Building AMD with Mark Papermaster: How Bold Bets and Breakthroughs Reshaped the Semiconductor Industry

Summary

Podcast Summary: Building AMD with Mark Papermaster: How Bold Bets and Breakthroughs Reshaped the Semiconductor Industry

Podcast Information:

• Title: Building One with Tomer Cohen
• Host: Tomer Cohen, Chief Product Officer of LinkedIn
• Episode: Building AMD with Mark Papermaster: How Bold Bets and Breakthroughs Reshaped the Semiconductor Industry
• Release Date: May 6, 2025

---

1. Introduction

In this episode of Building One, Tomer Cohen engages in a profound conversation with Mark Papermaster, the Chief Technology Officer (CTO) of AMD. The discussion delves into the strategic decisions, cultural shifts, and innovative breakthroughs that facilitated AMD's remarkable turnaround in the highly competitive semiconductor industry.

2. Mark Papermaster’s Background and Career Path

Mark Papermaster's journey into the semiconductor industry is rooted in his early passion for chip design. He began his career at IBM, contributing to mainframe computers, which laid the foundation for his deep understanding of computing's critical role in industries like banking and national security.

Mark Papermaster [04:39]: "I took a class that really helped me understand all of the physics and the math and how chip design could really solve difficult problems. And so I switched and did my last internship with IBM doing chip design and the rest is history."

His tenure at Apple, under Steve Jobs, significantly influenced his approach to product development by emphasizing bold vision and prioritizing the end-user experience.

3. AMD's Turnaround Strategy: Bold Bets and Exit from Server Business

Upon joining AMD in 2011, Mark faced a company struggling with plummeting stock prices and a loss of competitive edge. Recognizing the need for drastic measures, Mark and CEO Lisa Su charted a transformative roadmap.

One of the decisive moves was AMD's strategic exit from the server business to focus on areas where they could be truly competitive.

Mark Papermaster [00:26]: "We made the decision early on, over a decade ago we'd actually have to exit the server business. This business is over $5 billion per annum for us right now, but we had to leave it before we could come back in."

This bold decision was pivotal in redirecting AMD's resources towards more promising sectors, ultimately fueling their resurgence.

4. The Role of Culture and Leadership in AMD’s Success

Mark emphasizes the critical role of culture and leadership in AMD's turnaround. At IBM, he honed his leadership and management skills, learning the importance of systematic processes and minimizing failures.

Transitioning to Apple, he absorbed the value of a strong, bold vision and the courage to break norms, which he later applied at AMD. Under Lisa Su’s leadership, AMD fostered a culture of transparency, accountability, and collaboration.

Mark Papermaster [08:14]: "Culture is absolutely fundamental... There's healthy contention, you know, challenging each other but then aligning on that. Why?"

5. Innovation Post-Moore’s Law: Chiplet Architecture

With Moore's Law slowing down, AMD had to innovate beyond traditional scaling of transistor counts. Mark introduced the concept of chiplet architecture, a modular approach that integrates multiple smaller chips (chiplets) into a single package.

Mark Papermaster [30:56]: "We were the first in the industry to this type of chiplet deployment... the way that we put these chiplets, these Lego blocks together really provided value."

This strategy allowed AMD to enhance performance and efficiency without relying solely on shrinking transistor sizes, positioning them ahead of competitors in addressing modern computational demands like AI.

6. Risk Management and Big Bets

Mark discusses the importance of making big bets while simultaneously implementing robust risk management strategies. At AMD, this involved building multiple exit points and fallback plans to mitigate potential failures associated with large-scale innovations.

Mark Papermaster [37:16]: "You have to manage risk. You have to know what those variables are... and you have to put exit points in the process."

An example cited is the development of the Zen processor family, where AMD committed to a new CPU architecture despite inherent risks, ultimately leading to significant market gains.

7. Future of Semiconductors and AI

Looking ahead, Mark envisions a future where AI is embedded ubiquitously across all devices, driving the need for more specialized and efficient chip designs. AMD’s holistic design approach aims to integrate advancements in hardware, software, and system-level optimization to stay ahead in the AI revolution.

Mark Papermaster [43:23]: "AI is going to be built in everywhere. We're going to basically just have smart devices in all walks of our life."

8. Managing Talent in the Age of AI

The advent of AI is transforming roles within AMD, automating repetitive tasks and enhancing productivity. However, Mark underscores the irreplaceable value of human creativity, strategic thinking, and empathy in leadership and product development.

Mark Papermaster [41:01]: "We're brainstorming, looking at all the different variables... it's a completely creative process that is not just based on data that I can train."

9. Misconceptions and Myths in the Semiconductor Industry

Mark addresses common misconceptions, particularly the belief that innovation in semiconductors is reaching its limits. He asserts that relentless innovation continues to push boundaries, dispelling myths that the industry is nearing its physical constraints.

Mark Papermaster [46:23]: "They were wrong every time. Innovation is an indomitable force."

10. Key Takeaways and Conclusion

Tomer Cohen concludes the episode by highlighting the key lessons from Mark Papermaster’s insights:

1. Holistic Product Development: Success hinges on understanding the "why" behind product features, not just the "what." Combining systematic processes with bold, user-centric visions leads to superior products.

2. Strategic Big Bets: In challenging times, incremental improvements are insufficient. Making substantial, calculated bets can redefine a company's trajectory, provided there are robust risk management and exit strategies in place.

3. Paradigm Shifts Drive Innovation: Embracing new architectures like chiplets exemplifies the necessity of breaking traditional molds to address evolving technological landscapes, such as the limitations of Moore’s Law.

AMD's journey, as narrated by Mark Papermaster, serves as a testament to how strategic vision, cultural alignment, and continuous innovation can overturn industry challenges and drive sustained success.

Tomer Cohen [47:53]: "Mark is drawing a compelling contrast between IBM's systematic process orientation and Apple's bold vision and user experience focus."

---

Notable Quotes:

• Mark Papermaster [04:39]: "I switched and did my last internship with IBM doing chip design and the rest is history."

• Mark Papermaster [08:14]: "Culture is absolutely fundamental... There's healthy contention, you know, challenging each other but then aligning on that. Why?"

• Mark Papermaster [30:56]: "We were the first in the industry to this type of chiplet deployment... the way that we put these chiplets, these Lego blocks together really provided value."

• Mark Papermaster [37:16]: "You have to manage risk. You have to know what those variables are... and you have to put exit points in the process."

• Mark Papermaster [46:23]: "They were wrong every time. Innovation is an indomitable force."

---

Conclusion

This episode of Building One provides invaluable insights into AMD's strategic renaissance under Mark Papermaster's technical leadership. By blending bold decision-making with meticulous process management and fostering a culture of innovation, AMD has not only revitalized its market position but also set new standards in the semiconductor industry. Listeners gain a comprehensive understanding of navigating complex industry challenges, the power of strategic risk-taking, and the relentless pursuit of technological excellence.

Transcript

[Leah Smart]

LinkedIn news.

[Jessi Hempel]

Innovation ultimately is solving problems.

[Tomer Cohen]

It came back to my own personal pain point. It just takes a while to build that trust.

[Mark Papermaster]

The problem that I was trying to solve, it was all I was thinking about.

[Tomer Cohen]

You have to be obsessed with the human condition. I'm Tomer Coyne, Chief product officer of LinkedIn and this is building one.

[Mark Papermaster]

We made the decision early on, over a decade ago we'd actually have to exit the server business. This business is over $5 billion per annum for us right now, but we had to leave it before we could come back in.

[Tomer Cohen]

That's Mark Papermaster, the Chief Technology Officer of amd. He's talking to me about the hard decisions the semiconductor company had to make in order to achieve its remarkable turnaround. We're going to get into that and so much more. So stick.

[Leah Smart]

From LinkedIn news. I'm Leah Smart, host of Everyday Better, an award winning podcast dedicated to personal development. Join me every week for captivating stories and research to find more fulfillment in your work and personal life. Listen to Everyday better on the LinkedIn podcast network, Apple Podcasts or wherever you get your podcasts from.

[Jessi Hempel]

LinkedIn News. I'm Jessi Hempel, host of the hello Monday Podcast. Start your week with the hello Monday podcast. We'll navigate career pivots. We'll learn where happiness fits in. Listen to hello Monday with me, Jesse Hempel on the LinkedIn podcast network or wherever you get your podcasts.

[Tomer Cohen]

One of the most amazing and important moments in the history of technology, and probably the world, is the invention of the chip. That's the origin of Silicon Valley. I started my career as a semiconductors engineer and I remember the sense of awe when I internalized what it took to design a cutting edge chip. It's like building a city on a 5 millimeter surface. It was that incredible. Fast forward to today and chips are everywhere. Your coffee maker, your washer, your dryer, even your car. In fact, if your car was made in the last few years, then it.

Has thousands of chips in it.

Thousands. Chips are also propelling the incredible revolution of AI, leading to not only technological breakthroughs, but also economic and societal shifts. That's why I'm excited to have Mark Papermaster with me today on the show. Mark is the CTO at AMD and he is responsible for the company's remarkable turnaround. When he arrived there in 2011, the company's stock was plummeting, the company lost its competitive edge and its bets were not paying off. Together with the CEO Lisa Sue, Mark set the company on a roadmap to success making strategic bets that would shape the future. And along the way, those bets changed the industry as well. Even before he joined amd, Mark already had a remarkable career at IBM. He was accomplished enough that Steve Jobs recruited him over to Apple to work on the iPhone. He eventually moved on to Cisco before ending up at amd. And all along the way, those decades of experience shaped Mark into the man that would revamp AMD's product roadmap. Mark is full of great insights. And in this episode we'll learn how Mark led that amazing turnaround. Why AMD had to exit the server business before it could come back and succeed in it. The dangers of incrementalism, why it's important to take bold leaps but have strategic exit points, and how modularity can be the key to achieving customization and scale. And there's so much more. So let's jump in.

I'm curious, you started in the 80s, what drove you into microprocessors then?

Did you see the potential back then.

Or was it something else?

[Mark Papermaster]

I'd love to say I saw where chip design would be and its role in the world today back then, but I didn't, to be honest. I just fell in love with chip design technology. I was lucky. I had co oping with its intern program with IBM. I had been working on actually the space shuttle program and I really enjoyed it. But in my studies at the university, I took a class that really helped me understand all of the physics and the math and how chip design could really solve difficult problems. And so I switched and did my last internship with IBM doing chip design and the rest is, is history as they say.

[Tomer Cohen]

So you started with a passion, you were passionate about chip design. How is your thinking of the potential of semiconductors and processors evolve over time?

[Mark Papermaster]

I started with IBM. IBM was a leader in the computer industry. And so when you think about it, I started working on these massive computers called mainframes. Not everyone today even knows what mainframes are, but these are the big computers that powered the world for years and they're still out there today. Mainframes are still relied on behind the scenes running your banking and some of your other applications that need that type of, you know, off the charts reliability and relying the code that may have been written decades ago. So I really understood what it meant from the very beginning when people banked their entire businesses or their national security on computing. And my timing was perfect. I happened to join right, right when IBM launched the PC. And so here was, you know, these mainframes these massive computers and then the idea that you can do a lot with a very small computer, a personal computer, I realized right away, like, my God, there is so much innovation that can be made in the coming decades to, you know, drive that computing forward and have it be much more accessible to people. Obviously that happened, you know, when you hold a phone in the palm of your hand, that used to be a huge computer and look forward to the future. We're going to have devices embedded in our own bodies. Tomer so I'm very fortunate to have started right at an inflection point of not by any means the first inflection point. I mean the transistor was invented in the 1950s, but it was a real inflection point when the PC came out and showed that computing could become pervasive. And it's been a great ride and there is a lot more on this journey.

[Tomer Cohen]

Still going back to your journey into amd, obviously, IBM, Apple, Cisco, distinctly, you know, very different companies, different philosophies for building products, I assume. I'm curious how those environments shape your product thinking. You know, IBM's long era of innovation to Apple's consumer focus, were there distinct kind of building craft principles you learned from those experiences as you kind of brought all the way to amd.

[Mark Papermaster]

All of us are a product of our experiences. You start with your innate curiosity, your innate knowledge and let's say training you did in your schooling. But you're really impacted by your journey, the people that you work with. I'm a sponge. I pick up something from everybody that I work with and hopefully I am part my experiences to them. And likewise, every company has its own culture and I never strategized to learn from the company I was at the different elements that could help me be a builder, help me be a leader of getting great products out. But it happened that way at each stop. I learned a lot. I mean at IBM particularly when I was there, it was a storied company to train leaders. It, it prioritized. What does it mean to be a leader? What does it mean to learn management skills, to not only take a good idea or an idea what could be a great product that you want to build, but you know, to really have a mechanistic way that allows you to ensure success or minimize failure. You're always going to have failure, but you want to minimize failure.

[Tomer Cohen]

And when you shifted to Apple, what was the stink there for you?

[Mark Papermaster]

Well, it was incredible for me. I opted and he worked directly for Steve Jobs. And what impressed me was what can happen when your vision is so strong about what you can do with the product that it can be so game changing that it creates a level of boldness and courage to break norms, break paradigms, move more quickly and I would say prioritize the end experience above everything else. This belief that if you fundamentally change the experience you have with the product and give people people beyond what they even dreamed, so if you can break norms, you can really achieve great successes.

[Tomer Cohen]

I think about my experience. I kind of went all the way from semiconductors to kind of software consumer experiences. This might be my own experience, but when I was working semiconductors we rarely talked about the end user. And I'm curious, was that your experience as well coming to Apple when you saw such a focus on the end user experience or this was already there for you in IBM before?

[Mark Papermaster]

Well, I'll tell you, it's different. So I had really been focused on business applications, enterprise building computers that you know, most people never saw because they were, you know, behind the back walls of a, of an IT department. So I was focused on customers. That was certainly a part of the training at IBM, but not as much on the simplicity. Give those customers the simplicity of the experience. You know, deploying an IBM computer in the time I was there was not a simple task. You really needed IBM experts to deploy that computer. What I realized at Apple was the beauty of not only developing a great product, not only breaking norms or breaking assumptions to deliver something uniquely different that can catapult the product or capability. In the case of amd, it's, you know, our chip design getting out chip designs that really disrupt and bring capabilities that may not have been there before, but also how you deploy it. How do you just make it easy for customers to absorb that technology and get a better experience? It has to be front and center. When you build a product, you'd love to think that there's just one thing. If I can do this, just one thing, this product can be a winner. It's absolutely never the case. You really have to think about why are you developing the product. Not just the what, the what could be the specifications. Oh my God, I'm 20% better on the frequency of this chip. My graphics resolution is 20% better. And my God, what a great specification. That's going to win the day. Nope, it's how all of it gets put together so that you know someone using that product. Let's just say it is a graphics chip. It's what's the experience? Is it more lifelike? Did I integrate with that new graphics capability with all the game developers. So if I buy a game, is it seamless? And it can use that, and, oh, my God, I get a better experience. And if you keep not just the what, the specifications, but the why in front of you and the whole team, it's really motivating. And if you keep checking in, are we delivering that original vision and make sure along the way that you're still delivering that it raises the odds of success significantly on what it is you're trying to build, why you and not your competitors out there?

[Tomer Cohen]

So this is a wonderful segue into amd. You've led a remarkable turnaround. It's actually hard to talk about AMD's current trajectory and not acknowledge that, you know, a little over a decade ago when you joined, it wasn't looking great. I'm curious, when you come into this experience, how did you identify what was broken? How much of it was like, hey, I'm gonna bring what I've learned about, like, you know, focusing on the why and the process design to make it amazing. Like, was it like, product strategy? You had to, like, delve into culture technology with all of the above. How do you go about such a turnaround when you join?

[Mark Papermaster]

Well, the answer is yes, yes, and yes. It is definitely all of the above. When you go back to that period of time over a decade ago, you know, when you think about amd, it had a storied history. So, you know, big innovations to chip design, but often it wasn't. You know, every single product generation, there'd be gaps, like a big innovation, and then several years later, another big innovation. I'll say, the raw talent and the culture of innovation, and certainly great, great people were here, but it was indeed a very, very challenging time. The stock was going down rapidly and. And the board recruited Lisa Su and I for me to run engineering, Lisa Su to run the businesses. And then just over two years later, Lisa was promoted as CEO. And my God, she's really one of the top CEOs in the world now. She's done just an incredible job. But again, go back in time. We faced quite a challenge. And so I think what was really important Tomer when you face a challenge like that, a turnaround situation is two things. One, you have to be very, very candid, and you have to get people comfortable with having very, very candid conversations, because it's really easy to be in denial. Well, we had bad luck and look what happened here. This was beyond our control. And, oh, we were dependent on this semiconductor Foundry process. And look, there was a problem there and that impacted us. You know, there's a dozen things like that, and frankly, they're all true. They're all factors that could create, and in the AMD's case, did create headwinds. But you've really got to step back from that and say, okay, what are our strengths? What are the things that we can tap and make a big difference? And then what is a strategy going forward that we can build on that strength? So that was the story of the turnaround. I mean, Lisa and I remember were working at those early days, seven days a week on what do we not do? What do we lean in? What do we do? How do we change the processes? We talked earlier in this discussion about processes. They have to be efficient, particularly when you're in a turnaround. I will tell you, you go back 10 years ago, we're a fraction the size of Intel, a fraction of the size of who we had working on graphics chips compared to Nvidia at the time. So we had to be more creative. We had to lean in on our strengths and our processes had to be efficient. Not bureaucratic, but efficient. If I developed an innovative approach and some circuitry for a PC product line, could I use that in my gaming chip? Could I use it in a server chip? So it caused us to really rethink every facet about what we did. And culture, you mentioned all the multiple facets. Culture is absolutely fundamental. There was cultural roadblocks. I know I drove a set of cultural tenets in engineering, which was really about being transparent. If you have a problem, call the problem. You can't fix a problem if you sweep it under the rug. You gotta be transparent. And you have to move into problem solving mode. You have to identify risks and you have to then think about then how do you execute extremely well to get that great product out. So there's a set of engineering kind of culture of execution. And when Lisa became CEO, she brought the whole company together, rallying around three mantras for the company. And the first one was build great products. And that why that I talked about earlier comes in there. Two, it's about really listening to your customers. Are you going to deliver an experience they value? Don't decide it yourself. Make sure you're customer centric and you really understand why. Are you doing something that's going to help your customers? Is it giving them a better experience? Is it letting them run their business or their personal lives in a very exciting way? And then lastly, simplifying everything that we do so a set of engineering cultures that really helped us execute and then a company wide set of mantras that Lisa brought to bear that allowed us to come together. And it's a huge piece tome of our turnaround. We don't have infighting, we align. We've got a culture that drives collaboration. Do we have contention? Sure. But it's healthy contention, you know, challenging each other but then aligning on that. Why?

[Tomer Cohen]

I'm curious, from a product strategy or product philosophy standpoint, was it about kind of trimming down? I'm just reminded of the kind of famous Steve Jobs story where you know, he came in and said like, we can't do all of this, pitch me your product. And then at the end of the day, he killed 80% of it because he had to make a few things be great. And you can be great at so many things. I'm curious, when you look back, what was a product decision? You made a decision on the strategy or the portfolio or a bet on the future?

[Mark Papermaster]

Sure. We made two big bets right off the bat, I'd say in the first months of the turnaround. One, to fund the turnaround, we had to win game console because we had at AMD at the time been the first in the world to really integrate a CPU processor into and a graphics processor on a single piece of silicon. That drove a tremendous efficiency for gaming and game consoles. And that's exactly what we did. We won Microsoft with Xbox, we won Sony with PlayStation. You know, people have to realize the PC had been well over 90%, I think it was, you know, 97 plus percent of the revenue. Because at that time AMD server revenue had fallen off, we had lost competitiveness. So the game console was a very needed revenue boost. And so leaning in that innovation, what makes us different? How do we get a better experience winning game console and hanging in that fight on PC? This funded the turnaround. We actually made the decision in the first months. I remember it was a Sunday night. Lisa and I and a number of the technical and business leaders in the company, we made the decision early on, over a decade ago. We'd actually have to exit the server business. This business is over $5 billion per annum for us right now. But we had to leave it before we could come back in. Why? You have to be realistic again, candid, blunt. Our product wasn't competitive. We had to rebuild it. We needed an entire new CPU processor, became the Zen processor family that we kicked off in fall of 2012 and takes a while to get a brand new microprocessor out. So it first went to market in 2017, but we created a whole family. So every 18 months we've had a new processor that takes of that Zen family. It keeps it competitive. That was a huge bet we made Tomer on high performance.

[Tomer Cohen]

But that's very hard. Right. When you're a builder, you always have ideas about how I can do better. Right. If you're a good builder, you're always like, I can win this, I know if I can build this, I know I can win the market. And it's a hard call to say, hey, I don't want to participate, I don't want to compete. I want to win and bet on something that actually will scale. Like gaming.

[Mark Papermaster]

Yes.

[Tomer Cohen]

So that's a tough one when like builders are coming along and telling you, mark, you don't understand, we can actually build this roadmap. Is this like coming from the outside perspective helpful? Because you're not tied into roadmaps, you're not tied into legacy. Like you're coming in fresh.

[Mark Papermaster]

It's a fresh perspective. I had been a user of AMD technology at IBM. I had run a server team that was using a AMD chip that was late, it missed its cycle. I had to actually lay off part of the team I had working on that because AMD was late and missed a cycle. You can't do that. You have to show up for your customers each and every cycle and you can't fall into the trap of incrementalism. I think at that time, again, well over a decade ago, there was some incrementalism going on, meaning, well, we can catch up, but let's do it by just patching this, patching that, doing small changes. And what was key was to really take a bold step and design a new microprocessor again. That was that Zen family. And it proved very, very effective. And we knew that we had to stage these bets. We bet first on the CPU to take intel on, and we didn't lose sight of graphics and taking Nvidia on. But we had to stage that a little further out in time. So you had to bring in the revenue. You have to be pragmatic. So don't do incrementalism. Think about something big, big that can really change the fortunes of the company. Think about how much you bite off. If you bite off everything all at once, you're going to fail. We chose to reconstruct the CPU business first. Follow it with the GPU business. We hit $5 billion on our GPU for the data center. Business last year and it had been almost $100 million. So just a fraction of that the year prior. That's not incrementalism, it's thinking really about getting a product that can really make a difference for customers.

[Tomer Cohen]

You mentioned before, listen to your customers. It reminded me of a story I heard about how Nvidia stumbled upon AI for gpu. So when I was a kid, Nvidia chip or AMD chips, card were basically how I can play the best game on my PC. And now obviously when you say Nvidia, you say GPUs, you say AMD. You think a lot about AI, that's what comes to mind. And I think the story about how some Stanford PhD grads were playing with Nvidia chips because they realized they can do parallel processing and being able to get their work done much better because the CPUs are sequential, I think for me highlights the story about listening to your customer actually looking for those serendipity moments were those sums that you can see with AMD as well.

[Mark Papermaster]

So you have to give Nvidia credit. I mean, they did listen. Well. I was in the Bay Area 2012 and you started seeing those advancements with natural language processing. All of a sudden, the accuracy to listen to the voice as you and I are talking here and to be able to identify and, you know, recognize that. And then of course the recognition of images, that was all the early work of AI, of machine learning was rather than a rule based approach that had been around for decades to really use deep neural nets to much, much more efficiently perform these kind of tasks. And of course a GPU is ideal, it's a parallel processor. So if you're going to parse the task out and be able to distribute it, that's what a GPU is architected to do. Because it was architected to do gaming and visualizations. You know, that's all the pixels on your screen, that's all parallel processing. That kind of listening is, is also a story, of course, of what we've done. It was part of AMD's history. The first to think about how you could integrate some of these elements together, like the CPU and GPU and later techniques that we applied to combat Moore's Law to be able to have tighter and tighter integration. End customers don't care how you integrate it, but if you give them a great experience at much less power and a better cost, that's a game changer. So by listening to the customers and understanding that we could provide a much more efficient way of computing and combine different types of computing, CPUs and GPUs together. That became our rallying cry at AMD of how we could provide a difference. We had to earn our way to get a seat at the table of the customers where they would tell us, here's our problem. You guys have something.

[Tomer Cohen]

We're going to take a quick break, but don't go anywhere. When we come back, Mark is going to explain why it's essential to lead with big bold bets.

[Mark Papermaster]

I will tell you, this is the riskiest thing we've ever done. We had had multiple exit points on a number of the technologies. No one had done this before.

[Jessi Hempel]

From LinkedIn News, I'm Jessi Hempel, host of the hello Monday podcast. In my 20s, I I knew what career success looked like in midlife. It's not that simple. Work is changing, we're changing, and there's no guidebook for how to make sense of it. So come figure it out with me on the hello Monday podcast. I've been a journalist for two decades writing cover stories for BusinessWeek, Fortune and Wired. And now every Monday I bring you conversations with people who are thinking deeply about work and where it fits into our lives. Like Microsoft CEO Satya Nadella on growth mindsets.

[Mark Papermaster]

The learn it all does better than the know it all.

[Jessi Hempel]

Or NYU Professor Scott Galloway on choosing a career.

[Tomer Cohen]

I think the worst advice you can give a kid is follow your passion.

[Jessi Hempel]

Or MacArthur genius winner Angela Duckworth on talent versus grit.

[Leah Smart]

Your long term effort and your long term commitment are support surprisingly important.

[Jessi Hempel]

Each episode delivers pragmatic advice. For right now, listen to hello Monday with me, Jesse Hempel on the LinkedIn podcast network or wherever you get your podcasts.

[Tomer Cohen]

We're back and I'm speaking with Mark Papermaster, the Chief Technology Officer at amd.

Mark, you mentioned Moore's Law. So this is one of the most famous adages in technology for folks for the first time hearing it. It's basically the observation of the number of transistors on a chip on integrated circuit will double every two years with minimal rise in cost. Which basically means that in plain language you will see exponential growth performance with corresponding energy efficiency, which basically allows for incredible computing velocities, smaller devices and cheaper prices. I actually remember when I was in semiconductors, I clearly remember the CEO pushing for double the performance and double the power efficiency. That was the slogan. I think it was like I'm on double of both.

[Mark Papermaster]

And that's kind of the essence of Moore's Law. You got it.

[Tomer Cohen]

Exactly. That was the what I wish there was more of the why, but that was the what, what was interesting. It's not that this happens naturally, right? Like it was incredible innovation from people that made it happen. So it wasn't like Moore's law was such a natural law that would, you know, you see tight and it just happens. People actually had to push for that. Now with Moore's law slowing down due to physics, what does this mean for chip designers and builders? Like, does this require different type of innovation? Are you like, hey, those days are gone, I don't want to push for that anymore. That for me is diminishing returns. We need a whole different way of thinking.

[Mark Papermaster]

That's right.

[Tomer Cohen]

So what is that like?

[Mark Papermaster]

Yeah, so we talk about listening to customers and how that drives features into the product or how your product makes a difference. But what anyone building a product has to think about is the fundamentals underneath that whatever you're using to build. In our Chase chip design so we're the experts of the chip technology. AMD is a 55 year old plus company and so we are the ones that have to solve that problem. So Tomers, you described when that Moore's law of physics slowed down, it took whole new innovation. And it's been well over a decade that we've been in this era of slow development. So it's not like each new semiconductor manufacturing plant that's built and each new semiconductor technology node isn't better, they are better, they do make smaller transistors. Each of those transistors burns less energy, but it's a lot more expensive. So it used to be Moore's law. So you did all those improvements and kept the price the same. Well now the price is going way up, so you have to innovate. And the way that we think about innovation, it was from the get go at the turnaround at amd we said, you know what, we see this coming as we think about our processes and how we design chips. Let's think about it in a more modular way. Let's design our chip functions such that when the technology gets to where you could break apart, not have to build all of it on one chip, but use what we call the term chiplets, break it into different pieces where you could build one chip in an older technology node because maybe it didn't need that most expensive yet most power efficient and densest transistors. And yet let's say for the CPU and GPU where you're getting that computing, well, yeah, you want that on the cutting edge. So we innovated and we were the first ones to bring out chiplets. We started it in our CPU line. And then that GPU that I told you just broke 5 billion in sales last year. From very little to the prior year. It actually has 13 of the chiplets. Actually, I've got one here. I don't know if it'll show up on here, but as you look at that, it's 13 little chiplets put together and then a bunch of memory around the edge. And it's all connected over a silicon plate. And it's a whole mix of technology nodes. It's the latest technology nodes. It's an older technology node. It's memory from our memory suppliers all put into one chip. This is about 150 billion transistors. The computation you can do in this one chip used to be a room of servers back when I started my career, some decades ago. So it's that kind of innovation, Tomer, that combats Moore's Law. It's thinking out of the box, it's thinking differently. And you say, well, how is that a different experience? Well, the difference experience is power, is gating everything. I mean, if you look at what does it take to take on AI and get these new GPT or Gemini or Anthropic, all of these new models, they're all gated by how do you get enough processing power in the amount of energy that you can provide? How do we have a sustainable solution? This innovation of thinking more holistically, not about just one chip, but how do you subdivide it? How do you partner with a semiconductor company? How do you optimize all the way through the software stack? That's the name of the game now and going forward, so that we continue to move forward at an exponential pace of getting more computing power and more efficiency at every generation, despite the slowing of the gains that Moore's Law used to provide.

[Tomer Cohen]

Mark, what's a good analogy for folks who are less familiar? So modular, obviously this is the Zen architecture, right? That amd.

[Mark Papermaster]

It's Zen for the CPU and it's what we call instinct for our GPU in the data center and Radeon for our gaming. GPU's great.

[Tomer Cohen]

So modular basically allows you to start to bring in multiple chips, whether it's CPUs, GPUs, AI, accelerators into kind of one environment. The chiplets really allow you to take smaller components of them.

[Mark Papermaster]

Yes.

[Tomer Cohen]

So before that was not an area of focus because Moore's Law was just paving the way to do it better. And now that Moore's law is not providing the same dividends as before. You know, you're realizing, okay, there's more programmatic what I have to bring in. How should one think about the transition if they're not familiar with the industry?

[Mark Papermaster]

Well, I think it goes back to what we talked about earlier, Tomer. You have to be able to take on some big risks. You can get stuck in incrementalism. So we were the first in the industry to this type of chiplet deployment and some people ridiculed us. They said, oh, you're putting a bunch of Chiclets together like, you know, like candy. But no, the way that we put these chiplets, these Lego blocks together really provided value. It was the what underneath, providing the why of tremendous computing at tremendous efficiency. And now no one doubts that everyone's moving to a chiplet type of approach. You're going to have to take risk going forward. Things are constantly changing. Whatever product you're designing, if there's one constant, those paradigms will change on you. And if you continue at the incrementalist approach, you're going to fail. And so what we like to do at AMD is we like an underdog mentality. I told you we were always 1 10th the size of our competitors. Well, we're not 1 10th anymore, but we still are much smaller than our competitors. And we like that. We don't have a choice. We have to be more creative, we have to take some more risks. We manage those risks and it allows us to delight our customers to be able to bring them more value going forward.

[Tomer Cohen]

And the bet there, which I want to go into, because I want people to understand like how hard it is to make a bet in semiconductor because you are playing for years ahead. Yes, it's not a bet on the next year. You're playing three to five years ahead when you make a bet on modular design, which basically means it's more customizable. You can have multipurpose software or products running on top of it. You're not doing it as like a specialized use case. But then you allow for chiplets, which allows for specialization at the dice level, at the kind of chip level you're really building for customization. That's the bet you're building for. Allowing customers to think how they want to use this in the future and having the ability to go back and forth. Is this used for training? Is this used for inference? Is this used for this type of application? That application? If I go back to the idea of big bets, long time horizons in semiconductors, is not well understood by people outside of the industry. Remember when I was working on it and again, I'm a novice rookie versus you. We talked about five, six, seven years and for me it was hard to fathom sometimes. But I'm thinking about the leadership that requires to make those predictions about the future. Yeah, and a prediction gone wrong is actually very costly because the development cycle, the manufacturing cost is. Those are like, you know, one way doors. In a way, those are hard to reverse decisions. So how much of your time do you spend in the future when it's so hard to hedge? What does that mean from a product development standpoint?

[Mark Papermaster]

I'll tell you what we do at amd, we recognize how important this is. Again, if you know we're going to keep this underdog mentality and with an underdog mentality, you may not be perfect, but your aim better be pretty darn good. You better be hitting the center of the target or pretty close to it every time. It's not a great sexy answer, but you really have to be incredibly thoughtful about roadmap. You have to do two things. First, for whatever your product is you're building, you have to break apart what are the elements for your roadmap? For us, we build semiconductor chips. Our roadmap going forward, you can't fall behind in any of the components. If it's a gaming chip, I need to have the best rasterization, my cpu, whether it be from anything from a PC or embedded device up to a server. Am I absolutely competitive on that CPU roadmap? The same thing for the programmability devices, the adaptive compute. So you have to break apart all the engines, all the support circuitry of how you feed those engines, all those pieces you're managing in a roadmap process, I mean, we literally month in and month out, you know, crawl through each of those elements and make sure that we're on the competitive path. Then you got to think about what are the pieces you need to put that together. Anyone building a product, you rely on others. We rely on semiconductor factories and factories that can package all this together. So we manage the heck out of it. We've got people living with those manufacturers, understanding what's coming next, making sure we're absolutely on the cutting edge. And then you have to go up to the software stack, all the way to the applications. You have to manage that roadmap. And we have a roadmap process where we bring all of those pieces together, look at how it culminates. And we are very tough on ourselves. We Assess our competitiveness across every product line going forward. And the rule has to be, give all your competitors the benefit of the doubt. Assume that they're going to achieve or exceed what you think they might be able to do. And are we still competitive? Are we innovative? Are we delivering a better experience, a better value proposition? More performance, less energy. So it's complex. And the way that you deal with complexity is you have to tease it apart. You got to manage all the pieces of it, and you have to put it back together again and ask yourself the really tough question, does this make sense? Is it going to win? And having, frankly, some healthy, contentious debate along the way, Mark, I'm still going.

[Tomer Cohen]

Back to that point around, like the big bets, do you give yourself exit points? Because, again, the bet is large. If you're two years in, I could see the fallacy of sunk cost playing into it, saying, oh, we invested so much, you know, even if it's not the best ship, we can't. Like, do you give yourself exit points?

You bet.

[Mark Papermaster]

Well, it comes into risk management. And I learned that lesson right away when I was at Apple. I had learned elements of that lesson across my career, but it couldn't have been more stark when I had just been on the job probably a month or so at Apple, and we were driving a version of the ipod that was going to go to market, and we were going to add a camera capability into the ipod, and we said, my God, this is gonna be incredible. It's gonna be really disruptive. I think about everyone having a camera within an ipod. And it had to be inexpensive. It had to fit into an ipod price point. But there we were, and it was clear we could not ramp up that capability, prove it out, and get it in time for the holidays. Well, think about ipods. They have to hit the holiday season. Of course, there's no ipods anymore. So the analogy would be, let's say an iPhone. You can't miss a product that has a seasonality associated with it. There was no hesitation by Steve Jobs at the time. He said, well, then it's out. It waits for the next year, done, move forward. You can't miss the seasonality. And that was an extreme lesson for me in exit point. And it paid off in huge dividends. And what we were doing at amd, remember what I said? You got to show up and be completely dependable every cycle. You got to win that trust, and you can't lose the trust of the customers. I told you about the Zen processor family. We created Splash. We disrupted the industry with its competitiveness. But on the second generation of that, we were looking at the metrics and I give my technical leaders and management leaders credit, we saw that we were off schedule and we said, my God, we're going to miss our window. And we pulled the plug and we had a backup path that our engineers had developed that we could create a version that we had of that processor, add some of the new features that we had planned for second generation. We pulled our risk mitigation lever, ended up with an equally or better second generation of that Zen processor than we originally designed. So you have to manage risk. You have to know what those variables are that are the risk gifts. You have to measure them along the way, and you have to put exit points in the process. Think about that chip I showed you, the GPU chip with all these chiplets. I will tell you, this is the riskiest thing we've ever done. We had to have multiple exit points on a number of the technologies. No one had done this before. So we measured along the way. We had test chips before we built the final chip. We brought the risk down wherever we could.

[Tomer Cohen]

So there's so much to learn here. I don't think it just applies to semiconductors. Mark, before we go to the future of industry, I have one of more of a how conversation when it comes to specifically the talent you manage. You know, one of the most amazing things right now is a lot of, whether it's productivity, innovation, a lot of what we can do right now with the power of AI has been transformed. I think about, like the process of building from coding to design to verification. When you think about how roles will be disconstructed and then reconstructed, do you think of some roles staying, some roles just morphing? Do you think there's okay from, you know, whatever, 15 job families, we're going to end up with two or three, and this work is going to be left undone by AI. So it's really what hasn't been solved yet. We need a human to basically run through that process, whether it's creativity, whether it's quality, whether it's the empathy of listening to the customer. Obviously hasn't, you know, will take a long time for AI to build empathy for anything. But just curious about how you think about it from not a roadmap for product, but a roadmap for people on your team.

[Mark Papermaster]

Well, the way to think about it is what are people doing as they're building products? There's part of what everyone does today in getting a product out which is very time consuming. It's not the work you love to do, but you gotta get it done. So it might be that, okay, I've got this product and I've got to test it and there's an infinite number of variables that I have to throw at it to make sure that could work in all those variables. Well, that's an ideal application that AI can help you with because you can train it on what are those variables you can have it help generate, how you can test that product to make it very, very efficient. It does involve a number of steps. But now with the gentic AI you can train different models in some of those steps, chain it together, and we are seeing good productivity gains in those type of tasks. Now let me contrast that to sessions I might have with my technical leaders, management leaders. We're brainstorming, looking at all the different variables that are coming at in our industry where the competition might be. We're on the whiteboard and we're doing a lot of what ifs. And it's a completely creative process that is not just based on data that I can train, but quite experiential and bringing teams together to have different experiences or from multiple disciplines, including having talked to customers and users of our product and knowing what really matters for them. No, we don't have AI that can do any of that right now. And so yes, we're at the early innings of AI. But the way to think about from a job standpoint is what you're gonna see in the next few years. You're going to see great gains in productivity for everybody using copilots and agentic AIs. So I think many walks of life will see that different and make us all more productive.

[Tomer Cohen]

Super helpful Mark on the future, five, 10 years, which for consumer company that's two way far in the future. But for semiconductors, I think it's actually in the realm of normal discourse. When you think about the applications that will exist that you need to build unique semiconductors for. You mentioned before, you're going to have chips in your body as one example. What are you envisioning that's going to happen that AMD needs to be ready for from a innovation standpoint?

[Mark Papermaster]

Well, the biggest disruptive trend of course is AI. AI is going to be built in everywhere. We're going to basically just have smart devices in all of walks of our life. Certainly what we're doing is how do we drive the fastest we possibly can, making the engines of AI more efficient. We call that a holistic design approach. Meaning we're thinking about how do we make it more efficient from every facet of the semiconductors, to the design elements of the chips that we do, to the hardware, software, and how we're working all the way through that entire holistic design approach. So that has to anticipate the future. How do you make it more energy efficient? We're energy gated. You have to think about new ways that even these chips can communicate. So photonics, optical interconnect is starting to get cost effective. New types of transistors that are yet more energy efficient. New types of memory, how you store the data. So every facet of how you build computing has tremendous room for innovation. And then you don't have to look beyond deep seq, deepseek challenged. Everyone said, look, there's different software, algorithmic approaches we can take that can be more efficient. In AI, it's a constant race for innovation. I don't see any lack of incredibly disruptive ideas on every aspect of that entire sphere of holistic design that I see in our field. It is an incredibly exciting future. Tomer the thing I worry about is just the pace. AI is enabling us to be more productive and we need it because that pace of innovation is on a double exponential, as you said. I've been decades in the industry. I love what we're going through right now and the innovation, but we are on a torrid pace. We need all the, all the productivity help we can get.

[Tomer Cohen]

Awesome.

So maybe we'll shift to some quick questions to wrap the session with you mentioned before, the chiplets and happy pool used to call them Chiclets and they kind of underplayed the power of them. I'm curious, when you're thinking right now, what's underrated technology that people are not bringing enough attention to?

[Mark Papermaster]

You know, I think it's really just, I'll just call it the broad term of integration. How you integrate what you put together, how you connect it more efficiently. History is littered with people that said, oh, I've got a great new computer chip, oh, but the software doesn't work on it. Okay. It's dead, right? So how you integrate, how you bring all the pieces together to bring new experiences and much more efficient computing, yet let it be deployed in a seamless way and ensure that all the applications can run. That is the holy grail.

[Tomer Cohen]

I'm curious if there is a myth about semiconductors that you always had to dispel or it felt like there was a misunderstanding of semiconductors.

[Mark Papermaster]

Yes, we have a constant myth. We deal with in semiconductors. And that is, oh, we're up against the wall. This is the last new transistor we're going to get. We've hit the limits of physics. In my decades in the industry, probably three or four times I had people tell me, we're done, there'll be no more innovation. They were wrong every time. Innovation is an indomitable force. The engineers and scientists of the world, you know, live and breathe finding a way around any wall you put in front of them. Innovation always rules the day. And so when people tell you it can't be done, I would take that with a healthy dose of salt.

[Tomer Cohen]

And Mark, lastly, you chosen a remarkable industry. You had a remarkable career at college. Just curious, college. Were you looking at different career paths or it from the beginning, electrical engineering, semiconductors kind of grabbed you?

[Mark Papermaster]

I was lucky, Tomer. I was in high school and working in a summer job at a hospital where there was an electrical engineer who was using computers to help solve cancer. I was age 16 and so that lit my interest in computing from a very, very early age.

[Tomer Cohen]

Wonderful. The passion started very, very early. Yes, Mark, thank you so, so much. I've learned so much from this interview. I really appreciate you spending time with us.

[Mark Papermaster]

Tomer, thank you for having me on your podcast.

[Tomer Cohen]

Thanks to Mark for sharing his experience with us. Let's dive into my main takeaways. First, as Mark said, there's never just.

One thing that makes a great product.

It's the why, not just the what. Mark is drawing a compelling contrast between IBM's systematic process orientation and Apple's bold vision and user experience focus. Working with Steve Jobs taught him the courage to say no and prioritize the end experience above specs, a mindset that now deeply informs AMD's product philosophy. Second, the importance of big bets. If you're already struggling, incrementalism will rarely make a difference in your business future. You need to make big bets to improve your offering in a meaningful way. Once you decide on a path, you work really hard to reduce the chance of failure and build multiple exit points. This is a really important learning. Big bets require big exit plans, and Mark shares a rare look into risk management in the semiconductor industry. He talks about building multiple fallbacks into long product cycles, a really powerful discipline in an industry with irreversible high cost. This mindset reflects a fusion of bold vision with engineering pragmatism, making big bets safe enough to survive. Lastly, don't be afraid to shift the paradigm. For decades, the chip industry followed Moore's law setting its goal at doubling the processing power of chips roughly every two years. But a decade ago, physical limitations caught up with the business. This sort of innovation became exponentially harder and very costly to achieve. Instead of trying to get incremental gains by doing what was done before, AMD made a bet on a chipotle architecture integrating different types of processors to arrive at a modular solution. AMD broke norms and was even mocked for it.

[Mark Papermaster]

As Mark said, some people ridiculed us. They said, oh, you're putting a bunch of Chiclets together like candy.

[Tomer Cohen]

But eventually the industry adopted AMD's practices. This is an example of not just.

Technical ingenuity, but also cultural risk taking.

AMD navigated the slowdown of Moore's Law by reframing chip design, betting on manufacturing efficiency and scalability before it was obvious to others. I'm Tomer Coyne. Thank you for listening. I've learned a lot from this conversation with Mark, and I hope you did as well.

[F]

We'll be back in two weeks with IKEA of Sweden's Managing director, Fredrika Inger. If you liked this episode, don't forget to rate and review us on Apple podcasts. It'll help people discover the show. Building one is a production of LinkedIn News. Our host is Tomer Cohen, LinkedIn's chief product officer. This episode was produced by Max Miller. Our associate producer is Rachel Karp. We're engineered and mixed by Asaf Gadron, and we get additional production support from Alicia Mann at LinkedIn News. Sarah Storm is senior producer. Dave Pond is head of productions and creative operations. Maya Pope Chapelle is director of content and audience development. Courtney Koop is head of original programming. Dan Roth is the editor in chief of LinkedIn. If you know a product leader we can all learn from, send us a [e-mail].