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

Okay, great. We'd like to get started. Good morning, everyone. Welcome to day 4 of our Communacopia & Technology Conference. My name is Toshiya Hari. I cover the semiconductor and semiconductor capital equipment space. Extremely excited, extremely honored to have the team from AMD with us this morning. You save the best for last, in terms of our semiconductor portion anyway, the conference continues. So we have Forrest Norrod and Ruth Cotter. Forrest Norrod is a Senior Vice President and General Manager of the Data Center Solutions Business Group, and he's responsible for managing all aspects of strategy, business management and engineering for AMD's data center products. Forrest has more than 30 years of technology industry experience across a number of engineering and business management roles at both the chip and system level. Prior to joining AMD in 2014, Forrest was Vice President and General Manager of the Server Business at Dell. He joined Dell as CTO of Client Products in August 2000 and helped multiple engineering and management roles during his tenure. Ruth, I'm sure most people know, is a Senior Vice President, and oversees Marketing, Human Resources and Investor Relations. With more than 25 years of experience, Ruth's leadership has been instrumental in helping the company scale during a period of hypergrowth as well as the successful completion of its acquisition of Xilinx, where she led the integration planning for the largest acquisition in semiconductor history. Ruth is also the recipient of the National Diversity Council's Top 50 Most Powerful Women in Tech award and has been honored as an Irish Technology Leadership Group Hall of Fame inductee and was named a Top 10 Champion of Global Diversity by Diversity Global Magazine. And I'm just a sell-side analyst. Well, thank you for coming. I really appreciate the time.

Thank you.

Forrest, I wanted to start off with you. As we sort of went through your bio, you used to be on the customer side. Curious, what was your perception of AMD back in 2013, 2014 as a customer and the perception of the data center business, in particular? And how has that evolved over time? Clearly, we've seen your numbers, you've done a lot better, but what was your perception as a customer? And how do customers perceive you today?

Yes, that's a great question. I would say that when I was on the other side of the fence, or the other side of the table, leading the server business, the strong reputation of AMD was a company that was really an innovation leader and had driven significant innovations in the data center space. If you think about the architecture of the modern server, candidly, it was really defined by AMD, in terms of the integration levels, the 64-bit instruction set. And AMD had done a great job in driving these disruptive innovations and architectures for many years. So they had that and we tremendously respected that. But the thing that I think that AMD struggled with there in the last cycle, in the mid-2000s and up to the early part of the past decade, was inconsistent execution. And they would come out with a great part, and then the next part would be not so good or would be late. And so I would say, innovative, disruptive, but inconsistent execution. And so when I came on board, I retired from Dell and then very shortly thereafter, I got a call from Lisa, when she became CEO. The thing that I really wanted to address and the thing that Mark Papermaster, our CTO, sort of my partner in crime there, really wanted to address was that execution set. Hold on, retain, double down on the innovation side, but really focus on being a predictable, dependable supplier of technology. Because if you are late or if you don't do what you say, the downstream effects to your customers can be pretty bad. I mean they're basing their resourcing plans, their product plans, their business plans on your technology and is extraordinarily difficult to manage when you get a component like the CPU that's 3 months, 6 months, 9 months late. It's incredibly disruptive. So we locked arms and said, "Look, we're going to fix that." Both Mark and I, I think, very much respect and value execution. And so we said, all right, first off, we're going to lock in -- we're going to be thoughtful. We're going to lock in a long-term road map that will get us back into this business and get us over time in a leadership position. We'll come up with focus points of differentiation, and we'll build it step by step. And we'll do it that way so that we're managing risk, and we can adequately resource the products to come out right on time. The second thing we'll do is we're going to tell people what we're doing. And so I said on stage at the first launch of our EPYC processors back in 2017 in Naples, very unusually at the time, I laid out, here's what we're going to do. Here's our full tour of Italy, Naples, Rome and Milan for the next 3 generations. And I did it because of that perception of AMD is somebody that had inconsistent execution. We wanted to lay down a marker and say, "Look, measure us. We're going to come out with these innovative products. We're telling you now what some of those innovations are going to be, and we're going to tell you when we're going to do it." And I think that doing that, and then more importantly, delivering to that was hugely important in getting credibility back as AMD as a reliable partner. I think that was incredibly helpful to our success in the past few years.

Got it. That makes a lot of sense, and that's great context. We'll definitely come back to some of the customer engagement dynamics and how you're doing from a market share perspective. I just wanted to pick your brain on how you're thinking about sort of chip design going forward. There's obviously demand for more specialized silicon in data centers, cloud data centers in particular. At AMD, how do you see CPU architecture evolving in response to this trend? What are the implications of this on the x86 ISA? And how does the co-ownership aspect of x86 impact your ability to meet these needs?

Yes. I think -- so first off, we definitely see that the market is broad enough now and different enough that the needs of a cloud company running a massive scaled-out service like Google Search or Amazon Web store, it is very different from the needs of an enterprise that may be running 24-core per server virtualization form. And so there are different optimization points to be able to deliver the most performing, most efficient, most effective solution. And so we've long thought that, at some point, we would have to begin broadening our portfolio. And our analysis was, look, we can hit the ball down the center of the fairway with essentially 1 product, 1 socket, 1 product, 1 architecture until we get to about 20% share. And then at that point, we would -- if we wanted to continue to grow, and that's our ambition, we would have to begin broadening the portfolio. And so that's what we've done. We've announced with the Zen 4 generation that we're going to start announcing -- start launching later this year with Genoa, a broadening of the portfolio. So Genoa is a 96-core part, great for sort of center of the fairway and great for many cloud computing and HPC applications. But then we've also created Bergamo, which is tightly focused on those cloud throughput applications, bumps the cores up to 128 cores, is really designed for those sort of workloads. And then we've also introduced -- we're going to be introducing Ciena, which is really targeted at a lower-cost platform still very capable for telco, lower-end enterprise SMB. But -- and we think that the innovations that we've -- the distinctions between those products really provide extremely well-optimized platforms for each of those segments. But there's a lot of commonality. So it's the same Zen 4 logical design. We do a different physical design of the part for each one of the different segments. It's a different socket. But all of the software, all of the code is the same. Software compatible across the board. It's not like we've got a different e-core that has a slightly different instruction set architecture. We don't do that. It's too hard for us. It's too hard for our customers. It's just -- and it's unnecessary. And so we've tried to drive as much commonality in the underlying base elements, and then drive innovation into the configuration of those elements to optimize for each one of those segments. With respect to the ISA, the ISA is going to continue to evolve. It's in both Intel and our best interest to keep the core of the ISA synchronized. And so I think you're going to see. There'll be slight differences in ISA between us, but generally, I think you should expect it's going to remain pretty consistent.

At your recent Analyst Day, you introduced a couple of new server chips in your road map, some optimized for certain workloads, certain end markets, including cloud, edge telco and HPC. Can you just discuss the varying workload characteristics across these segments? How you intend to meet these needs? And to the extent you've got sort of TAM estimates for these respective applications or workloads that would be super helpful?

Sure. And that sort of foreshadowed part of my answer in what I said a second ago. We see -- certainly, we see a growing TAM continuing to grow -- continuing to grow TAM for those cloud scale-out workloads, things, again, YouTube, social media services, search, you name it. And so Bergamo is the part for that. High core, 128 cores, still multi-threaded. So you can get 256 socket -- threads on a socket. But everything is sort of scaled. Those are -- you pack them any cores in, you're not going to be running them at 4 or 5 gigahertz. You are power constrained. And so that part is tightly defined to run at a certain frequency range to be highly power-efficient, which is super important for these guys, very predictable in its execution performance and just really optimize for that segment of the market. I think that over time, that's probably 20%, 25% of the market is right there, is right there in that application set. Enterprise and many other cloud workloads, particularly ones for third parties, people that whip out the credit card and run a virtual machine by the hour, single-threaded performance and ultimate bursting performance is more important. And so things like Genoa, bigger caches, more designed to be able to provide more bursting capacity to each one of the core, so they can burst up to a higher frequency when necessary. That's how we're optimizing there. And then what's a little bit more speculative for us is the telco and low-end enterprise market, but particularly the telco. We do think that a lot of the networking infrastructure is going to continue to move to virtualized infrastructure. And so we've produced a product there that sort of halfway in between those characteristics that I talked about before, very power efficient, but still a pretty high single-thread performance and a lower number of aggregate threads.

Forrest, you guys have gained significant share over the past couple of quarters. If you look at the June quarter, at least based on external third-party data in terms of revenue dollars, you gained about 600 basis points on a sequential basis. I think the momentum you have in cloud, hyperscale is very apparent. We see it. The enterprise side tends to be a little bit slower in its evolution. Can you talk about some of the stickiness there and how you intend to overcome that? And any catalysts sort of out in the horizon that could lead to an inflection in your momentum in enterprise?

Yes, I think we've started to see the inflection in enterprise already. I mean, if you think about, again, our strategy, our strategy was Naples was our proof point that we're back in the market and something that people could really test and begin kicking the tires. Rome was a part that, with its 64 cores, with its performance level, it was really optimized for cloud. And so we wanted to come out with a part that just killed it from a TCO perspective for cloud. And it was good for other workloads as well. And it was really designed to be extraordinarily competitive in cloud and leading in cloud versus Ice Lake, which is what we thought was going to be competing with. But it wasn't really designed to be dominant across enterprise workloads. That was the job for Milan. We wanted to get to Milan and say, okay, we'll build credibility with Naples, cloud leadership -- begin cloud leadership with Rome, and then Milan would be the part that could, no apologies, competitive to leadership performance on every workload, including enterprise. And that was designed to compete against Sapphire Rapids, which it does very well. Now we wound up with a different competitive environment than we thought. And so I think that Milan has seen a rapid uptick in -- we're starting to see our acceleration of our enterprise growth maybe -- and I think that's going to continue. The one other dynamic, though, enterprise is -- there's many factors that go into purchasing a server. In the cloud, it's pretty simple. These guys are buying parts to power their business, right? Their factory is their data center. And so driving every bit of TCO is critically important and almost nothing else matters. Once they know that you're a reliable partner, whoever has got the best part wins. In enterprise, it's a little bit more complicated. They've got a wider set of applications. They've got long-standing relationships. You've got to educate a large number of customers. And so our ability to cover those end customers to do that education and outreach is one factor that somewhat limits our growth, and that's somewhere we've been making investments. But you've also got the issue where the Intel sales guy has been buying me lunch every quarter for the last 20 years, factor also comes a little bit into play. And then in the cloud, these throughput applications dominate into our top of stack parts, a 64-core or seem to be 96- or 128-core parts dominate. And our lead is so great there that there's just no -- there's really no alternative. In enterprise, many workloads look like that, but many others really want a 16- or 24- or 32-core part. And there, we absolutely still lead against our competitor, but it's not a 2x lead. It's more like a -- if you look at Milan versus Ice Lake, same number of cores, it's like 30%. And so there's a strong pull, but it's not quite as fast. And so for all those reasons, we anticipated and we're seeing the enterprise ramp take a little bit slower, but we saw a clear inflection with Milan, and we believe it's accelerating strongly.

Right. So to your point, the performance advantage, the TCO advantage is very clear. And I think many of us would agree that your share momentum stays really strong. The big question that we collectively have is, what are you seeing in terms of the macro, right? You deal with a lot of customers, both on the cloud side and the enterprise side. Any sort of change in demand signals near term? To your earlier point, you've got 2 very strong products coming out over the next 6 to 9 months in Genoa and Bergamo. Are those 2 powerful enough to potentially offset a more difficult macro environment?

Yes, we definitely think we've got -- we're going to extend the leadership product portfolio and incredibly gratified with the reception that we're getting from our customers on Genoa and Bergamo. I mean, they see everything. They see what we've got. they see what our competitors have got. And based on what they're telling us, we think we're going to continue to have an extremely strong position in TCO. So we're very excited about that. We think this is going to be a strong transition for us. We think that Genoa and Bergamo are just stellar parts. And so I think we're going to continue to take strong share. So first off, in both cloud as well as enterprise. That said, I think the macro concerns are absolutely there. We don't see -- I don't see a slowdown right now in the cloud. I think that they are -- they do tend to be bursty and we see some of that burstiness, but nothing sort of out of the ordinary at this point and nothing that we didn't anticipate. So I think that from our perspective, we believe we're well poised to continue to grow there. The enterprise side, I would say the same. We still see strong end customer pull and demand. We are seeing some decisions slide out. I can't tell how much of that yet is people hedging because of macro concerns or there's significant product transitions coming up right now. Some of this is people, when they realized you're so close to the next product generation, well may slide their buying window over into the next one, and for both us as well as our competitor. And I can't fully tease that apart yet. But we're watching it very -- obviously, we're watching the macro very closely. But I think we're going to keep growing and taking share really independent of the environment.

Right. Your point about deals taking a little bit longer, I think Lisa had made on the earnings call. So this is more of a continuation of what you were saying a couple of months ago?

Yes, saying the same thing. I mean, again, we're seeing still extremely high win rates when we look at the enterprise side. So I'm very pleased with that. But some of the deals we've seen some -- take longer to close or were pushed out. But I'm not seeing a stop.

Right. Okay. Makes sense. Supply has been a big headache for not just you guys, but for the broader industry. And I think given how fast your business has been growing, I'm sure it's been hectic or nightmare even in terms of securing supply over the past 12, 18 months. What are some of the key pain points for your business specifically today? What kind of steps have you taken as a data center business as AMD? And when do you see supply potentially catching up to your demand profile?

Yes, I think for us, we have been completely -- in the data center business, we've been completely supply gated for quite some time. So we're -- our growth is utterly modulated by supply. And it's not wafer supply. We can get all the silicon we need. Our partnership with both TSMC and GLOBALFOUNDRIES has been great, and those guys have been very responsive. For us, it's really about substrates. So the underappreciated piece of fiber glass and metal that connect the chips, encapsulate the silicon dye and connect them to the motherboard. That's been the constraint. We've been investing very strongly over the last couple of years, in conjunction with our substrate partners, in broadening the aperture of our substrate partners to make sure that we secure supply, and that's come through. I mean, again, the business has been doubling quarter-on-quarter quite a bit or high double-digit/doubling for a long time. So that supply is coming through, and we expect it to continue to grow. I think -- it was a bit of a pain point for the rest of the business as well, but those substrates are a little bit simpler to manufacture. And that constraint, I'd say, is gone for the rest of the business by the end of the year. But for servers, I think we're still going to be somewhat modulated by that into next year. But again, modulated, it's growing in a pretty stiff growth.

Okay. That's helpful. Shifting gears a little bit, a couple of questions on your data center GPU business, where you've had good success in areas like supercomputing, cloud gaming, virtualization. I feel like yourself and Lisa have been talking a little bit more about the visibility you have into AI applications in the cloud, call it, over the next couple of years. Can you speak to the customer engagements that you're having there and kind of the opportunity set as you think about that business on a multiyear basis?

Yes, you bet. I think in GPU, we again took a phased approach to attacking the market. And we thought that the easiest -- not the easiest, but the most accessible portion of the market was going to be in the exascale, somewhat non-intuitively at the very highest end. The exascale systems and high-end HPC systems were somewhat counterintuitively, the software lift to have the systems optimized and the codes converted was more tractable than the broader AI market. And so we focused there first to begin building up the silicon port road map, and then also begin investing in the broader software road map. So we've been working on our ROCm software, a Radeon open compute, for about the last 4.5 years. We've got now with ROCm 5.0. We've got a production release of the AI software with production already for both PyTorch as well as TensorFlow. And now AMD and NVIDIA are the only 2 companies that have production branches on those core trees. And so I think we're getting there. Our competitor, NVIDIA, has done an incredible job in building out an incredibly rich software ecosystem around AI that really spans the gamut. Trying to duplicate that as our entry to the market would be a fool's game. So we chose to focus on quite candidly where the bulk of the market is with the mega data center guys. Because that's where most of the consumption of the AI GPUs is, and that's where the breadth of the software support is narrower and they will co-invest with us to get that software support for their environment, for their applications. And so we've been engaged. We've already publicly disclosed. Microsoft has talked about the MI250 running production training applications in the Azure environment. We're working with some others as well very closely for multiple years. And we're really looking forward to MI300 as the first part where we think it's a leadership AI part and the leadership HPC part, and that's coming out next year. And so all of our focus and effort is really directed towards the partnerships to make sure that all lines up.

The export restrictions on high-end data center GPUs that sort of came out in the news a couple of weeks ago surprised many of us. Your competitor in their 8-K spoke to a specific number in terms of how big the impact could be to their business. I'm guessing the potential impact to your business is minimal, but if you can kind of give some context around that, that would be helpful.

Yes, it's certainly we see a very minor impact to the business. Part of that is quite candidly -- so we didn't know this letter was coming either. I think it took us both by a surprise, but you could see the trends. And so we didn't make GPU in China focus of the business. And so I think it's a minimal impact to us.

But it captures our MI250 product force and does not capture our MI100 or 200 for more that we see right now.

That's right. That's right.

And we do not see our product development as it relates to future products being impacted currently from all that we know as well.

That's right. Well said.

Got it. And then maybe we can bring Ruth into the conversation here a little bit with the Xilinx acquisition, which is your baby. It's been, I guess, 7 years since your competitor acquired Altera. And I think there's an industry consensus that they've done little in terms of truly integrating those 2 assets and coming out with something powerful. Given the dollar opportunity set that you presented at your Analyst Day, obviously, you're thinking about this partnership differently. Can you talk a little bit about the opportunities there?

Yes, we're obviously very excited to have the Xilinx family part of AMD, having closed the transaction in February of this year. They have incredible IP as we think about our opportunity to really diversify our revenue, which is part of the strategy that we've been leaning into, not just as it relates to the data center, but the depth and breadth of other markets that they reach into spanning everything from test and measurement through communications all the way through to aerospace and defense. I'm also bringing FPGAs and SmartNICs back into our portfolio, along with the complement of the Pensando transaction as we bring DPUs into our portfolio, really providing that extensive portfolio of offerings is exciting. We've shared that there is north of $10 billion of revenue synergy opportunities. The teams are working closely and carefully together to try to figure out where and which ones we really want to go after. And as we think about that IP portfolio where we're best positioned, we've shared pervasive AI is one of those strategies and focus areas, and Victor is taking the lead there as we think about how we can spread that out across the entire AMD portfolio over time. The integration has been going very well, and the teams are very synergistic in how our engineering organizations came together under Mark and Vance's leadership. And now we're just looking at how we harmonize our hardware and software and road maps as we think about where can we place the AI engines across the entire product portfolio. So all going well. Obviously, significantly large transaction. We didn't lean into cost synergies too much, but we were happy to share that we're actually exceeding our original expectations and now tracking to about $400 million of cost synergies, which you'd naturally get through sort of duplicative public company engagements as well as the harmonization of some of the teams. And then culturally, very, very aligned as we think about, very customer-facing, very customer engaging, long, sticky relationships in terms of what would have been the embedded businesses, design win cycles at their extreme or 20-year design wins. And as we think about tracking and learning from the fostering of that length and breadth of customer relationships, there's other things that AMD can also apply to our business. That's very interesting. So far so good, so much to do. It takes a lot of time to bring road maps together, but we feel well positioned and on track.

Staying with you, Ruth, we are getting a bunch of questions on the PC market. We've had quite a few companies attend our conference this week, and it's pretty clear, you've got headwinds there. I think the thing with AMD is that you've been fairly conservative from the get-go in terms of how you've been thinking about the market. I think your most recent assumption for the market this year is down mid-teens, as Lisa put it on the earnings call. Since you talked about the market a month ago, 1.5 months ago, how have things gone? How are you seeing the market as of today?

Yes, I think we continue to see mixed data points in the market. Those are widely publicized. And I would say the PC market continues to track lower than expected. Typically, you'd see the second half of the year higher. But as it relates to the PC market, I think it's lower than expected, it's kind of the current view. However, we sort of feel more broadly well positioned as it relates to the other growth pieces of our business, which we had sort of leaned into as it relates to data center and embedded. But there's a lot of puts and takes, and the PC market is pretty messy. So we're working through all that, staying highly engaged with our customers as we continue to just work through all those data points and the messiness in the market right now.

Yes. I guess a similar question for your consumer GPU business. Again, it's a tricky market where you've got a lot of moving parts. And I think your nearest competitor has had their own issues as well. But how do you see that market evolving into the second half and potentially 2023? I know it's really hard. You need a crystal ball for that, but...

Yes, if we all had it, we might not be here together right now. The -- we actually -- just to take a slight step back, we adjusted our financial segments and that we reported last quarter, and we created a gaming segment. So there's 2 pieces to that. There's our semi-custom business, game console business with Sony and Microsoft, great partners. And then there's sort of the discrete consumer gaming piece. The game console cycle is very strong, very pleased with how that's going and continue with the momentum there. On the discrete graphics side, to your point, we headed into 2022 coming off of very strong 2021. I think as we were in the second quarter, we did see the market soften and softened quite a bit, which again has been sort of widely discussed, largely driven from what we could see in terms of supply versus demand and how that was playing out. Consumer spending under the macro backdrop, all impacting factors. We have said for the second half of the year, in particular, in the third quarter, we expect that portion of the market to continue to be weak as sort of supply and demand and some inventory in the channel continues to be worked through. The fourth quarter should be better for us. We have new product launch before the end of the year, which we're pretty excited about. And we think that will be helpful. But certainly, the back half of the year is having to work through those sort of machinations and again, that kind of messiness in the consumer graphics market. Product launches will help, and then we'll just have to see how we head into 2023 from there.

Okay. I certainly don't want to put any words in your mouth. But since the beginning of the year, it feels like the data center business, the game console business have trended higher. And the PC market, at least, has been weaker and consumer GPUs, again, the outlook has been weaker. So those dynamics kind of continue into Q4 and potentially early 2023. Is that the right way to think about the puts and takes?

Yes, there are a lot of puts and takes. I think we continue to work through those with some of the data points in the market continue to shift. But you're right, as you point to data center as well as the embedded business, which Victor is leading, also very strong. And we're delighted to have that in the portfolio. Game consoles held in there very nicely. And then it's more the consumer-exposed portions of the business, that there's a lot of messiness in and where we're just working through all that right now.

Okay. Great. We have about 5 minutes left. I wanted to pause here and see if you had any questions from the audience. If you do, please raise your hand, and we'll get a mic to you.

Too early in the morning for the...

Yes, it's day 4. Okay, great. I guess, let's talk about process node leadership. You recently announced some products that will leverage 5-nanometer at TSMC. At this point, just given the growth that you've gone through and you're showing, you're a very large customer, you're a main foundry supplier. I guess the question that we often get from investors is why not accelerate your transitions from a process node standpoint and be at the bleeding edge? Is that something that you think about internally? And if so, what kind of advantages would that give you?

Well, I mean, we think about it, sure. I mean, it's a very -- we've got a very close partnership with TSMC to make sure that each new process node is well tuned for our needs. And the tuning for the high-performance segment is a little bit different from what you might need for, say it, a cell phone. I think we're pretty comfortable with where we've been. We were fairly early in 7-nanometer. We're not quite on the bleeding edge on 5-nanometer. Part of that, quite candidly, was dictated by the timing of the markets. In the end, we want to have this regular predictable cadence of products in the market at the right time. And so I would say, on the server side, it's -- there's a cadence of 18 to 21 months that we've got to hit. And that's what drives us. Everything else backs up from there. But we're incredibly pleased with 5-nanometer. We think it's going to drive absolute leadership performance and performance efficiency, power performance, so that node is just fantastic. And I think you'll see us meet the market with the process node that we think we need at each given step at the time. But it's not -- we're not driven by the process node, we're driven by the needs and timing of the market.

And I think I would add to that. Obviously, the transition to every leading-edge node is incredibly expensive. So making sure that you're getting that right ROI at each node while meeting customer needs. But it's just beyond the node, right? I mean, we're very focused on our architectural advantages spanning both hardware and software and then adopting a lot of the new technologies and IPs, whether that's AI engines all the way through to advanced packaging, which you're seeing increasingly more, whether that's through stacking, whether that's through our chiplet strategy and the sophistication that we continue to do there, in particular with the chiplet to have a multi-node strategy. Or as you look at our embedded business, that spans 10 different technology nodes as we sort of think about that business. But really, our primary focus and competitive differentiation is how we're putting the building blocks that you hear Mark Papermaster so regularly talk about together. And then our process node strategy is one element of that. But we start primarily with the architecture and then build out from there in terms of servicing customer needs.

All right. I guess in the last couple of minutes, I wanted to give you, Forrest and Ruth, the opportunity to speak to anything that we may have missed or through your conversations with investors, you're a well-covered company, obviously. But anything that, as a group, we collectively missed or overlook or underappreciate about your markets or technology, your opportunity set long term?

Maybe I'll start, Forrest.

Okay. Sure.

Look, I think for us, it's the long-term strategy. AMD has not only had such a great trajectory -- growth trajectory and journey, it's about where we go next. And I think the revenue diversification journey that we're on is something that's tremendously exciting as we think about the next 5 years and how we take our portfolio from some of the more consumer-centric and exposed businesses that create this choppiness depending on the market environment, more into sustainable data center and embedded opportunities that can be deep and sticky as we think about relationships that we've touched on in the questions today and driving AMD's revenue to be greater than 50% over the long term with this data center and embedded portion of the business combination is very exciting and takes AMD into a new gear that you heard us talk about a little bit at our Financial Analyst Day, but that's very much a focus for Forrest and the leadership team as we think about where we're driving AMD to be over the long term.

Forrest?

Ruth did a great job. I think one of the exciting things -- for the data center guy, one of the exciting things is our ambition is to have constitute alongside of the Victor's embedded business, greater than 50% of the revenue of the company. I believe we've got the product road map, we've got the track record of execution that we're not going to let off, and we've got the customer relationships to absolutely get there. And that's an exciting company -- for a data center guy, that's a very exciting company.

Awesome. Thank you so much. Really appreciate the time.

Thank you.

Thank you.

And thanks for having us. Appreciate it.