable you to take the
next big step. And making those decisions sort of at the right time are the things that I think
about, because you do have these fundamental limitations that are there in terms of physics.
But boy, we have just incredibly smart people. You give them a problem and they figure out,
well, there's a way to get around that problem. It's just you have to kind of invest in it.
Yeah. And so for folks who aren't chip people, let me see if I can try to explain what you just
described, and you can correct me where I get it wrong. So for when you open up a PC, like,
and you see the thing that is the CPU, like, that is not the chip. That's the package that
for the longest kind of time, a single monolithic die sat in. So like you would print a bunch of
semiconductors onto a single piece of silicon. You would put it inside of that package and like
you would very carefully connect the perimeter of that piece of silicon out to the pins or the
balls or whatever the connectors were on that package. And like what you are describing now
is like a bunch of the innovation is in the packaging technology and in like these components.
So chiplets are like little pieces of semiconductors that you place inside of that big
package like they're very complicated ways to interconnect them. And then a separate set of
ways to connect all of that out to the outside of the big package so they can sit on a motherboard
and do its job. So how close did I get? That was perfect, Kevin. Perfect. And I think that
to the casual observer, I think they don't have to care about any of those things.
All they have to care about is you're going to get more performance at sort of a better cost
point. So we talk all the time about in computing, we want to ensure that Moore's law was that you
would really double the performance every couple of years. Just going sort of the normal way,
you can't do that. But using all of these techniques and technologies, you really can
extend that performance curve, which just gives people like you, Kevin, much more compute for
all the great things that you're doing and building. So that's our goal in life is to make
sure that the compute does scale, that you do get more and that enables applications to do
lots and lots more. God knows we need it. Actually, I was a little bit depressed
for a while with computing because it seemed like we had lost our imagination for what you could do
with dramatically more compute. And we had sort of gone inward and you're like, okay, well,
how can I power optimize compute to get a certain set of capabilities available on a small
battery and things like, which is very, very important. But like the thing that's always
really excited me is like what can you do like if you just like cost efficiently had so much more
compute? And I think this current generation of generative AI is like one of the answers to that
question. And it's just remarkable. And I don't think we're anywhere near the end of the scaling
laws for things. So it feels exciting to me more in the way that the internet did than mobile,
just because we are so quickly deploying so much compute and have so many people with these very
inventive ideas about what they can go do with like this vastly expanded compute landscape that's
available. I completely agree. I mean, for the last sort of 10 years, I mean, we've had a lot
of computing evolution. It's been more about form factor, frankly. And we've done lots of great
things. But I think going forward, this is to me like the most exciting time in my career. Absolutely
hands down. I would have said a very similar thing. I think AI is the most sort of important
technology of the last whatever, 40 or 50 years. And it's because you can see there's so much
untapped potential. And as much progress as we've made in what computing can be used for,
it's still like hard to use. It's still like it's still not quite as accessible as it should be.
And generative AI just brings a whole new dimension for how we can use computing. So
I completely agree with you. It's just amazing what's in front of us right now.
Yeah. And the thing that always excites me is when you let people have less constrained
imagination. Like when I was a little kid and I was reading science fiction books and watching
the first Star Trek series and all of these optimistic science fiction movies and books
that were around then, like the computers were incredibly powerful. And maybe this is before
personal computers even existed, like these folks were imagining what a computer could be.
And maybe some of the reality of what the computing revolution actually became in a
way constrained people's imagination. And I think what's happened in the past
handful of years has unconstrained people's imagination again, sometimes in weird ways,
but I think in mostly incredibly optimistic ways. And that's exciting because I don't know about
you, like I'm of the firm belief that my imagination is not enough. We need everybody
imagining. I completely agree. Completely agree. Yeah. So let's talk a little bit about like where
you think AI compute is going. So you just announced this really like what I would call
a breakthrough in AMD's AI compute roadmap with the MI300, which is a very powerful GPU
tune for AI workloads. Like we've done a lot of work together to try to figure out how to
get the most powerful AI workloads working on this system. But like that's just one point in time.
So obviously you all see the same future trajectory that we see and like you must be
thinking all sorts of like awesome things about what that future looks like for the semiconductor
world. Yes, absolutely. So first of all, Kevin, again, thank you for your partnership, the
tremendous partnership between Microsoft and AMD. From my perspective, AI is such an empowering
technology and it's empowering in many dimensions. We talk a lot about the data center
from the standpoint of these large language models, the great work that Microsoft and OpenAI
and others are doing in just training the largest models in the world. Like you need lots of compute
for that. And that's where we come in. But I also look at it as it's actually quite a continuum as
to where AI is going to impact our lives. And so we look at it as in the data center, you need the,
let's call it the big iron so that you can train and infer the most complex models. You have the
opportunities at the edge when you think about all the data that's at the edge. And then you have the
opportunities at the client and kind of reshaping the sort of what PCs do, what mobile phones do.
And all of that requires AI capability, although it's, it may not necessarily be exactly the same
technology. I think they all want to interoperate together. So it has been a super busy year,
several years from the standpoint of, you know, really extending sort of our roadmap from let's
call it more general purpose processing to more of the AI capability. And as we go forward,
I think we're going to see AI in all of our computing products, like whether data center,
edge, client, and it's a fun place to be. But we're super happy with the work that we're doing,
certainly in Azure, but also on the PC and Windows side, like very excited about what's going on with
the Windows evolution and the co-pilot capabilities there as well.
Yeah, I think one of the interesting things, and again, it just sort of gets abstracted away
from folks because, you know, you don't need to think about all of this to make an API call to,
you know, the Azure OpenAI API service. But you all have understood this for quite a while because
you have some of the most powerful supercomputers on the top 500 list for scientific workloads.
When you're building these big systems, you have to think about everything. So,
how you get power into the data center, like how you cool things, how you design racks,
how you build your networks. And I think that's the other like really exciting piece here. It's
not just about the chips, but it's about everything that you have to put around the chips.
And you really have to have partnerships, you know, like the one that we have to
think about the full system design because otherwise you have one highly performant
thing. And if everything around it is not equally high performing, then, you know,
it just doesn't work. Yeah, I think that's what, you know, we all view as the opportunity, right?
The opportunity is, you know, through, you know, deep partnerships like what we have
to really, you know, sort of move computing to the next level of, yes, chips, systems, and then,
you know, also, you know, just what you're doing in terms of the model development itself and,
you know, bring those things synergistically together, we can build, you know, better overall
systems, you know, going forward. Cool. Well, we are just about out of time here. And so,
the last thing that I ask everyone on the podcast, even though your job is like one of the
most demanding in the world right now, and it's got nothing but more demanding over the past
year, for sure. But like, I do want to ask you what you do outside of work for fun.
Well, I have to say you're right. I think work is fun. I think you think work is fun, too.
Work is tremendously fun. But outside of work, I like to play golf. So, I have to say my golf
handicap, though, has definitely gone up in the last few years. I haven't played enough. And I'm
kind of like a foodie, Kevin. You know, we love to have, you know, great food and a little bit of
Bordeaux wine once in a while. It does well. And it's just an opportunity to kind of relax and
enjoy all the wonderful things in life. Awesome. Well, thank you so much for taking time out of
your very busy schedule to do this with us today. It's been great to hear more about your story.
Again, I'm so grateful, not just for the partnership, but for, you know, everything
that you've done in your career and the inspiration you are for, like, hopefully, a generation or
generations of young engineers. Like, we need more Lisa Sues in the world for sure.
Well, thank you so much, Kevin. We need many more Kevin Scotts in the world, too. It's really an
honor and pleasure to be here with you today and look forward to all that we'll do together.
Yeah, awesome. Thank you.
What a fantastic conversation with Lisa Sue. There were so many things that stuck out to me.
Her background, you know, it's impressive just to, like, kind of see her career arc and all the
different places that she's been. And I loved listening to the two of you kind of reminisce
about how the hardware industry and the, you know, silicon industry has evolved over the last
30 years or so. But one thing that really, I think, was great for me as a woman, and I think
also for just for people in general to hear, was her comment about how important it is for people
to ask for what they want their careers to be and to be open with their ambitions. What, like,
a remarkable and kind of simple comment, but also something that I think you made this comment,
a lot of people don't do. And I was just really struck by that.
Yeah, it is so important. Yeah. And it's two parts, I think. Like, you have to be clear with
yourself what it is you want and why you want it. And then you need to tell people and act on it.
You know, and I think this is consistent good advice that a lot of people give. Like,
Warren Buffett's advice is the best investment you can make is in yourself.
And it always amazes me, like, how self-conscious people feel sometimes about
expressing their ambition and asking for help and opportunity and, like, things that I think
would be really surprised at how open people would be to, like, getting on the journey with them.
No, I mean, I totally agree. And I think it's a few things. I think, one, it's that
there is kind of this idea in society that maybe you're not supposed to do that, right? That you're
supposed to be more humble, be less outward with your ambitions. I know that's true, especially
for women, which is one of the reasons why I really appreciated what Lisa said. And I appreciate
especially having people like her as examples to the rest of us.
But I think the other thing, to your point about asking for help, I think a lot of us are just
afraid to admit we might need help, even though when someone asks us for help, I think very few
of us ever would look at that as anything but wanting to offer that if we can, and certainly
not looking upon that request with judgment. Or if you make it a judgment at all, like,
thank God this person is asking for help because we're aligned in wanting to get something
ambitious done. And if they need help, please ask for the help as soon as you need it.
No, no, you're exactly right. And so I think even though these things make sense and we hear very
successful people say these things, I think that it can be hard for us to sometimes take that action
on. But again, I think it's so great that we have people like Lisa, who are examples of this.
Her career is so interesting. And what's been happening in Silicon and computing GPUs
over the last 5, 6, 7, 8 years has been just remarkable. And AMD has been leading the way
with that. And for you as a person who is both into hardware and software, was there a moment
when you kind of started to notice, OK, we're now back to where things are interesting again?
Yeah, I really do think it's been the past two or three years.
And look, I don't mean to throw any kind of shade whatsoever on the mobile revolution,
because I was one of the leaders in one of the companies that helped with the mobile
revolution. I was very excited about it. But there is a thing that we are returning to now
that I haven't seen since the beginning of the Internet, where you just have an abundance of
compute all up and like lots of interesting ideas with this new AI platform about what to go do
with it. And it feels really liberating in a sense, like just the things I can imagine doing
with it and then just seeing everybody around me like having these great ideas and then having
a platform that they can go use to try these ideas out. It feels very Internet-ish to me.
Yeah, I think I agree with you. At first, I was a little bit
hesitant because I think I have a stronger affinity towards what the mobile revolution meant.
But if I'm being honest, a lot of the ideas of the mobile revolution, we were not ready.
We didn't have the compute to really take advantage of what people wanted to do.
And I think maybe that's the difference. Whereas when the Internet became a thing,
we were at the point where hardware technology and the software were aligned. And so it felt
limitless. And with mobile, it took a few years for mobile chipsets to become powerful enough,
for wireless Internet to become powerful enough. But now I think, to your point, I do feel like
maybe we are at that correct alignment where we have these massive amounts of compute,
thanks to companies like AMD, that people can access from clouds or they can build their own
threadrippers if they want to be like you. But they have access to the compute. And so we can
finally make the two goals are aligned. Yeah. And it's just it's exciting to see
entrepreneurs be entrepreneurs. Like it really, really is. Like I can't even imagine what I
would be doing if I were, you know, like a 20-something-year-old young engineer right now.
Like it would like it's exciting for me is, you know, like a like an older engineer later in my
career. But like I would just be so happy right now and so energized, you know, to have all of
this power to go do crazy things that were hard to imagine just a year ago. And then just
imagining like everything that is still ahead of us. It's so much fun right now. I completely
agree. It's so much fun. And I'm with you. Like if I were in my early 20s, I think I would
definitely I'd probably be at an AI startup. Right. And because there's so much exciting things
happen. And it and as you said, like it does just kind of feel like one of those moments where
the possibilities are limitless. And a lot of that is because of the hardware, like hardware
is sexy again. And that hadn't been the case for a long time. It's absolutely true. And yeah. And
like that to my compiler nerd, computer architecture nerd beginnings, like makes me very happy.
Definitely. That's all the time we've got for today. A big thanks to Lisa Sue for joining us.
If you have anything that you would like to share with us, please email us anytime at
behind the tech at Microsoft dot com. And you can follow behind the tech on your favorite podcast
platform, or you can check out our full video episodes on YouTube. Thanks for listening.
See you next time.
