AMD CEO Lisa Su on the “Dead Sexy” AI Chips Race - On with Kara Swisher
52:47 Share

Look at all these people who are here in the evening. I know, to talk about chips. Yes, we're very excited. It's on! It's on!

Hi, everyone from New York Magazine and the Vox Media Podcast Network. This is On with Kara Swisher and I'm Kara Swisher. I've spoken to many people here on the podcast about the AI race and a lot of the focus has been on generative models and the companies that make them. But the real power and a lot of the tension in the AI space is in the hardware, the semiconductor industry that is building all of that compute.

My guest today has been a leader in the chips industry for more than a decade. Lisa Su, CEO of Advanced Micro Devices, better known as AMD. When Su took over AMD in 2014, which a lot of people warned her not to do, the company was almost on the verge of bankruptcy. By 2022, AMD was thriving and has surpassed its biggest rival, Intel, in market cap. Last year, Su was named Time's CEO of the Year.

I really always like talking to Lisa. We have a great rapport, largely because she's a straight shooter and she actually tells me exactly what she thinks all the time, even when we don't agree on things. She's also, unfortunately, one of the few women CEOs in tech, and so she stands out for that reason alone. But she also stands out because she's incredibly qualified for the job, and it's a big job these days.

You would think that the increased demand for the best high-powered chips to fuel the AI race would mean that Sue and AMD were just coasting, but not so. There are a lot of headwinds, not the least of which are the back-and-forth terror threats, export controls, funding cuts, and the

and all the other chaos coming from the Trump administration in the past 100-plus days. And instead of just two major competitors, Intel and, of course, the now-popular NVIDIA, there are a slew of other newcomers in the space to contend with, including in China. I talked to her all about that and about her strategy for this new future at the Johns Hopkins University Bloomberg Center this week.

This was the third event we've done there in our joint AI Discovery Series. It was a great conversation with a lot of insights about the state of the industry today and where it's going, and it was actually very funny. You'll see what I'm talking about when we talk about size and GPUs. So have a listen. ♪

Support for Pivot comes from me, Kara Swisher, and Johns Hopkins University. Recently, I interviewed AMD CEO Lisa Su in the third of four live episodes of your other favorite podcast, On with Kara Swisher, that I'll be recording at the Johns Hopkins University Bloomberg Center in Washington, D.C. In each episode through 2025, I'll be hosting timely discussions on AI, policy, copyright, and intellectual property, and more as part of the Hopkins Bloomberg Center's Discovery Series.

Listen to my conversation with Lisa Su in the On with Kara Swisher feed and stay tuned for more live discussions to come from our partnership with Johns Hopkins University. Lisa, thank you so much for coming. The last time we sat down for an interview was 2021. It seems like forever ago. It seems like forever ago, yes. It wasn't that long ago, but a couple milestones have passed since then. In 2022, AMD passed your biggest rival, Intel, in market cap for the first time. Congrats on that. Thank you. Thank you.

A few months later, ChatGPT came out and basically started the AI race. And of course, most recently, President Trump launched a trade war with China. I don't know if you've heard about it, but it's happening.

Maybe a little. Maybe a little. You would have to pay a lot of attention to it. So rank these three in terms of importance of how you're thinking about your business right now, your place and your competitors, innovations, political challenges. Well, first of all, Cara, it's great to be here. Great to see everybody in the audience. I like talking about chips. You're going to see that from me. Yes. I started in the chip industry. I mean, I'm going to date myself, but I've been in the chip industry for 30 years, and

Back when I went to school, chips were not exciting. Why did you do it? You were at MIT. I was at MIT, and I liked the idea... It was cool at MIT, but okay. It wasn't even that cool. It was sort of cool at MIT, but not amazingly cool. All right. But it's the idea that I like hardware. I like to build things. And so the idea of...

chips, you know, these are the brains of computers. And at that time, they were small computers. Now they're very, very big computers. I really enjoyed but at the time, you know, a lot of people were like, hey, software, that's that's a much more exciting field. So it's kind of foundational to say that right now chips are really, you know, sort of the underpinnings of this entire tech community, you know, where AI is going. And you know, that's one of the things.

that I do feel very fortunate about, that I'm in an industry that makes that much of a difference. Now, back to how do I think about these things? I mean, you've seen these technology arcs, Cara. They go in waves, let's call it tenure waves. And you saw that when the internet came out, and then you saw that when PCs came out, and then when the cloud came out. And AI is like one of those waves. But I would say that it's probably the biggest wave,

that I've seen in my career because it has so much potential. So yeah, that chat GPT moment was a very special moment. Why was that? You know, AI has been around for a long, long time. But what was true about it is it was sort of so hard to use because not everyone had access to AI. And what that chat GPT moment unlocked was that, wow,

Now everybody can interact with AI. Like my dad can interact with AI because you're using, you know, spoken language, you're using text, you're using a way of expressing how do you get a computer to do things without having to have deep knowledge of, you know, programming languages and things like that. I think that moment just unlocked the fact that if we can make computers easier to use... Mm-hmm.

then we can get so much more productivity, capability, actually, you know, sort of new opportunity. How is it being used from your perspective right now compared to what it's going to be? So I still think we're in the very early innings of usage. I would say, you know, this is a 10-year arc and maybe we're two years into a 10-year arc. But the way I think about AI is there's, let's call it regular usage. And regular usage is,

We think about companies trying to use AI to become more productive, right? Okay, how do I make something that is right now very manual and very labor-intensive much more operational? That's sort of one part of AI. But I think the other part of AI, and it's similar to the internet era, which is you first had companies just using the internet, and then you had companies that were truly changing their business model because of the internet. Mm-hmm.

I think you're going to see the same thing in AI. And we see that in our own work, too. I asked it to write a tooth fairy letter. Okay. All right. It did okay. In that realm. It did not do great, I would say. I would say it got worse and worse as it went on, as I kept arguing with it. And then it started misspelling everything egregiously. And I just said, go back to the first one.

It should have done a little better than that. No, it really didn't. I'll show you. I'm going to send you now. I'm going to text you. I'm going to see which model you're using. It should have done better than that. I was using ChatGPT. Is there a better tooth fairy model for someone? No.

I'm certainly not using XAI for tooth fairies. I'm just not going to. I'm worried about what I could get back. So I want to start with the challenges for you, for a company like yours, especially a trade war. You just come back from a trip to Japan and Taiwan. A majority of your chips are built in Taiwan, the world hub for the semiconductor production. That should be the 51st state, honestly, in some ways.

The administration has been all over the map on this one. In his executive order, President Trump said there would be no so-called reciprocal tariffs on semiconductors, and he said there would be soon. Same day, Commerce Secretary Lutnick said a semiconductor tariff would be coming in a month or two, but who knows. How are you dealing with that? Talk about the challenges here. Do you feel like you need to react to these things, or are you just sort of quietly sitting in the corner crying into your...

We are not sitting in the corner crying. All right, okay. We are recognizing that semiconductors are now...

very high priority list on, you know, sort of all national agendas. And that is new. As nationalist agenda that is key to the security of the United States. That is definitely new. I would say 10 years ago, that wasn't the case. Five years ago, maybe a little bit, but sort of in the last few years, that's become top of mind. And what does that mean to us? Okay. So first of all, the investment arcs are not changing. At the end of the day,

In a technology business, it's about making long-term bets. We decide what technology goes into our products like three to five years in advance. And at the end of the day, our job is to build the best. So we have to keep pushing the envelope on new innovation. That doesn't change.

What does change are a few things, right? And it's not even just what's happened in the last 90 days. It's been what's happened for the last few years, right? Supply chain resiliency becomes really top of mind. So when, for example, during the pandemic timeframe, that was a completely unexpected thing. I think we talked about it then. Yes, exactly. And when that came up,

It was like people were like, why can't I get chips for cars? And it's like, well, you know what? It's because it's a nine to 12-month supply chain. And if you want to change locations, it might take several years. And you need to plan for that in advance. And so we did a lot of work around the pandemic to give our supply chains more resiliency. And that's been positive.

Now, the fact that, you know, advanced manufacturing is mostly in Taiwan, that's absolutely true. And that's absolutely one of the things where we don't have enough options right now. So, you know, the focus on trying to bring more manufacturing back to the United States, I think we're all in favor of that. Like, that's a positive, right? And just put it in the resiliency category, which is... Nine to 10 years. Yeah.

It will take... If at all. It will take certainly five to 10 years. And when I say that...

We're going to manufacture in the United States advanced capabilities like this year, for sure. Our latest generation server chip will go through TSMC in Arizona, but it's still small. And there's so many steps. It's such a complex supply chain that you just have to plan every step of that. So when you have these tariffs coming in and out, Taiwan is facing a 32% tariff. Officials there said they had talks with the administration earlier this month and offered a zero tariff regime. Yeah.

Is that something that might be agreed to? Have you had conversations with the White House about this? Obviously, it affects your business. We've certainly talked about the implications of tariffs and what that would do overall. And I think our... When you say we've talked about the implications of tariffs, how does that conversation... Mine would begin with, what the fuck? Are you kidding me? But...

That's why I'm not Lisa Su. Anyway. We would say, look, we appreciate semiconductors are really important, and we appreciate that there's a long-term vision of bringing manufacturing back to the U.S., but it won't happen in a short amount of time. So let's just make sure that we're balanced in how we think about this. So...

If the administration levies that 32% tariff on Taiwanese semiconductors, and it keeps tariffs in China in place, which at last call were 145%, I think, China has ramped up its own tariffs, although it has said chips manufactured in Taiwan will be exempt, which can help you. On the other hand, Beijing has also suspended export of rare earth minerals, which are essential for semiconductors. So how do you deal with that as a CEO? I mean, these are

incoming from all sides? I think what we do is we stay very agile. We spend a lot of time analyzing what we can do in different places. And look, at the end of the day, we have to be part of the solution. And part of the solution is making sure that people understand this supply chain is complicated, that any changes will not happen in the matter of months. It'll happen in the matter of years. And we're here to

be as good a U.S. company as we can be, knowing that these things will have some impacts. The way I look at it, though, Cara, is these are a little bit short-term impacts. And so I don't want to say we don't spend time on short-term impacts. Of course, we spend time on short-term impacts. But it's a shock to your system. Yes. But over the medium term, if what we do is we get to a balanced trade system, which is

What we want is as a U.S. semiconductor company that has a lot of products, we want to sell it as broadly as possible and ensure that we still stay competitive in that regime. And so I think those are kind of the key high level bullet points that are really important.

So you aren't just dealing with tariffs, by the way. The Commerce Department also issued an export licensing requirement for sending your MI308 chips to China. AMD says that export controls could result in charges up to $800 million. To be clear, China was AMD's second largest market. As you just noted, you want to have markets everywhere. It made up nearly a quarter of your...

Do you think there's room for negotiation with the administration, and where are you with that? Well, I would say overall when we look at export controls, it's really complicated. And it's one of those things where, look, we get it. We get it that we have some crown jewel technologies and some of our highest performing chips are—

are, you know, frankly in the largest supercomputers. So I'm actually really proud of the fact that if you look at the top 10 supercomputers in the world, you know, five out of 10 of them have AMD technology in it. The number one has the number one supercomputer at Lawrence Livermore National Labs is our technology, the number three at Oak Ridge. So we have really powerful things and they're going to be export controls. That's a fact of life. What we want to do is ensure that, you know, when you look at the,

the framework of chips that we build. There are, let's call it consumer chips. There are, let's call it competitive but not bleeding edge chips that we would like access to. And then for the highest performing chips, we understand we're going to be export controlled. And our job is to make sure that we are

really spending time with those who are making these decisions that they understand. Is it appropriate, this more focus on national security over other things? Because they weren't paying attention at all, but most of those chips went into, like, games and cars and things like that, which is like, who cares, essentially. Yeah, and I think those are fine. Like, those haven't changed. Right, but is it appropriate to focus so much on the national security aspects of it? When you say, is it appropriate? Look, I think we...

We understand that it is just part of sort of national policies and national strategy, and it has to be a part of the conversation. We want to make sure that the regs are done in a way that give us as much access to the market as is not a threat to national security. And how do you do that? I'm really curious what you actually say in these meetings. LAUGHTER

I think a lot of it is intelligence about the market. So there are two ways to approach, I think, leadership in AI. And I'm a big, big proponent of the fact that the U.S. leads in AI today. The U.S. leads in AI chips today. The U.S. leads in AI models today. We want the U.S. to continue to lead. And to continue to lead, it means run faster.

And every day, like that's what we think about. When you actually ask me what I think about every day, I'm not necessarily thinking about tariffs every day, although I have people who are. I'm thinking about how do I go faster? How do we get the technology out faster? How do we get the next big innovation out faster? How do we collaborate better with our customers faster? And that's

Probably where we want to spend the majority of our time. Where you'd like to spend the majority. Yes, and also that would be our recommendation to the administration, frankly. We should spend the majority of our time thinking about how do we as a country go faster? And yes, there's a piece of this that says we need to also protect our crown jewels, and we want to be part of that conversation. And the part of the conversation is...

These rules are actually very detailed. They're not like you can ship this and you can't ship that. They're a combination of what you think the crown jewels are. And I don't think it's a question of should there be export controls. I think there's a question of where do you draw the line on export controls. And that's where I'd like to believe that I can take off my AMD hat and put on my U.S. semiconductor CEO hat

and be helpful in that process of defining where to draw the line. Where to draw the line. Now, the Biden administration also put export limits on semiconductors to China. Other countries tried to prevent China from getting ahead in the AI race, which was a goal. Then DeepSeek came out. Did you,

Do you anticipate the chance to be able to train an AI chap out for less than $6 million? That's numbers. What's the real number? I don't know that we need to focus on, you know, was it cheap? Cheap is what it is. With less. With less. With less compute. With less compute. Did it change your thinking internally? No.

I don't know that it changed my thinking so much internally. I'll tell you what I believe. What I've seen time and time again in tech is that necessity is a mother of invention. Like if you put constraints on a problem,

People will find a way to engineer around it. I mean, I'll give you sort of an example. There was a time when AMD was much, much smaller. And people used to say to me, Lisa, are you crazy? How could you possibly compete against Intel because they had 10x the people that you do?

And I was like, you know what, it's not about how many people you have. It's like, do you have good ideas? And in some sense, we were constrained because we had less people. We had to out-innovate and come up with new ideas and a new way of doing things. I think that's what we saw a bit with the DeepSeek moment. They didn't have the compute. They didn't have the gobs and gobs of compute that we have at USAILabs.

They didn't have the chips. They didn't have the compute. And so what they had to do was think of a different way to innovate. And I mean, the truth is, you know, when you're trying to do something, let's call it a little bit after the leaders, it takes less resources. And what they did was they aggregated some very good ideas that came from other top AI labs. What I think we learned from it is

You can slow down with restrictions, but you're not going to stop progress. I mean, you have to imagine that there are really smart people out there across the world. And if you give them a set of constraints, they're going to innovate around it. Did it change your thinking internally? It forwarded something that I really believed, which is...

two things. One is that, um, that there are lots of ways to solve problems and, you know, you learn different ways. Um, the second thing, which actually was really positive is that, um,

We saw the power of open source. Like, I'm a big, big believer in open source because it allows people to collaborate on a problem. And what you saw with DeepSeek is, you know, put aside DeepSeek being a Chinese lab. You know, DeepSeek was an open source model. And now you see, like, every research group learning from some of the techniques of DeepSeek to make their models better. And I actually thought that was really cool. Like, we saw...

In the realm of a short number of weeks, people kind of changing the way they thought about

you know, adding some innovation from what they learned. From what they learned from others. So despite all this, Huawei is reportedly about to start sending out its own advanced AI chip to Chinese customers in May. Baidu said last week that its cluster of its Kunlun chips can support deep-seek model training. Has it fueled a rise of new players? Are you worried about losing the Chinese market?

or other Asian markets completely, especially in the mode of terrorists, because I think the terrorists have advantage as China in many ways. So, look, we give Huawei a lot of credit. You know, they have a lot of smart engineers. I think we have better technology. Our goal in life is to continue to stay ahead, and that's why I said...

At the end of the day, we have to spend like 90% of our time thinking about how do we go faster? How do we innovate faster? How do we use all of the great capabilities that we have to continue that trend that we have and recognize that, yes, there's competition out there. Does this create a more competitive, especially with the tariffs on top of it and the language that goes back and forth? Yeah.

I don't know that I would say it's more competitive. I would say that, you know, in the grand scheme of things, I'm not a big believer in decoupling of ecosystems. I actually think, you know, we work in China. We learn a lot from that interaction. I think, you know, we have teams in China. And so, you know, my view of the world is there's a good opportunity for

to recognize that there should be some limitations, but also the ability to continue to have an integrated... Yeah, I'm a big fan of not decoupling either. Part of the Trump administration's stated goal is to increase manufacturing here, which I think a lot of people... He was not the first person to talk about this. Lots of administrations have talked about that. President Biden's 2022 CHIPS Act was over $52 billion, bet to bring semiconductor R&D and manufacturing back to the U.S.,

Trump is essentially rebranded as investment accelerator and executive order. He writes, it shall focus on delivering the benefit of the bargain for taxpayers by negotiating much better deals than those of the previous whatever. Okay. Um, can you give us some insight into those deals look like and what's changed in that regard? Because, um, uh,

You all profited indirectly from the Chips Act. TSMC's new production site in Arizona that you noted was awarded $6.6 billion in funding from the Biden administration. Some of your key processor chips will soon be made. Your historic rival, Intel, got over $7.8 billion in funding to advance their production products.

Mintel, just for people who don't know, after reporting loss in the first quarter, recently announced massive job cuts and cuts to capital expenditure. So talk a little bit about the impact of the CHIPS Act and whatever you want to call it. You can call it Phyllis if you want, but talk about... What did you say? Call it what? Yeah, Phyllis. I'd like to call it Phyllis. Yeah, whatever. I don't care what you call it. Investment accelerator on the CHIPS Act. I love Kara. Don't you guys love Kara? I mean, like...

Talk about the importance of this. Yeah, look, I'm a fan. And call it Phyllis the whole time. Okay, no, sorry. That'll distract me, Carrie. Is it okay? Can I call it the Chips Act? Chips Act is good. Okay.

I thought the CHIPS Act was pretty historic in the sense that for semiconductors to get to the level that it requires industrial policy for the United States, that's a big thing. That's industrial policy, exactly. It is absolutely industrial policy. Look, we didn't benefit directly from the CHIPS Act. We didn't get any particular funds, but I'm a big believer in U.S. manufacturing is a good thing. And...

The CHIPS Act was a good catalyst to start people really, you know, taking, you know, the bold steps to bring manufacturing back to the U.S. This is expensive, the chaos involved, training. It's expensive and, you know, frankly, I mean, we're all human beings, right? So if you don't have incentives to do something, you won't necessarily do it, even though you know that it's a good thing to do. Mm-hmm.

I think it's accelerated a bit, right? I was very pleased to see TSMC increase their investments by $100 billion. That's huge. I think there was a big thing that people learned from this thing.

Because it used to be that there was a belief that you couldn't build the same performance, power, yield here in the United States. That frankly, it was hard. We didn't have all the systems built and all of that. What has been really good over the last decade

sort of several years as we've started bringing some of the manufacturing back to the United States. Like TSMC has now said on record that their yields and capability in Arizona are as good as, you know, the chips that they build in Taiwan. And we see it, like we see it in our own products. Like it's fantastic. And I think that's like a clear existence proof, like, yeah,

actually, you can bring advanced manufacturing back to the U.S., and it's a good thing to do. Now, we need much more of it. Right, because they've got a 20-year, 30-year. Yeah, that's right. They would say the scale is not what it needs to be. So the idea that we need to go faster and we need to have more manufacturing, let's call it eggs in the U.S. basket, make perfect sense. It is more expensive. We just have to realize that it's more expensive. But the resiliency aspect of it and the sovereignty aspect of it make for...

good business. And that has been a change in people's mindsets. Does it change your way of thinking around your business when this is the case? It does. Because you want to make investments. I mean, one of the things, people make investments over the long term, especially in chips. Yes, I think it does. I think, you know, there was a time when

you know, sort of chip guys would normally think, hey, let me just go to the lowest possible cost because everybody wants to, you know, get the lowest possible cost. And I think that has absolutely changed. Now it's like, hey, we need to have a very resilient supply chain. And that's a global supply chain. But by the way, I don't think we can bring everything back to the U.S. That's also not practical.

But having a good portion of manufacturing here in the U.S. and around the world gives us resiliency in the business, which is a good thing. So let's talk more about the competition, too. There's been an AI arms race going on since 2022. It's shifted the characters. Now NVIDIA is the one to beat in GPU and data sector. In a Time article, you noted announcing you 2024 CEO of the year. You were quoted as telling your engineers that you have negative slack performance

You said, whatever we do organizationally, we cannot slow down. Talk about that. How has the AI changed your strategy and where do you put your energy then? So AI is the most important technology that I believe we'll see over the next 10 years. So with that, as a tech CEO, my job is to make capital allocation decisions. Like I decide where do we spend money. And over the last few years, we have really pivoted the company so that AI is in every product.

We talk about big AI a lot, like we talk about the AI that goes into, you know, meta and open AI and Microsoft. But the truth is, you're going to see AI everywhere. You're going to see AI in the cloud. You're going to see AI in the enterprise. You're going to see AI at the edge, sort of in telco edge things. And the reason I say it's negative slack is because the industry is moving so fast. Like, I've just not seen an industry move this fast. And that's where I go back to

It's not just the big players. I mean, there's some, you know, amazing, you know, startups that are coming out of, you know, the Stanford's, the MIT's, the Berkeley's, the John Hopkins of the world that are really coming up with really good ideas. Give me an example of one that you were like, that's pretty cool. You know, there's many, like Luma, for example. You know, when you look at text-to-video and, you know, what they've, you know, been able to do. Like the...

you look at where the technology was a few years ago and where the tech is today and then where it's going, going forward. These are, you know, some of the examples that are out there. Like I'm a big believer in, you know, healthcare and life sciences and that, that, that,

Though for the record, Lisa never brags. It's always the dudes. Sorry. They're always like, it's going to solve cancer. You've never done that to me. And I thank you for that. I'm not going to say it's going to solve cancer, but I'm going to say it's going to do a lot. Absolutely. That's why I like you. It's going to do a lot for drug discovery. It's amazing. I think they're often doing it because they're like, pay no attention to this other stuff we're doing, but we're going to solve cancer. Yeah.

But yes, no. So it is an example, though, of where we're just at the very early innings of what can be done. And so why do I say we have to go faster? It's because, like, we want to be the enabler of this technology. Negative slack. What is that exactly? I love the term. What is this? Oh, it's just the idea that if you build a schedule... Mm-hmm.

Like the idea that everything goes in order? No, you can't do that. You actually have to make it go faster than if you laid it out on a piece of paper. I see. So one area where you've been expanding is in data centers. When you took over AMD in 2014, the company's share of data center chip market was basically zero. At the end of 2024, your sales in that sector had increased. Maybe 1%. 1%. Okay. Okay. Well, basically zero. Okay. Yeah.

I'm good at math. No, you're good at math. I'm not good at math at all. I'm not even good at making tooth fairy AI stuff. Your sales in that sector increased 69% year over year to a record $3.9 billion. In March, you bought ZT Systems for just under $5 billion. They designed data center equipment.

And talk about this move to AI factory market, which is what NVIDIA has also been expanding. And in these data centers, a lot of people feel there's been a bubble in this. Some people think there's not been enough of one. And you've said you want to offer customers an open data center system rather than a proprietary system with all-inclusive pieces. Explain what you mean by that and how that makes you more competitive. So talk about the data centers and then the open idea. So...

I believe companies need to know what they're good at. And what we're good at is building great big computing systems. And so that was the reason that we were focused on data center business. Before data centers were sexy, it was the idea that... They're still not really sexy, but go ahead. Well, you're talking to me, so they're not terrible, right? Okay, sexy. Sexy it is. Okay.

So before people cared about data centers, the fact is we cared because I believe that if you can get compute at the most efficient capability, that you'll be able to really touch the lives of billions of people. And so what we see today is exactly that, right? I mean, the cloud is powering so many things and we continue to be the underpinnings of the cloud environment going forward. So that's kind of...

the view of, you know, why data centers and where we're going from there. And I think when you, you know, think about this move to AI factories, it's really hard to actually make all of these pieces work. And so our job is to make it easier for people. And so that is exactly what we're doing. That's why we acquired ZT Systems, a great system company that allows us to put together more of the components. And then, you know, to your conversation or question about

about open versus closed, you know, I guess I'm a believer in like choice is a good thing. Like people want choice. People don't only want one choice. People want the ability to choose kind of what their systems should look like. And it's almost a little bit like if you look at the phone industry, you know, Apple has a very vertically integrated system between hardware, software systems, ecosystem. And then you look at the Google Android ecosystem,

and you have lots of choices. That's sort of the difference between what I would say a closed ecosystem is and an open ecosystem. Apple's done pretty well with that one. They have. Not initially, but then they did. Yes, but there are also lots of other people in the ecosystem. And that's my belief, is that in AI, because there's no winner-take-all in the AI market. There shouldn't be. You're going to have big AI, you're going to have medium AI, you're going to have small AI companies,

Big models, medium models, small models, big compute, medium compute, small compute. It's like the three bears, but go ahead. And the ecosystem that puts that together, having choice is a good thing, and that's what we are working on. So NVIDIA also has 90% of the AI accelerator market last year, but you were trying to compete in this space. In March, you inked a multibillion-dollar deal with Oracle, a big company.

data consolidator, I guess, I don't know what you call it now, for a new, I'm going to get this right, MI355X AI accelerators. So talk about this deal and explain AI accelerator market and why you decided to get into it. Yeah, so the AI accelerator market, you might hear it called GPUs. I mean, think about this as truly the foundation. Like when you think about

training the OpenAI models or training the Grok models or what Meta's doing with Lama. You need lots and lots of GPUs. GPUs are very nice parallel things, so we call them accelerators, versus CPUs, which are more, let's call it traditional compute. And

I'm a believer and you need all of these things. There's no one size fits all in computing. You need all of the various pieces. Roadmaps are our lifeblood. When I talk about negative slack, it's like how do I get products out to market faster?

We're very excited about our roadmap. We are just about to launch a new series, MI355, as you stated. We'll be launching that in June, and it's the next generation, which will allow you to think about from year to year, we're going to see like a 35 times improvement in performance. In performance of these chips. Yeah, so that's pretty cool. These GPUs. That's correct. That's correct.

Can I ask you a side question? It's not in here. Why do men always talk about GPUs like they're talking about their penises? Why is that? That's a side question she asked me. Seriously, I've got 10,000. No, I've got 20,000. No, I put 100,000 together. Like, it's so strange. I don't even know how to answer that question. How should I answer that question, guys? I mean, I'm good with big, but what is...

You must be like, oh, stop it, boys. I mean, I know you do. You know what I'm like? I'm like, if you need GPUs, I can help you. Yes. That's what I said. Okay. All right. Sorry. I just had to ask.

Again, has never done this to me. Never called me about her big GPUs, but right now she just did. She kind of did in a very classy way. All right. A lot of big tech customers, Google, Microsoft, Apple, Meta, Amazon, have started designing their own chips. This is something I've heard Tim Cook talk about it. Does this concern you? Talk about that. I would think Apple would be the one

would be the most forward of those, but maybe perhaps not. But talk a little bit about that because they want to control their destiny. Sure. Look, I mean, Apple's done a great job, certainly with that vertical integration, what they've done with chips. My view of the world is, look, everyone realizes that like the semiconductor industry or the semiconductor capability is so important as you think about, you know, long-term innovation.

you know, sort of business resiliencies. It's not surprising to me that, you know, these large companies want to design their own chips. What I will say for sure is they're not designing the same things that we're designing. Right. And so there's more than enough chip companies

chip market for all of this. You know, our job is to partner well. I think that's one of the things that AMD does well. Like, we believe that our job is to make our customers better and to work more collaboratively with them. And so, no one's going to design all the chips they need. Like, it's impossible. Like, we can't design everything. There's...

Thousands of chips in these systems. There may be cases where the application has gotten mature. Like, for example, today we design custom game console chips. That application is very mature and you want to do something custom. But when you have something like AI where the algorithms are changing every day,

you may not want to do something custom because it may not run the next model as well as the general purpose chips will run them. Right, but so you don't see it as a threat to your... I see it as a fact of life and there's more than enough market to go after. So I want to switch gears a bit and talk about R&D. You're one of the few Fortune 500 CEOs with a PhD. You said you fell in love with semiconductors when you got, as you said, an engineering student at MIT and got a part-time job in a lab. Your alma mater and John Hopkins are

Two of the universities now suing the Trump administration are trying to cap research funds. How do you look at these governmental caps and other freezes to university research, which presumably is very critical in your mind, I would assume? Yeah, I would say I'm not fully...

with exactly what's happening with some of the governmental caps. But if I put that aside for a moment and just talk about university research, I'm a big believer in university research. I'm a big believer in research, especially that is early on and not tied to, let's call it commercial interests, is actually a great thing. So, you know,

The government has been a big funder of that. Frankly, the private sector probably should do more than we do today. And we need to work through this and ensure that there's a good pipeline of students. Students are incredibly important in this industry.

And then that, you know, we also, you know, continue, let's call it, you know, leading edge research that is, let's call it, not necessarily tied to, you know, to commercial interest. How under siege do you think that is right now? It could ruin a whole generation of scientists, whether it's healthcare or medicine.

tech? I'm certainly hoping it doesn't do that. How do you prevent that? Do you have to step in? Do companies have to step into this? Look, I think private sector does have to do more, you know, frankly. And I think public-private partnership in general is, you know, a good thing. So from that standpoint, we get a lot of benefit from the work that we do with universities. You know,

relatively small dollars. Like, for example, we give a lot of universities GPUs. Back to the big ones. Everybody wants GPUs. We give a lot of universities hardware because we want them to learn on our capabilities. And computing is expensive. And so there are places where we can do more and we should do more. But no one can replace government. For sure. For sure. I think, look, all of these things are cycles.

I don't want to overreact to any given cycle. I think the long-term interests are, you know, U.S. must continue to lead. And there's multiple aspects of what U.S. leadership include. It includes work that we have to do in the private sector. It includes work that we have to do in universities. It includes work that we have to do in national labs. All of these things are important. And of course, we want to do it as efficiently as possible, right? And

We all know that there are efficiencies to be gathered, but fundamentally, university research is very important. Very important. So I'm going to move on to the next thing, which is you grew up in New York, but many top tech CEOs with immigrant roots first came to the U.S. for college or grad school. I'm thinking of Jensen Huang of NVIDIA, Microsoft CEO Satya Nadella, Alphabet's Sundar Pichai, IBM's Arvind Krishna, and yes, Elon Musk.

Many have come to this country, to name a few. And so many. There's so many everywhere. The Trump administration has been targeting some foreign students, although that's moving back and forth, trying to take away green cards. Talk about the importance of luring top talent from around the world. And tech industry has depended on it almost completely. If I had to go through the various people who have invented things in this country, it's really pretty much...

It's heavily oriented towards people who have come from elsewhere. Yeah, absolutely. I think the U.S. was the place to come. It was the place to come to go to school and the place to come to do bleeding-edge technology, and I think we would like it to stay that way. So certainly skilled immigration is one of those things that we care a lot about, and the ability to kind of keep the best people in the United States is a good thing. What do you do to stop?

particular trend from happening? How do you lobby against that? Because I just heard from three major technologists who are people you do not want to lose in this country.

France is offering them visas and tons of money to move there. And they're scared, too, personally, because they happen to be vocal supporters of Kamala Harris, for example, and they're thinking of moving. That's at the top echelons. How do you get that back in the bottle? How do you get that? Because other countries are now moving in to take advantage, France, many other countries. Yeah, no, look, I think we all agree that this is an important topic and it needs to be addressed.

You know, it's always a pretty difficult issue to address. I mean, we've been talking about high-skilled immigration, you know,

Forever. And it's been one of those things. But I think we need to continue to go through the process of recruiting and maintaining the best students and the best researchers in the U.S. Yeah, I have never heard these people say this. This is something that was new. It was fresh for me. And I don't like to use the term unprecedented, but it felt like it when I heard it. So one of the things I've noticed—you're incredibly diplomatic. I appreciate it—

Because it's intense time. She's like, yeah, right. Yeah. But you're one of the top tech CEOs in the country. Do you have a role or responsibility to speak out? Someone told me, I'm saying a lot of things just recently, a pretty high-ranking, well, very high-ranking CEO, tech CEO. I'm saying a lot of things privately, Kara. Yeah.

Like, trust me. And I was like, what are you saying privately? What is going on? Do you feel like you have to speak out? Or most of them have not been. And in fact, some of them have been front and center at the inauguration and stuff like that. What do you think your role is? And how difficult is it right now? Especially when some irritating reporter is asking you this question. Well, I'm sitting here with such a very nice reporter. I did not say irritating. You know why? Because you have bigger CPUs than those men. But...

Look, I think as tech CEOs, we do have a responsibility to speak out. But I think it has to be authentic. And authentic means I don't weigh in on every topic that I don't have an opinion on, frankly. Like, I can't. Because...

And every day that we have, yes, we all read things, we all do things. But at the end of the day, I spend, you know, 95% of my time thinking about the semiconductor industry and what we can do within that view. And so I think the opportunity in places like, you know, this is how do we be, let's call it,

Not necessarily polarizing. So you can say I'm diplomatic. I wouldn't use the word diplomatic, but I'm not out to make a headline. But I am out to be pretty authentic. So if you ask me a question, I will do my best at answering it. And recognize that the world and sort of the country goes through waves and changes.

As you go through all those waves, our job is to run a good company, or better than run a good company, run a great company, and continue to be the best at what we do. Is tech spending too much time on politics? You've seen the impact on Musk's companies, for example. And it's not only because the cars aren't innovating. To me, it's hard. I think innovation—no, I mean it. I mean it. You can talk all you want. But as you said, innovation matters. And if they had a cool car, it would sell.

Instead, they have the Cybertruck. You know not have to comment, but it's ugly. I know you think that. By the way, Tesla is a customer, and they're a good customer. Look, do I think tech CEOs are spending too much time on politics? I think as a tech CEO, I spend more time...

not on politics, but on government-related things much, much more time today than I did three years ago or five years ago. And I have to, right? Because you just went through the litany of the reasons why. We need to understand tariffs. We need to understand export controls. I think the most important thing that we need to do is to make sure that people understand our business. And if you think about other industries, like you think about the automotive industry or the aerospace industry, they've been doing this forever. Right.

because they were much more entwined in sort of the national economy. We weren't. Like we were, you guys just keep doing what you're doing. Now we are absolutely in the middle of sort of the national economy, sort of the national security. And so, yes, we need to spend more time. And that doesn't mean like you're not going to see me on sort of on the cover. There's no chainsaw for Lisa Su, right? Sorry, I'm not going to make you insult Elon. Go ahead.

It's not about being seen, but it is about being heard. And I've learned this notion of not, you know, whereas the semiconductor dynamics of supply chains and manufacturing dynamics and design dynamics are second nature to us. For those who are kind of establishing policy, it's not second nature to them. And we need to help educate. And that's the way I view this. This is like, hey, we

We get it. We're important. We're foundational. Now let me help be part of the solution and not just be someone lobbying for my company. Right. It seems like often a giant waste. It's exhausting to a lot of people in the technology itself. Well, I don't think I'm good at it. Right, right. Anything that you're not great at, you find a little bit exhausting. They're not good at it either. But you're learning. And you're learning how to make an impact on something that's very important for the country.

Okay, so let's finish up with two more questions. When you took over as CEO of AMD in 2014, people don't realize that the company was close to bankruptcy. Its stock was around $3 a share. You famously decided not to go for mobile, which was the hot new thing at the time. And you're right. Lisa Su and AMD is a two-part Harvard Business School case study. Very exciting. As I mentioned, you were a time CEO of the year in 2024. This is my complimentary period.

when AMD had a record year. She didn't start with this, right? But your stock, like a lot of stocks, has had a tough time this year. Your quarterly reports come out next week, which you can't talk about, but you said you have to think long-term in the business. So what are you banking on in the next...

couple of years, what is your biggest worries as a CEO? You're obviously always looking at the stock market. The stock market is going all over the place, largely because of the tariffs, mostly the tariffs, actually. What do you have to think about? Because the values, I mean, NVIDIA got hit hard. People don't know what to make of these things. So how do you look at that? Yeah. So I look at it a couple of different ways. First of all, we cannot

sort of rank our own value based on what the market is doing on any given day. And I tell my employees that, right? I tell my employees, hey, like, don't spend too much time looking at the market because the market will go up, the market will go down. We're still the same company that we were. And, you know, frankly, it's all about long-term growth. And for us as a tech company, it's all about the product roadmap. So great products, great

deep customer relationships end up with long-term growth, and that's what we're trying to do. I also am quite motivated by this notion that I believe in tech for good, and tech for good is the notion that the products or the technology that we're building actually does make the world a better place. I know that sounds really like... No, I believe you, actually. I don't believe most. No, no, no, but...

I really mean that. No, I think you do believe. I really, like, why am I in tech? Like, why am I in semiconductors? It's for that reason, because I believe the stuff that we're doing. Because they're dead sexy, as you noted, but go ahead. Because the stuff that we're doing actually makes a difference. And if you actually make a difference, then that should show up in your market cap over a range of time. And I really do believe that. Like, I'm super excited about what we have in front of us. And yes, AI is a big piece of that.

But AI is, you know, a piece of a larger story of how do we use tech to make the lives of billions of people better. So on that, you have an event coming up in June called Advancing AI 2025 about the vision for AI. But tell me...

Tell me what you think. What is the thing that people aren't thinking about right now? And what's the thing they're thinking too much about right now? What would you say is overhyped and underhyped, I guess? I think people are spending too much time trying to predict exactly what's going to happen in the short term. The bubble. Yes. Is it bubble? Is it digestion? What's the latest model? Is some model going to come out and change the world? When is AGI coming? I think...

We're probably spending too much time on those types of questions because everybody wants the answer, and there is no answer. There is no answer. I think where it's underappreciated is how early we are in the cycle. It's really underappreciated that we are...

Two years in a 10 plus year cycle that the AI that we're experiencing today is nowhere near as good as it can be. And that I, you know, I truly believe that it will kind of change the norms of what we do.

you know, what we consider, you know, tech to do for us. And, you know, for enterprises, it's about business model innovation. And for, you know, us as, you know, sort of people experiencing AI, it's about, you know, making, you know,

you know, our lives, you know, just better. So my very first question, be a big visionary here. What's a thing that you would love to see it do? You know, that it would... I can't say cure cancer. You can, but I don't believe you. But I think as a more generic thing, I think AI can really accelerate the rate and pace of research beyond what we're...

Thinking about today. Today, you talk about AI agents and you think about agents as like customer service agents. Right. Here, let me get your ticket. Yes, these kinds of things. Like, I think about AI as, you know, PhD researchers. Mm-hmm.

That are helping us solve the next set of problems that we have. And I don't believe that they're taking over people's jobs. Actually, I really don't. I believe that they are making us, they're helping us accelerate the rate and pace that we learn at. And of your things that you're most nervous about? So that's techno-optimist, techno-nervous. Techno-nervous is that we get too scared of the technology. Yeah.

Meaning? Meaning, you know, I know there's this whole conversation about how much to regulate, how much not to regulate. Yeah, and a lot of very important AI researchers are part of that. It's not just crazy people or Hollywood. But I think you can work through those things. Like, I really believe you can work through those things. So one shouldn't be too afraid of it. One shouldn't slow down. Like, that's my belief. All right, everybody. Thank you. Thank you.

Thank you.

If not, you've got negative slack at this point. Go wherever you listen to podcasts, search for On with Kara Swisher and hit follow. Thanks for listening to On with Kara Swisher from New York Magazine, the Vox Media Podcast Network and us. We'll be back on Monday with more.