Advanced Micro Devices, Inc. (AMD) Earnings Call Transcript & Summary

Why don't we go ahead and get started? It's my pleasure this morning to introduce Dr. Lisa Su, President and CEO of AMD. Also in attendance is Laura Graves, who's IR, VP of Investor Relations. The format today is fairly straightforward. We've got about 30 minutes to go through a fireside chat, which I will start. There are mics that are going around the room. So if you have any questions, please raise your hand, and we'll try to get you a mic and kind of give you a question. Lisa, first, welcome. It's great to finally do one of these things again in person, face to face. So I'm glad that you're here participating in the conference this year and very much appreciate it.

The first question that I always try to ask is somewhat open-ended. You've had a tremendous journey over the last 5 years, reshaping AMD, turning it around, giving it a better footing. I guess the question that I get asked often is given all the success you've had, what's in store for [ AC 2 ]? What's the value proposition for the people in this room from here? If you could kind of walk through that, that would be extremely helpful.

Yes. Great. Well, thank you, John. It's good to be here, and thanks for having us. Look, I think for AMD, we really have been -- our mission is around high-performance computing, and really driving high-performance computing. It's an area that we think is extremely exciting. It's become even more important, more central to everything that we do. And so when we think about that in the core markets that we're in, PCs, gaming and data center, these are large, really exciting markets that need more computing every day. And we've driven a very strong road map over the last 5 years. I think we're excited about sort of the road map, the customer set, the traction that we have. And what I see over the next 5 years is really a continued expansion of our market opportunity because computing is only becoming more important as well as more complex, right? We're seeing a lot more workload optimized, computing need for CPUs and GPUs as well as ASICs and other accelerators. We're very excited about our Xilinx acquisition and closing that by the end of the year. So overall, as we put all of that together, I think we see just a tremendous opportunity as we go through the next 5 years.

Well, Lisa, I've got a lot of company-specific questions, but one of the overarching themes right now in the semi industry is just the supply shortage, the logistic constraints. If I look at where you started your guidance at the beginning of the year and where you're ending up at the end of the year, you've done a tremendous job of finding incremental supply in an extremely tight market. A couple of questions around that. How do you do it, one. Two, what's the outlook over the next couple of quarters into 2022? Because it still feels like as well as you've done this year, supply did kind of hold back further revenue growth.

Well, I would say when we really saw the sort of the pandemic-driven demand, which was probably really the second half of last year, we were very aggressive in ensuring that we work with our supply chain partners. And it's every aspect of the supply chain, right? It's wafers, it's back-end capacity, it's substrates, it's logistics, all of that stuff coming together, and we have made progress. I would say the entire semiconductor industry has made progress. I mean a lot said about supply chain shortages, but I think we also have to recognize that the demand in 2021 is so much higher than any of us thought 4 quarters ago. We've done well. We continue to work very closely with all aspects of the supply chain. We brought on a lot of new partners, and they performed well. And we've also made forward investments. So investments for AMD-specific capacity over the next 3 or 4 years to support our growth. So as far as I see today, I see that we're going to continue to put on more capacity as we go through 2022. The 2022 demand environment looks like it's shaping up very nicely across all of our market segments. And so we continue to believe this is a great growth market for us. And on top of that, it's not just the market, but I think we have also -- our product portfolio is well positioned for us to continue to gain share.

Without giving a forecast, you've thrown out a long-term revenue CAGR of 20% and you've sort of beaten that number handily every year since you've given out that target. When you look at the supply environment today, is it at least supportive of that growth rate? Or are there constraints to that long-term growth rate because of where supply is?

Yes. So it's hard to imagine that it was only 18 months ago that we gave that long-term guidance of 20% CAGR. And we said with Xilinx, we expect to continue that 20% CAGR. And from what we see, we see that that's absolutely achievable. From a supply standpoint, from a demand standpoint, from a customer standpoint, I think there's a lot of desire to expand where we are positioned in the market, and we're putting on all aspects of the supply chain to ensure we can do that.

Lisa, I kind of want to go through your different business units and kind of some of the revenue drivers there. And let's just start with the core PC market. I think the big concern that the investment community has is that pre-COVID, PCs were running about 250 million, 260 million units a year, and now we're up to 350 million and change. I'd like to joke that pre-COVID, work from home meant mostly just returning e-mails and not actually doing real content creation. So I'm of the camp that COVID has structurally sort of reset the value of the PC and the notebook. I think there's a lot of skeptics out in the audience. What's your perspective on kind of the PC market? Is this the new sort of floor? Is this a growth market again? How do you think about PCs?

Yes. So I think the PC market has been really remarkable in how important it has become as a result of, let's call it, the new world. And I think that's great. I mean I think that's great. I think the idea that people used to have 1 PC per household and now it's 1 PC per person. And by the way, it's not just a basic PC, but it's a PC with a lot of feature function, collaboration capability, content creation capability. So I think PC is a good category. There's no question the last 2 years were very strong for PCs. I think as we go into 2022, we'll see exactly how the supply chain -- the total supply chain shapes up for the PC market. But overall, I mean, 350-plus-or-minus million units is a great market. And we see within the market as well, there is a preference for higher performance, more capability, more graphics capability, more integration capability. And all that plays into our strength of really driving sort of a stronger mix within the PC market as well.

Can you talk specifically about the opportunity in enterprise? You've always had a solid share within the consumer market. I think over the last several years, with your product portfolio, you've done a much better job kind of penetrating into that enterprise market. Is there still good share gain available to you within enterprise?

Yes. I think in the overall PC category, our focus has been on, let's call it, the higher end and really sort of improving sort of our mix of product. I think commercial is at the higher end. And commercial PCs, it's as like enterprise servers, it's one of those areas where track record is important. Multiple years of strong execution is important, a strong commercial sales force is important. And so I think we've made excellent progress. I'm really, really pleased with the progress we've made in commercial over the last 12 months. But I still think we're quite underrepresented for our capability. And so as we look at places we're investing, it is around the entire commercial go-to-market cycle as well as the fact that, frankly, our server road map has performed very well and enterprise is also a nice sort of other piece to the commercial PC discussion when you're talking to Fortune 1000 companies.

Lisa, when we get into the server market, it's clear that the third-party benchmarking still has you in a very solid, strong position, leadership position in a lot of different workloads. Going back to the PC, one of the questions I'm getting asked more frequently is now with Intel introducing Alder Lake, having a big little core, perhaps being more aggressive at TSMC to get that part to 3-nanometer more quickly. Help us understand kind of your road map within the PC market. And I think on the last quarter, you talked about increasing your investments specifically in that business unit. What is that going towards?

Yes. So again, our road map, whether you're talking about PCs or gaming or data center, it is about being the best and driving towards that leadership point. On the PC side, we're excited about our 2022 road map. We'll be talking about that more at CES. We think it's extremely competitive. I think the progress that we've made in notebooks over the past 4 or 5 quarters has been tremendous, both in terms of not just the quality of our product, but frankly, the quality of the partnership with the top OEMs. Because as we all know, the PC ends up being what you deliver at a system level. And so there's been a lot deeper optimization of the product portfolio. So I would say from a competitiveness standpoint, we feel very confident that we have a strong road map, not just into next year, but we have lots of things planned. And the idea is looking for ways to continue to enhance the user experience as we go forward and also broadening perhaps some of the reach that we have in the market.

And then staying inside of that business unit, I wanted to kind of pivot towards the graphics business. Had some interesting announcements earlier this -- I guess, this month at your data center event. And one of them is the idea that architecturally, you're moving graphics to sort of your tile chiplet strategy, similar to what you did in sort of the core CPU market. Help me understand how important that is to continued share gains within that market.

Well, graphics continues to be a very strategic area and one, I would say, that we still have a couple of steps to go. So when we look at our graphics and maybe we'll talk a little bit about gaming graphics a little later on, but on the data center graphics side, this is an absolute invest category for us. We see it as with everything going on around supercomputing, high-performance computing, AI, machine learning, all of those aspects, they're going to need GPUs. And we've taken a lot of the learnings from our CPU road map and frankly, sort of used some of those learnings on our data center GPU road map. So we just launched MI200, which is a flagship data center GPU for us. It's using a new architecture for our data center GPUs. It's also the GPU that's in the Frontier supercomputer that's currently being assembled and put together at Oak Ridge National Labs, and it's also the basis of a number of other wins that we have over the next couple of quarters. But more importantly, I think it's another data point of how we're evolving the road map. So the chiplet architecture, the tile architecture, that's been very successful for us in our CPU road map. It's a key piece of our GPU road map. Frankly, it's a key piece of everybody's road map now because it's the way that you get around some of the scaling challenges from a Moore's Law standpoint. I will say that we continue to innovate and partner very closely, and we have further sort of enhancements to this chiplet or tile architecture as we go forward.

Well, you brought up the supercompute HPC market. And I think that's been an area where a lot of your successes have probably gone a little bit less noticed on Wall Street. How important is it that you have both the CPU and the GPU? You're really the only company that can go there. And you were one of the first, if not the first, to talk about this notion of heterogeneous compute. To me, it seems like what you can do around resource sharing, owning the CPU and the GPU really has given you a strong leg up in that HPC supercompute market. Is that a fair assessment?

Yes. That's exactly right, John. I think if you think about the idea of heterogeneous compute, it's the idea that you have so many different workloads in the data center. And so you want to use the right combination of compute for the right workload. And so having the CPU and GPU capability, being able to optimize the interconnect and the packaging between them and all of the resources, as you said, so you're not having to, let's call it, suboptimize resources, but you can actually share and optimize resources. We think it's a key differentiator as we go forward. I think we've shown it in supercomputers. It's a place where in working with the National Labs, we've really thought about how there could be significant optimization of their code as they go forward, and we'll get more developers used to that environment going forward. And I think you'll see more from us. And I think I might just add, John, it's not just about CPUs and GPUs, right? But we see this as really sort of a continuum of compute. And you can imagine ASICs or other accelerators, they don't all have to be from AMD. Frankly, we believe in an open ecosystem. We see FPGAs and adaptive computing SoCs with Xilinx coming in as well. All of these then become part of our sort of tool chest as we're talking to the largest hyperscalers or the largest labs in the world and saying, "Hey, what are you trying to accomplish? And how do we work on that together with the most optimized compute ecosystem?"

Before shifting over to service, I did want to spend some time on the gaming graphics business. And you mentioned it earlier in one of your responses. The perception has been that what your competitor there has done around software has made it very difficult for you to kind of break into a broader footprint into that market. And now they've got ray tracing. Can you talk a little bit about kind of the road map you're seeing on your graphics business and what the expectation might be over the next year or 2 for shares in that business?

Yes. So the gaming graphics, first of all, gaming overall, I think we should all recognize has been just remarkable in terms of the overall growth, right? When you look at gaming, whether you call it across the PC form factor, the console form factor, whether it's in desktops or notebooks or as cloud gaming comes on board, there are just more and more gamers coming into the ecosystem. And that's good for any company that has graphics, right? So we view gaming as a very strategic segment because of the number of users it reaches. In terms of our discrete graphics road map, I actually think it's done very, very well. We started the rollout of our RDNA to architecture. I think, from a performance per watt and a performance standpoint, it's done extremely well. It has been supply constrained this year. I know lots of gamers continue to want new graphics cards, especially as they have a bit more time on their hands. We've ramped up capacity over the last couple of quarters. I think we're quite competitive. And I think we're even more excited about sort of our RDNA 3 or our next-generation road map. So you can imagine this idea of having a top to bottom stack that services desktop through notebook form factors, and it's the same architecture that goes into our PCs, our PC APUs. And so we've been doing quite a bit around sharing resources there as well. Same market sector that goes into game consoles. And so I think the architectural capability of our graphics has continued to evolve and be very competitive.

Lisa, you mentioned how tight supply is in that market right now. Despite that, I always get the question asked, how much of an influence crypto might be in the GPU market today? It seems a little bit silly given how lean channel inventories are, how tight things are to think that crypto is a meaningful driver of the business. But how would you couch that kind of risks-loss concern?

Yes. So I think it's a very different market than it was a few years ago with crypto. I think the -- certainly, for our graphics business, we don't believe it's a meaningful driver. What we are trying very hard to do is make sure that our graphics cards do, in fact, go to gamers and go to the retail channel and places where are very strategic. And everything else is being handled separately.

I wanted to switch gears to servers. Clearly, the most exciting part of the business over the last couple of years, both from a revenue growth perspective, and I would argue a multiple perspective on the stock. You had your Data Center Premiere event earlier this month, a lot of interesting new announcements. I'm wondering if you can just kind of recap, help us understand you're well into the 7-nanometer ramp for EPYC. You introduced Milan-X. I'm kind of curious about how that fits into the portfolio. You talked a little bit about Genoa. And then interestingly enough, you also brought out a piece of optimized sort of silicon for the data center market. Maybe you can talk about that part of the strategy as well.

Yes. The overall server sort of business evolution has gone just extremely well. I think I'm very, very happy with how it's gone. I think from a customer standpoint, the customer adoption has been tremendous. And many of these customers you work on for years. So we've been working for years with all of the cloud companies. We were happy to be able to announce Meta as the newest user of EPYC. But it just underscores the fact that we're in 10 of the largest hyperscalers. I would say we're nowhere near max penetration. So there's a significant opportunity for us to continue to expand our workloads in the partnerships with the cloud vendors. I think Milan has ramped extremely well. I think our socket compatibility between Milan and Rome and Naples has worked well for customers, so they've been able to bring it on quickly. We've seen just the adoption rate and excitement around the product go well. And then as we look forward, one of the benefits of the scale and the growth that we have is that we can invest more in R&D. And so this notion that workloads are different and workloads are going to want different compute is what drove us to have Milan-X. Milan-X is really sort of, let's call it, the highest-of-the-high performance for technical computing, especially there are workloads that need fewer cores but need more performance per core, need more cash. So Milan-X is a great option for EDA and things like that. And then as we go to Genoa, I think Genoa looks great. We're excited about Genoa Zen 4. I think that is yet, again, another big leadership message for us as we go into the next-generation I/O, the next-generation core, next-generation scale-out. But we thought that at that time point, it would also be useful to have sort of a cloud-optimized core, which is what Bergamo is or our Zen 4c. So a lot of code names, but I think that the takeaway for people is that, look, this business is a great growth business from the standpoint that the market is very strong. And then AMD within the market is participating in more and more of the segments. And by workload, by optimizing sort of the compute for the various workloads, I think we just become a deeper partner with our largest customers. And so that's t kind of the overall point.

Lisa, within the hyperscale market in your server business, one of the questions I get asked broadly, less about you than the industry, is this idea of general-purpose compute versus optimized silicon and which way the world is going. I think it's somewhat of a false premise because to me, from my perspective, a lot of people in the general-purpose compute world are doing more optimized silicon. I think that's been a big part of your strategy around heterogeneous, around your M&A targets and the like. But help us understand how you see that dynamic playing out. If we're in a 90-10 world today, are we going to 70-30? Or is that too much of an exaggeration? And kind of how do you leverage your IP to participate in that optimized world?

Yes. It depends on what you're calling sort of optimized. If you're talking about optimization being -- optimization, there's like this sort of very wide spectrum, right? And I would say pretty much every hyperscaler optimizes today in some way, shape or form, whether it's just depending on what they're trying to do. And we're working with each of them in terms of what optimization points are important. But if you go from that to -- yes, you might choose, for example, some of the compute may be more cloud optimized in the sense that, hey, you may not be so focused on a single-core performance, but you're really focused on throughput. Some may be more looking for the leading-edge performance. Some may want ASICs. Some may want some different combination of CPUs and GPUs. That's the whole range of optimization. And our view of the world is, yes, there's going to be optimization. And actually, optimization is a good thing. I mean optimization is the idea that everyone needs more compute, and you want to do it as efficiently as possible. But there's no near -- like nowhere near saturation of compute, whether it's general purpose or specialty. And our view of the world is there are some people who want to develop their own silicon, and that's totally cool. We'd want you to use our general-purpose compute in our ecosystem, and we're going to work with you on that. There are a lot of people who just want optimized compute, and we think we can be a great partner for that optimization. And so I don't see it as anything other than the natural evolution when you have an area where the demand has gotten so high that optimization actually can make a big difference.

Lisa, you mentioned Meta/Facebook earlier. It was great to be able to announce them as a customer earlier this month. You've been working with them for a while. Can you help us size that opportunity? And if you don't want to talk about a specific customer, can you help us understand, once you land at a hyperscale company, what's the opportunity to expand once you've gotten your foot in the door?

Yes. So again, not talking about any specific customer, but just as a more general statement. What I would say in general with the hyperscalers is the opportunity is tremendous. And the way it typically works, what we've seen with some of our other very deep partnerships is in the first generation of adoption, you might expect some number of workloads, let's call it, a piece of the portfolio that is -- that has AMD in it. And as we go through the next generation and the following generation, the number of workloads expand. So my view of the matter is we have a significant growth in the cloud. Although we've grown a lot already, we have significant growth in the cloud going forward because we're deepening those partnerships going forward. And all of them need more compute. So yes, I think we see data center, particularly the server CPU market as a very strong growth driver for the next few years.

Shifting gears towards the enterprise side of the market. It's a little bit harder of a nut to crack than hyperscale, but you've done a great job kind of getting traction within that side of the business. Can you help us understand the opportunity over the next 12 to 24 months? And at an end state, how would you think of the split of your server business longer term between sort of enterprise and hyperscale?

Yes. So our business is more cloud weighted today. Although I will say here in the third quarter and sort of in the second half of the year, the enterprise business has picked up very nicely. I think Milan is a sort of a strong -- a very strong product in there, not just again because of the product, but also because of the coverage that we have. What's most important in enterprise is, one, ensuring that the OEMs get their platforms up and available as soon as possible. And so our OEM partners like Dell and HPE and Lenovo and Supermicro, they've all been great in sort of their focus on AMD. And then as I said, we also have a strong software optimization team, application optimization team, together with an enterprise sales force that is approaching enterprise customers. So we made great progress there. I think we expect the enterprise business to continue to be a strong growth driver for us. In terms of the mix, I mean, it's hard to call mix far in advance, but I think in general, cloud will probably still be a larger piece of our business, just given the dynamics of the business. But I would say enterprise would be a very significant piece of it.

A year ago, virtually, I complain that I couldn't find an Xbox or a PlayStation. We're a year later, I still can't find them. I'm pretty cheap. I don't want to pay 2 or 3x the retail.

You can find them, you're not just willing to pay for them.

I'm just curious, the strength of the business there. How much longer does it persist? And I guess importantly, expanding beyond just sort of the gaming console in an environment where supply is this tight, do we just throw seasonality out of the window? And is your growth over the next couple of quarters is just going to be a function of how much incremental supply you can get?

Well, I think -- first of all, I think the console business, this console cycle and all when you look at it in totality, has been amazing. I mean if you think about the strength of the cycle, now we're sort of deep into the second year and the fact that the demand is so high. I mean we've been shipping a lot of product. So the fact that the demand is still high says something about just sort of the product capability that's been captured by both Microsoft and Sony in this console cycle. We're continuing to ramp up production. We expect 2022 to be another strong growth year for consoles. If you look at the typical console cycle, the peak year is actually the fourth year typically. And so you would expect maybe 2023 would be the peak year. But to the overall point is, look, it's a great platform. We love the fact that we have so many users that are using AMD technology on the console platform and we'll continue to ramp up supply. And from a seasonality standpoint, yes, you're right. Seasonality does become more muted in this environment until we get to sort of a more normalized environment.

Do you think that holds true for your PC and graphics business as well as we turn the calendar to 2022?

I think if you look at those businesses, you might see a bit of seasonality, maybe not typical seasonality, but I learned there's a little bit of clarity here that better than seasonal means that typically, if you go from Q4 to Q1, you might see a little bit more of a -- you would usually see consumer down. You might see it less down than normal just given the supply patterns. But that being the case, I mean, what we see is just very good visibility across our market segments. And in data center, you sort of expect that visibility. But in PCs and gaming, we're seeing just very good visibility all through 2022, which gives us the view that the market is strong. I mean the market is strong, and we're preparing for that strength.

Lisa, we, in this room, take all of your hard work and boil it down to a P&L and that's sort of your report card. One of the things I wanted to highlight on the P&L is just the gross margin. It's been a fantastic story. I would argue it's being understated by the gaming mix, even though gaming gross margins are coming up. How should we think about kind of long-term gross margins? Because I can make the argument that if I x out your gaming business, you're well above 50% gross margin today in the rest of the business. Where is the ceiling in your mind?

Yes. So I think that's true. I think the gross margin performance for us, and I've said this from the beginning, it's all about mix, right? It's all about mix. We've done quite a bit to be very strategic in how we're approaching the market. So if you look at sort of the ASP expansion that we've seen, frankly, across all of the businesses, it's because we're providing more value and we're targeting sort of the places where we can differentiate more. So we're pleased with our gross margin progress. I would say it's a bit ahead of our -- the financial model that we had previously talked about. And we'll update our financial model as we complete the Xilinx acquisition. So next year, we'll update the model. But yes, we're going to continue to focus on the premium segments of the market, and that should be positive from a gross margin standpoint.

Great. With that, we've ended the time in this room, but I want to thank everyone who joined us, especially Lisa for joining us this morning. This was great. Thank you very much.

Fantastic. Thank you, John.