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

Good morning, everyone, or good afternoon for those of you not in the West Coast. Again, I'm Ross Seymore from Deutsche Bank, and we're very pleased to have Devinder Kumar, the CFO of AMD, here with us for our next fireside chat. So Devinder, thank you so much for attending. We'd love to have this be in person, but better to see you one way than not at all.

So why don't we just jump into some of the questions that I have? The last year has been -- well, last few years, but most particularly, this past year has been a great one for AMD. You started the year guiding to 37% revenue growth for the full year, then you upped it to 50%. And then the most recent quarter, you upped it to 60%. How much of that is due to the supply coming on? And how much of it is due to demand and kind of secular market share gains that AMD is being able to take advantage of?

Thank you. And it's really good to see you in a virtual mode. I know we're looking forward to doing it in person, but next year, we will do it in person, and that'll be good to do that from that standpoint. To your question, we started the year with 37% increase in revenue. And really, it was early in the year, we wanted to make sure from a supply standpoint, in particular, that we could be confident in terms of everything we need, from wafers to substrates and also to the ATMP capacity. We increased that to 50% and then 50% to 60%. I would say that the demand for the products was strong and actually has gotten stronger, in particular, with the products that we have introduced over the year, including the data center space and the PC space, the demand has really gotten stronger. And from a supply standpoint, it become more fortified. And right now, when people ask the question about 60% growth year-on-year in a space where there is significant shortages in some areas of the semiconductor industry, how do you do that? It's all supply supported. The revenue growth is there. We feel good about 45% growth last year and guiding to 60% this year, and the supply is there from an overall standpoint. We also believe that we can continue this from an overall standpoint with all the work that we are doing from an overall AMD standpoint, both with the customers on the demand side and also supply with our foundry partners, subject suppliers and ATMP capacity and feel that there's significant room for us to grow out into 2022 and 2023.

So the supply constraints, just to follow up on that, you rattled off a few of the areas, substrates, et cetera, where there are shortages. How do you see that being resolved? Obviously, it might be holding back some of your growth, but 45% last year and 60% this year plus are nothing to sneeze at, to put mildly. Do you expect it to get resolved in 2022? And if so, is there a catch-up that AMD should benefit from even more revenue growth as you catch up to true demand?

I wouldn't be smart enough to say when it exactly may get resolved. But I will tell you that I think the situation continues to get better, as there has been the realization that this demand is for real and there are customers demanding the technology and the products, in particular, where the competitiveness of the products is there. It continues to get better. And even from our standpoint, as we get to the 2022 and work with our partners across the board, the supply situation continues to get better. Exactly when it will get resolved, really depends upon the balance from an overall standpoint because it's not just within our industry, but many industries are affected given the usage of semiconductors across the board. And I think when you look at it from our supplier standpoint, they basically just had to put more capacity in place given the demand/supply imbalance right now.

And how does AMD prioritize where it allocates that supply? You have your client business, your graphics business, the server business, the semi-custom business. You have a bunch of other details within that. But in those 4 buckets, is it something that there's a clear hierarchy of where you prioritize? Or does it rotate around depending upon seasonality and specific customer demands?

Well, there are 2 things, right? The first priority is if we made a commitment to our customer, we want to make sure that we can supply the parts to the customers. And overall, from a viewpoint of forecast from our customers and us putting in place the capacity and getting the wafers or substrates or ATMP capacity, we want to make sure that we balance all of that. And in some sense, the relationship of sharing that information within customers and suppliers has gotten better. Many of our customers know that if they don't give their signals with the specificity of timing in particular as to when they need the parts, they may not get the parts. So that's priority #1. You made a commitment, you go ahead and deliver it with the parts. Then the second thing you look at it from an overall standpoint, demand versus supply where, obviously, we can prioritize the data center is a very high priority from that standpoint. And then the PC business, in particular, with the higher end of the stack, we want to make sure that we can maintain and continue to grow the business in that area. That's a very high priority, and then you take it from there. It is not as simple as I make it to be. It's a pretty complex equation because you have customers and you need to make sure that they have the rest of the parts they need to go ahead and take your products and actually ship customers to the end. And customers, for example, in the PC space, you hear about [ max outs ], you can get the chip but you may not be able to get other components. And that is something that we take into consideration as we go ahead and supply the units. Because we also don't want to have a situation in any particular area where there's an over-inventory situation just because we put the parts out there from our product standpoint.

Yes. If you can get to an over-inventory situation, you will probably be the only one in semis with that problem, at least as of now.

That's right.

Maybe a problem down the road for semis, but in the near term, it doesn't seem to be an issue. Why don't we pivot over to some of that area you just talked about in prioritizing and specifically go into your data center business? You break it out in its entirety, but I want to just split it a little more finely into what you do on the server side in EESC as well as what you do on the data center side for graphics. So on the server CPU side, where do you think we are in that cycle? There's been some debate about digestion periods and source of demand from cloud versus enterprise. So by my math, you don't have to endorse it all. It seems like it's almost doubling again this year for AMD and roughly $3 billion in revenue. If we were just to say what the drivers of that growth would be, is that mainly cloud? And then where do you think that profile of demand looks like as you look forward?

We believe the industry is entering a high-performance megacycle with potential for very strong growth, in particular, for AMD and the data center TAM is very large. Product leadership of EPYC, we'll talk about, I'm sure, with the transition of Rome to Milan and Genoa coming next year. And the current sense of revenue, even with the numbers that you just referenced, is fairly small from an overall revenue share standpoint, and we feel there is opportunity for growth. We've seen strength across the cloud and HPC, which have been traditionally strong for AMD. But we've also seen with Milan, in particular, strong interest on the enterprise. So if you look at it from a first half to second half standpoint, yes, the cloud was very strong and continues to be strong. But the enterprise growth, from an overall standpoint, second half to first half is where it's coming from in the enterprise space. We expect in the second half that the Milan product, in particular, will ramp very quickly and crossover Rome in terms of revenue. Very good momentum in the enterprise space, strong breadth of OEM and enterprise platforms with all of the work that we have done with our customers, in particular, to put in platforms across the globe and with our customers. So second half 2021, we believe enterprise will be a stronger component for us than the first half, with cloud also continuing to grow in the second half compared to the first half.

And I would assume, and there's no shame in it, far from it, that the enterprise side will still be relatively smaller as a percentage of total sales, even though in the second half, it sounds like it will grow faster.

Yes. So I think overall, the cloud has been very strong. The business is very strong there. Enterprise continues to grow. And obviously, that's benefiting us from a revenue standpoint going into 2022, in particular, with the Rome product transitioning to Milan crossover happen in the second half. And then the new product, Genoa, coming out in the 2022 time frame.

And without going too far down the technology side of things, what is the reason that the enterprise is starting to adopt Milan? Is it an architectural change to some level? Is it the fact that enterprise wasn't spending in aggregate and now they are? Or is it just a comfort level with AMD and it takes longer for enterprise to ramp versus the cloud?

I think it might be all of what you just referenced, Ross. Really, it is the trust we have built up with our customers and our customers' customers in the enterprise space. But performance of the product is very, very competitive. We said earlier this year that we feel very good about Milan. And now that Milan samples are out there and they have a platform that they've launched and people are buying the product, they can actually see the competitive sort of product. And that's driving not just the excitement, but also the trajectory in terms of demand. And obviously, we are supplying the product from that standpoint. And that is something that's very, very important. Because in the end, performance matters and performance is what people want to get their hands on. And the Milan product is delivering on what we had committed to deliver from a performance standpoint, and that's really driving the demand for the product.

And then the last question on the data center CPU side of things, in particular, specifically, is the competitive landscape in aggregate, it's not just your other x86 competitor trying to rejuvenate itself. It's also some vertically integrated folks doing ARM-based processors, ASICs, et cetera. It doesn't seem to be slowing down AMD's business whatsoever. But from a CFO's perspective, as you look at it, are you seeing any sort of signs of competitive intensity rising as far as what that means for pricing or profitability or kind of your traction? Or is the market just so big that you have plenty of room to grow even if those dynamics do assert themselves down the road?

I would say, Mark and Lisa are better people to ask those questions, but I'll tell you from my standpoint, when you look at compute solutions, whether it's x86 or ARM or even other areas, that is an area of focus and investment for us. You read about the Google Tau VM announcement that Google selected AMD's EPYC product, and we feel good about that. Because in the end, we want to deliver high-performance compute solution to our customers, and it's really the solutions that are important. We know compute really very well. Even ARM, as you referenced, we have a very good relationship with ARM, and we understand that, and our customers want to work with us with that particular product to deliver the solutions. We stand ready to go ahead and do that even though it's not x86. Although we believe x86 is a dominant strength in that area, but there are other areas we're willing to partner with customers of ours to go ahead and deliver those compute solutions. Because that's what we believe, from my standpoint at the CFO level, it's all about deliver the solutions, deliver what the customers want on a compute standpoint, and obviously, you have a very good relationship and good product demand in that particular area.

So I want to stay within data center, but maybe pivot a little bit to also including the GPU side of things, with the AMD Instinct accelerator side. What's the customer reaction to that been? We can see the design wins in the supercomputing side, where AMD has done very, very well. And I want to ask about some of the timing of that because I think it should be rolling out, at least with Frontier, very soon. But in aggregate, what's been the response of AMD's entrance into that data center accelerator market from your GPU side?

I think the customers love it. They want choice. Last year, we launched AMD Instinct MI100 product, the world's fastest HPC GPU at that time based on our CDNA architecture, and that's very well received. The supercomputer wins Frontier revenue in the second half of 2021 with AMD EPYC CPUs. And then obviously, you have the CDNA 2 architecture from that standpoint. And our approach really is if you have those kind of solutions, shared access of data and memory, people like that and customers like that. And that is what's driving the revenue. The Frontier revenue, like I said, starts in the second half of 2021. CPU and GPU content is a portion of that, and that is something that you will start seeing us talk about as we get to the second half in the data center GPU space.

So was that the Frontier win and the revenues from that, is that included in the 60% revenue growth guidance you gave for the full year?

Yes. That's included. That's included. It's about all in, call it, about $600-plus million revenue opportunity largely in the second half of 2021. It's included in that.

Got it. And how do you -- you have multiple of these exascale compute wins, high-performance compute wins. So maybe the lumpiness can be smoothed out. But how do you view that? If you have, say, $600 million in the back half of this year, that's great. But does that become a problem in the first half of next year for those that are looking at the world on a sequential basis? And I realize that might not be your target investor.

I think it's all about scale, right? If you talk about going back a few years ago, $600 million relative to our total revenue would have been significant in the half of the year. But the revenue of the company has grown quite significantly, where we were, if you go back to the 2019 time frame, under $7 billion of revenue. In 2020, the revenue grew 45% this year, on target to grow 60%. So all of a sudden, you're talking about $15-plus billion of revenue. And when you take that and smooth that around, it is lumpy from that standpoint, and we are very clear and have been very specific that it is lumpy, and we will be very transparent in terms of any particular area where there is lumpiness in the business. It is a lumpy business but is a very good business from a standpoint of what you can achieve there and also the margins that can come with that kind of increase in revenue in the data center GPU space.

And again, maybe the last question on this and not going down the technology road too far, but you're unique, in general, as of at least today, that you have the CPU and the GPU to address these markets. I know Intel has aspirations of doing the same thing, but how beneficial is it to have both sides of that market to get the design wins that AMD has had, which seemed to be bigger or a greater number of them than most of your competitors.

I think it's a realization. As we've said consistently for many, many years that when you're working with one company with both the CPU and GPU, and it can deliver the performance as needed, whether it's Frontier and then in 2023 time frame, we have the El Capitan product coming on board. That is something that's very unique, and AMD can provide that today from an overall standpoint. And that is obviously, we recognize, in particular, in the supercomputer area, which are very competitive. And these were the first wins where one company got selected to deliver both the CPUs and the GPUs for wins like the supercomputer Frontier and then El Capitan coming later in time.

So last question on these supercomputer rollouts. Like you said, they're lumpy, but they're still great business, not only from a financial point of view but an endorsement of your technology. What's the duration of those ramps for any single 1 of them. When you say roughly $600 million in the back half of this year, is that all it takes and then you wait for a year, 1.5 years before you would have the El Capitan? Or is there some longer tail to these types of design wins? It doesn't -- your answer doesn't have to be specific to any one.

Yes. It's always several quarters without the specificity, but it is over several quarters that you go ahead or recognize the revenue from that standpoint.

Got you. Last question then on that. So is the cadence Frontier will be late this year and then into next year to whatever degree, and then we have about a year of wait after that before El Cap takes off?

That's correct. That's true.

Got you. Okay. What's -- as far as accelerators go, data center exposure goes, it's a little bit of a tough segue here. But how does Xilinx fit into that? Obviously, you guys are trying to buy them. You're still waiting for approval. Knock on wood, hopefully, it happens later this year on schedule. I don't need an update on that because I know that's a bit of a black box. But how does Xilinx fit into that data center strategy?

I think on the question that maybe you want to know, I think we are on track, we believe, to close the transaction before the end of the year. We have obviously looked at Xilinx, announced the acquisition last year and believe we'll close it by the end of this year, strong leadership road map for us in CPUs and GPUs. And Xilinx brings us the ability to participate in a larger footprint in the technology space with the FPGAs and adaptive computing, as you will hear a lot more about that as we talk about it as we consummate the transaction. From a financial standpoint, the TAM of the company combined, $110 billion. And also from a financial standpoint, Xilinx has very good margins. And from my standpoint, as you know, Ross, we've talked about it over the years, AMD's margins have come up from the low 30s and this year to the 48% and going on to the 50-plus percent. But acquiring a company that has significantly higher gross margins is very attractive from that standpoint. But really, it is about the diversification, complementary product space, adaptive computing and FPGAs that Xilinx brings to us. And as a combined company, as being able to participate in a greater footprint in the technology space and therefore, really the right step for us in consummating the transaction.

Then the last question on that deal and Xilinx specifically, I believe when you announced the deal about a year ago, roughly, Lisa talked about still being committed to 20% growth, the CAGR target that you have for revenue. I don't expect an update on that today, but just want to make sure everybody is interpreting that the correct way. Xilinx has not grown 20% in any CAGR basis over multiple years, and that's fine. They're still a very, very good quality company. But to the extent that grows closer to 10%, is the interpretation that the combination between the 2 companies can accelerate the Xilinx side? Or is the interpretation that Lisa and your confidence in AMD's growth rate being well above the 20% is how you get the combined entity to still grow 20%?

I think it's the latter rather than the former. I think Xilinx, there are some opportunities in terms of what you think about 2 companies coming together and getting some benefit from a revenue standpoint in combination with customers and technology. But really, it is AMD from an organic standpoint that drives the growth. And Lisa did say at the time of acquisition that, "Hey, we think from an overall standpoint, the combined companies can grow and the levels of that, definitely." You've seen that from where we sit, a 20% compound annual growth rate in the March 2020 time frame when we had our Financial Analyst Day, and we've done 45% and 60% and believe that we can continue to do that because we are underrepresented from an overall AMD organic TAM standpoint. The TAM is roughly about $80 billion. And even with the guidance that we have right now, if you add up all of 2021, it is still $15 billion plus so we are representative. Customer preference for our products and the road map, in particular, is growing. And we believe, therefore, we can continue to grow much more significantly than the semiconductor industry as a whole and deliver to the revenue growth that you just referenced.

Great. Why don't we pivot over to the C&G side, the Computing and Graphics side of your business? And let's talk about the computing side or what you guys would call the client side. Intel talked about a high level of confidence in the PC market remaining a structurally faster-growth segment going forward post COVID and work from home than it was pre. The flip side is many people are concerned that work from home really pulled in demand as everybody had to buy extra PCs to work from home for themselves and their kids, et cetera. Where does AMD fall in that view of the PC market going forward?

It is a debate. I mean the folks I have talked about, very high growth and continue at the high level. And obviously, there are others that say, well, at some point, this income will stop and you could have lower PC demand out there. From our standpoint, as I said earlier, now we are significantly underrepresented from a revenue standpoint. We have gained revenue over the last couple of years. And if you look at revenue share just on a 2-year basis, a couple of years ago to now, we have essentially doubled the revenue share from an overall standpoint. Unit share is important. We have moved up the stack. The premium product that we have introduced helps us from a viewpoint of the overall revenue share. And we believe that we can continue growing our revenue in the PC client space out in time. And yes, there will be the ups and downs in terms of the demand situation and the total unit standpoint. But for us, we are very focused on revenue share to go ahead and continue growing that because we believe even at this level that we are in right now, we have significantly lower revenue share in terms of where we're going to get to grow our revenue share gain, in particular, with the premium products and the comparative products that we're introducing.

So how -- AMD has another interesting dynamic in the work-from-home period that it ended up being significantly -- a significant accelerator for the consumer side of things, while the enterprise side weakened. You guys have more exposure historically to the consumer side, so that was a great tailwind. But now you had a significant increase in design wins on the enterprise side. So how do we think of that trade-off in the sources of demand? Even if consumer side slows post work from home, is the enterprise side enough to keep the share gains going at the same pace? How does that trade-off work in your mind?

I think the commercial and enterprise space, where we have made strides in the revenue share, really benefits it from that standpoint. It's a tailwind. It's a tailwind to the revenue that we just talked about. Having the product from top to bottom, working very actively with customers to design platforms that meet the premium needs, whether it's a notebook space. And even in the consumer space, the premium notebook space, we've invested in that. You've heard us talk about R&D investments. You hear us talk about go-to-market investments in terms of the platforms with our OEM customers. And whether it's commercial or gaming or premium consumer segments of the market, that's a big focus for us. And as you observed, very strong in the consumer space historically, and this is the area that we continue to make very good progress in partnership with our customers.

So the other side of the computing and graphics is obviously the graphic side of things. I know you talked a little bit about the data center side earlier, the data center GPUs. But in general, by my math, the G side of C&G is actually percentage-wise grown significantly stronger over the last 3, 4 quarters than has the client side. Dollar-wise, they're much more even. What's driven the acceleration on the client graphics side of things? Or maybe more precisely, what's driven the acceleration in the graphics side? And is it client or is it data center? And then is there any crypto in it? So lots of questions for you.

So I'll answer the third question first, crypto, negligible. That's not a priority for us. We do not prioritize our product or make them for the crypto folks. It's more for the gamers, and that's a high priority from that standpoint. What's driven the growth is we had the Radeon 6000 Series high-end GPUs introduced, very competitive. And that is driving the growth in the GPU space. And then on the architecture side with the CDNA 2 architecture that we have that we are coupling with these products, and those are really driving better competitive products overall, and that is driving the revenue growth in the graphic space. Now to be fair, on the client PC space, if you go back into the 2019 to 2020 time frame, there was significant growth from an overall standpoint. And obviously, we continue that becomes percentage-wise, as you observe a lot smaller, but it's a much higher base that from a revenue standpoint as compared to the graphic side of it within the Computing and Graphics segment.

So the competitive landscape on the graphics side, the client graphics side is different. It might get to be more similar if your compute competitor succeeds and what their aspirations are. But generally speaking, from a high level, do you see any structural reason why the share gains you've delivered on the client side can't be at least directionally replicated on the graphics side?

I think it's really the growth trajectory, right? Those businesses have done very well with the competitive landscape, larger base of revenue, and we continue to grow revenue in that area. We expect to see continued strong demand for gaming graphics, but it is not the largest driver of revenue given the growth trajectory of these other businesses coming off of a higher base. You referenced the data center CPU numbers earlier, and we talked about the client PC side, but that's kind of the way -- is the mix of revenue from an overall standpoint.

But if the graphics keeps doing what it's done this year, especially if you get the back half of the year benefit from the data center graphics side of things, at least by my math, you're talking about a multibillion dollar business. So I appreciate it's relatively smaller than the others, but it's not meaningless. That's still a pretty meaningful dollar amount in my book.

And you're absolutely right, Ross. If you take in combination gaming graphics and you take the data center graphics piece of it, it is a significant portion of our business.

Got it. So the last question I had on the revenue side of things is the semi-custom business. Obviously been a huge success story over the years, one of the biggest parts of the early stage of the turnaround of AMD. Talk a little bit about where we are in that rollout. I would assume those are the exact customers that you have to meet the promises for given the ASIC nature of that. But more importantly, where are we in that rollout of the generations of game products?

Product introduced in the back half of 2020, very strong first half of the year for the semi-custom game console product. And you observed right. I mean the commitments we make to our customers since in those areas, we are the sole supplier. Very important for us to meet the commitments in terms of the product and the product demand continues to be very high. We expect it to continue to stay strong in 2021 and into 2022. Seasonally, patterns in future years will come back to the -- it's high in Q4 and then low in Q1. But right now, at least for this year and next year, we don't see that. We see continuing strength, second half to the first half from a semi-custom game console revenue standpoint. And we are still ramping the product life cycle in the first couple of years. And typically, after the year of introduction, the first couple of years, revenue grows and then it grows again, and that's exactly what we are seeing. And actually, on higher demand than what had been there than prior cycles.

So why don't we switch over to margins in the last 5 minutes because I don't mean to only give it 5 minutes, but the revenue side is worthy of a lot of discussion. But the result is the margins have been great as well. And so last year, I think your gross margin grew 3 or 4 points or this year over last year. That's especially impressive considering what's happened in the semi-custom side of things and how fast it's grown. What have been the biggest drivers of the margin improvement? Is it mix between segments, mix within segments or both?

I think the biggest driver is data center on the CPU side. That revenue growing helps the margin for the company because those margins are significantly higher than corporate average. And that, from an overall revenue mix standpoint at the company level, very favorable to the margin side of it. Then when you look at it within the segment, let's say, you look at the client PC space that we talked about earlier, higher-end commercial, higher-end gaming, premium consumer products helps drive the margin in that particular business, and that helps your overall margin for the company. Client PC, as I said, we are prioritizing commercial gaming, premium customer working with our PC vendors for higher-end products that command better ASPs and therefore, benefit the margin side of the equation. That's very important. And then graphics, we talked about earlier, high-end Radeon 6000 for channel and gaming notebooks helped the margin from that standpoint, even in that business, improving the margin healthy overall. So really, if you put it in a nutshell, it's across the board, right? You have data center becoming a larger part of the revenue of the company. That helps margin. And then in each of the businesses, client PC and graphics continue to improve the margin. You referenced semi-custom earlier. Initial launch in semi-custom, their use tend to be lower, then they get better. And even that business, even though from an overall standpoint is below corporate average, as we improve the margin in that area, that also helps. The one thing I feel good about and, Ross, is interesting you say in 45 minutes or so, we talked 5 minutes for gross margin, I've been in sessions like this where 90% of the time is spent talking about gross margin than anything else. We've come a long way from that. Just a few years ago, the margins of the company was in the low 30s. And here we are guiding this year at 48%. And we believe there's more to come from an overall standpoint to continue to improve, not just the revenue, but also the margin of the company from an overall standpoint. The customer preference of our products helps a lot, the fact that we are underrepresented, in particular, in the premium space in many of our business areas helps a lot. And that is something that can continue to drive the revenue of the company, but also improving the margin out in time.

So the last question I have in the last 2 minutes. I'll make it a quick one, I hope. You guys have also allowed that gross margin goodness to flow through nicely to the operating margin side of things. Your OpEx intensity is actually below your 26% to 27% target for this year. If I put you on the spot, do you think there's more or less further leverage on gross margin or operating margin?

I think gross margin, there's definitely leverage. OpEx, we've been very disciplined in the way we invest. We have chosen targeted investments, which have paid off, and we'll continue to do that. And OpEx is an area that Lisa, myself, Mark and others in the company are very focused on investments to make sure they pay off over time. But there's definitely leverage on the gross margin side, given the fact that we've said before, we want to get to the greater than 50% gross margin from a longer-term standpoint. So that obviously gives us leverage from an operating margin standpoint.

So I guess another way to ask it, it sounds like there's upside. You purposely put the over symbol on your 50% gross margin target. Is the 26% to 27% likely a fixed number because the opportunities are large enough that you'll continue to invest? Or is there a scale benefit at some point that you will even get OpEx leverage beyond the 26%, 27%?

I feel we've got a lot of OpEx leverage. I mean we've said consistently revenue growth and OpEx growth lower than the revenue growth for the company from a percentage standpoint, and that gives us leverage from an operating margin standpoint.

Got it. Well, Devinder, we are exactly on time with the allotment for our fireside chat. So thank you so much for taking the time to join us this Friday. Great story at AMD. Congratulations on all the progress over the years, and we look forward to watching a continuation of that going forward. So thanks again, and have a great weekend.

Thank you, Ross. Have a good weekend yourself. Bye.