Applied Materials, Inc. (AMAT) Earnings Call Transcript & Summary

Great. Good morning, everybody. Usually, when I have done the first session, they gave me a script of things I was supposed to say, and I don't have that script. So welcome to the Morgan Stanley TMT Conference, and I'm sure we'll have something smarter to say there. Just quickly on the research disclosure side. For important disclosures, please see the Morgan Stanley research disclosure website at www.morganstanley.com/research disclosures. If you have any questions, please reach out to your Morgan Stanley sales rep. So with that out of the way, I'm very happy to have kicking us off, Brice Hill, the CFO of Applied Materials.

So you've had a great year. Maybe you could talk just a little bit about the outperformance that you guys have seen. The strategy that led to that, and then we'll get into the details.

Sure. Joe, thank you for inviting us. Great to here, San Francisco. For us, it's been a great year. And what we described to our investors was we spent $3 billion approximately in R&D, and we point those toward inflections that the company sees, technology inflections that the company sees on customer road maps coming up in future years. This year, we had some markets that were very strong this past 4, 5 quarters. The ICAPS market, which we use ICAPS, the acronym to describe the more mature technologies, IoT, communications, auto, power, sensors, those types of devices. That market was very strong for us, and it was so strong over the past years that it outweighed weakness we saw in the NAND market, weakness we saw in the leading-edge market, et cetera. ICAPS has been an area where the company started a group that focused 5 years ago on developing new products for that market. And that's an example of an inflection that is manifesting in today's market. Another area has been HBM. We talk about high bandwidth memory and how that's growing in the marketplace. The company has greater than 50% share in HBM. And the DRAM market in general was also strong in the past year, and that allowed us, again, to sort of offset weakness that we saw in leading edge and in the NAND business. So Joe, it's really a function of the company investing its R&D in areas of inflection where the technology in the market is growing. And in the past year's case, that led us offset some weakness in NAND and leading logic.

That's great. And you've characterized this year, WFE, you sort of talked about strength in DRAM, NAND maybe a little better. And then this kind of handoff that you've talked about from ICAPS to the cutting edge. How do you think that positions applied relative to the WFE market and your ability to outperform WFE with that context?

So going through that and thinking about the shape of the year, even in this quarter outlook that we have for the quarter that we're in right now, it's still a strong ICAPS. It's still a strong DRAM market. We benefited from being allowed to ship some technologies to China DRAM customers that weren't allowed last year. So we've had 3 quarters, including this quarter, of elevated DRAM shipments. I described that as approximately $500 million incremental per quarter over what we would normally expect shipping to those DRAM customers. That strength and the ICAP strength, primarily in China, although we saw strength in all the markets last year, that strength has offset the weakness we see in leading edge. We are very clear about that and then also NAND. The handoff that you're mentioning is, we have articulated that there should be some digestion in ICAPS, especially in China. There's been a significant ramp. ICAPS grew, we said more than 40% 2 years ago -- or approximately 40% 2 years ago or even faster last year. That's a lot of capacity that has been put in place. A good portion of that was put in place in China, and we expect that market to remain strong in the coming years, but we think there'll be some digestion and it won't grow as fast. It will be slower. The handoff that you're alluding to, if ICAPS is slower and China is slower in the second half of this year, then what can replace it from a customer perspective? And that's really leading edge from our perspective, and we also expect DRAM and HBM to remain strong. The leading edge should be driven by the ramp of gate-all-around technology with some of our customers. And then also just new nodes to serve the newest designs on leading edge.

Great. And so I guess the natural question is the durability of all this into 2025. You had such a -- I feel like other people had a downturn in 2023 that you didn't really have any. I mean, but do you feel like we're sort of at a foundational stage where we can continue to grow from '24 on? Just -- obviously, it's early to call, but any thoughts you can give us on that?

Sure. Most of the people who ask us the durability question asked us in terms of ICAPS. And I'll look back to a couple of quarters ago. The way I think about it is, we saw the ICAPS process nodes in general, worldwide running at healthy utilizations. When we think of the $1 trillion device forecast for 2030, semiconductor devices for 2030, that requires a mid- to high single-digit growth rate for all device types. I think the analysts who provide forecast, data will be across all device types. If your mature node technologies are running at good utilizations, high 80% utilizations and the devices are going to grow at mid- to high single digits over the next 5 to 7 years, you have to add capacity in order to serve that device growth. And so when we look at that, there is a little bit of lower utilization right now. We think that's some of the digestion, but when we look at the next 7 years, next -- to 2030, we expect capacity has to be added globally in order to serve that. I'll stop there. We could go into -- we can break it. We could think about memory and we could also think about leading logic, but I think the largest body of wafer starts globally is in these mature node technologies, and that's where a lot of the equipment sales are going to come from.

Can you give us a qualitative sense for where ICAP is now as a percentage of logic? And then it's very clear that there's a structurally higher level of spending required in these mature nodes if only because we're not waterfalling the cutting-edge nodes down the way we used to, they're converting advanced nodes, and so we don't get that free capacity. So that's higher. But when you mentioned utilization coming down, it seems like in the higher technology areas, you need to spend a lot even in downturns because you need to keep progressing on technology. If we have this utilization hiccup, could it be a more severe decline? And, I mean, obviously, that's somewhat offset by geopolitics. But just how do you think about that dynamic?

Okay. A couple of questions in there. So the first part, about 2 years ago, we thought -- or we said that the leading node technologies and the ICAPS technologies, the mature node technologies were about 50-50. And we've had significant growth in ICAPS since then. So ICAPS is actually our largest market at this point. We don't get more specific about that. But it's the largest market. As we think about the business that we have, those mature node technologies are still evolving, just like in the leading edge. So we're developing technologies you might think about power, as an example, silicon carbide, different technologies that we're deploying in the mature nodes. It continues to grow its capabilities, grow its TAM and search for new solutions, new material solutions, not unlike leading edge. Leading edge is much stronger in terms of the need because you're pushing the edge of the envelope and the process steps are getting a lot more complex. And you're right, that the leading edge we view as just a consistent flywheel of demand requirements, new technologies, the company talks about the new transistor, which will be gated all around. There's another transistor after that called CFE technology. There'll be another technology called backside power delivery on the road map in the next few years. That leading edge is also a flywheel that requires our R&D, requires close collaboration with customers so that we can pick projects that will land on those customers' road maps and be deployed with end designs. It's a very close Nick community that's working on this road map so that it's effective and it makes our financial returns very efficient because that $3 billion R&D that you think of deployed in developing new technology that's in collaboration with your customers. So you are -- you have a higher probability maybe than if you're doing sort of blind R&D, looking for a market for a new solution.

Great. So maybe I'll focus a little bit more on ICAP in China than cutting edge. On this ICAP piece, you guys have obviously been very strong there. Is there a definitional thing where you guys -- I mean because you count everything, the top 3 nodes is more cutting edge. Does the ICAP denomination shift over time, first? And are you able to -- as we do advance a little bit more as we are probably able to maintain that very high share that you have?

Okay. We were shifting the definition in years past, and we -- I think we've stopped shifting the definition. So the cutoff point is between 10-nanometer and 7-nanometer. Our perspective is 7-nanometer requires the EUV that's more leading edge from that perspective. And then we'll -- at least for now, we'll keep that dividing line there. So 7-nanometer and newer will be leading-edge technologies, 10-nanometer without EUV and more mature and larger from that perspective will be the ICAPS technologies. Sorry, what was the second piece?

Yes. And it does apply maintain share as that spending shifts to more advanced parts of the ICAP market.

Well, we do think that, that goes back to what's the engine of the company and what should investors understand. $3 billion of R&D, working with the customers to identify the future inflections and then investing for the solutions that enable those technology inflections. And so our perspective is for the ones that we've named, gate all around backside power, $3 billion of R&D, working with the customers to identify the future inflections and then investing for the solutions that enable those technology inflections. And so our perspective is for the ones that we've named, gate all around backside power, CFED, HBM memory, DRAM in general, we've gained 10 points of share in DRAM. Our perspective is we've been successful at identifying some important inflections and that we'll gain share in those inflections. It's approximately 50% in each of the ones that I named except for DRAM. And that is what an investor is looking for. Are you effective at calling those inflections? Is there R&D or PhDs or engineers? Are they effective at developing the solutions? And then do you have the customer relationships to put them in the field in time to ramp new devices for end customers?

Great. And with regards to China, I mean, you've talked about the DRAM element that's got a little bit the timing all got pushed into the second half. But when you think about the broader logic investment in China, numbers are pretty high. I mean you're up over 45%, I think, of your revenue. If you think in terms of China achieving self-sufficiency that they would have to maintain those levels of investments for quite a while, but they also -- there's a lack of building, fab is only part of it. They need to have product, they need to fill those factories. So how do you think about those dynamics? Do you look at this as something that endures beyond this kind of export control anxiety environment? Do you see something that -- where there's a multiyear phase of very strong investment in China?

We do think that the investment will continue to be strong in China. That's our internal information. We have, of course, a lot of accounts in China. We have a forecast from all those accounts. When you think about a factory of a particular customer in China, we have our perspective, which shows each factory, maybe they've put 5,000 wafer starts per month in place. And then they have a plan for the next 4 years to add capacity. And so we have a perspective of who are all these customers, what have -- what factory and capacity do they have in place today? What are their plans in the next 4 years? And that gives us a perspective of what the aggregate demand is. We test that, Joe, to make sure that it's rational, at least from our perspective. Does the overall equipment market makes sense relative to the overall semiconductor market, that's the intensity equation that people have talked about in the past. And then you alluded to, we also look at how much capacity has China put in place relative to local consumption. We think that still makes sense. If you're looking for any softness in the equation, I think what we also think is that the China capacity is a large number of new customers, they're coming up the learning curve. Their yields are lower than world-class foundry yields. So as that capacity is put in place and those yields mature, we do think there's some digestion and that's actually what explains why the market we've described over the near term will be stable and not growing at that 7.5% to 8% because we're going to grow into that capacity, both yield and utilization wise.

But their investment wasn't predicated on super high utilization to begin with, right, at least, I mean, from the public SMIC, [indiscernible]. We had the numbers we know those utilizations weren't that high and they were investing anyway in the same way that TI is in the U.S. So it seems like even if utilization falters a little bit, the underlying commitment to spend is filled out.

We think so. We think it's a national strategy. It's a national priority. We think there's subsidies in place. We know there's government grants that are in place in China. And so our perspective is that the road maps that we see from the customers, they'll follow through on those road maps. Now there's a large number of customers. We do think there'll be some consolidation. If you have multiple different customers making image sensors, for example. Some will be more successful. We think there'll be some consolidation and some rationalization of the market over time. But we do expect that, that will be a continued growth in capacity in China.

Okay. Great. So you see the decline from the mid-40s to the mid-30s that you've guided to as an exposure that's mostly the DRAM side and then a little bit of consolidation.

That's right. So for Applied, our business was 45% revenues to China in our Q1. And we highlighted that as the year goes on, that should return to a more normal level, which is about 30% for us on historical average that will be reasonably linear through the course of the year. And this Q2, we have another quarter of elevated DRAM shipments. After this quarter, we expect that to slow down in the second half.

And on the China DRAM piece, can you talk a little bit about that because that was much stronger than I expected it that it would be? Now you have a send customer spending quite a bit there. The sustainability of that and the viability of building a DRAM business with the limitations that they have in terms of what nodes they can address.

Yes. You might know better than me. I know logic a little bit better than the memory side, but using logic as an analogy, and I don't know if this will be fair for memory. But in logic, you can find different ways to extend the performance of a node. You can improve your design for several cycles and get more performance from it. So I don't view it as -- at least I personally don't view it as a brick wall for the China memory customers. They'll innovate, and they'll find ways to use that technology and extend the performance of that technology, that would be my perspective. It may be different than what the other DRAM customers are doing as they move their nodes forward.

Okay. Great. So speaking of DRAM, you guys have had a great progression of DRAM revenue. You've talked about gaining 10 points of share. Can you talk about the drivers of that share? And how you see that playing out going forward?

Yes. The big drivers, so applied, I think we've said about 10 years ago, we were a 13% market share in DRAM. And today, we're closer to 23%. And that's really because as the technology has become more and more complex, it's deployed some of the same techniques that are used in leading logic. So advanced and much faster I/O capabilities, what's required for capacitor scaling, that requires some of the same techniques that we're using on leading logic for DRAM and some of the same types of equipment and equipment modules including our e-beam capabilities to give high resolution and fast-speed examination of different structures in the devices. So really, it's leveraging some of the same developments that have been used on the logic side.

Okay. Great. And maybe if you could talk about NAND a little bit. The NAND business spending is down a lot. Your revenues are down a lot and the economics are improving. They're still not probably where they need to be to see a big reacceleration. But can you just talk to the NAND business? Do you still see a large opportunity there over time?

We do see an opportunity in NAND. And it's an important market. It's hard to explain why it's been as low as it has been, at least personally, I don't see an architectural reason why we would need less storage memory than we have in the past. It doesn't have the same sort of demand driver that we've seen just recently on DRAM, which is this HBM. On the DRAM side, you have a lot of excitement around artificial intelligence. And the companies that are building systems, high-performance systems are searching for ways to use advanced packaging to add a lot of different devices, CPUs, GPUs, accelerators stacks of high-bandwidth memory. That's been a big driver on the DRAM side. You don't have that same driver yet on the NAND side, but there's nothing that has changed the system architectures that means we need less storage, at least from my perspective. So we would expect the NAND market to normalize over time and new investments will be required in NAND to advance those technologies. And just one more thing I'd say, for both NAND and DRAM, at least my perspective is, every year, you have more big demand. You need to provide more bits in order to serve the memory demand that's growing. That growth has been delivered by new process technologies. So when we sell equipment to DRAM and NAND customers, it's really equipment to upgrade the nodes. And those nodal upgrades are providing more bit density and that bit density is what's giving you the memory bits to serve demand. And I say that just to say that I think it's been going on for several years that we haven't been adding a lot of wafer starts in either memory technology, it's really driving the road map so you advance the technology, get more bits on every square millimeter. That's really what's been driving both memories and what we expect will continue to drive it for us going forward. So it's -- both of those are still very important opportunities, just like leading-edge opportunities where there'll be a lot of innovation going forward.

Yes. I mean it looks like by contrast, you've talked about DRAM being driven by China and HBM, but it seems like there's a pretty healthy underlying DRAM spend as well.

There has been. And that's why when I highlight that the incremental over the last couple of quarters has been $500 million incremental. Underneath that incremental, if you subtract that from our DRAM, there's still a very healthy DRAM market, both normal China and in the rest of the world, of course.

Great. So maybe we could talk a little bit about AI and advanced packaging. I think you've talked about $1.5 billion of revenue this year, of which DRAM is going to be $500 million. Can you talk about those opportunities and the visibility that you have?

Yes. So the advanced packaging business, another important one for Applied. We put a number on it and said it was approximately $1.1 billion in '23. It should grow to approximately $1.5 billion in '24. And most of that growth, at least this year is going to be driven by the high-bandwidth memory application and solutions. So you've got on the order of $400 million of growth driven by HBM. For HBM, the way I think about it is -- or the way I've been -- people help me understand is there's approximately 700 steps in a normal DRAM process. And you add 15 to 20 additional steps to give a sense of scale in order to add the HBM, high-bandwidth memory, capabilities onto a DRAM line. For those additional steps applied is providing the majority over 50% of the equipment for the additional steps. And I won't walk through it. It's through silicon via, it's billing, it's plating, gap fill, different films, different technologies. If you think about micro bumps that connect those devices and the through silicon vias that allow the connection through the devices, Applied is providing a significant amount of the capability to produce those.

And I assume we've had this kind of rush to catch up to the demand for AI. Any change in that visibility going forward?

It's interesting for both HBM and AI in general. When we think of WFE, we would say that only about 6% or so of wafer starts are actually today directed towards AI in the WFE, but it's going to grow at a 30-plus percent CAGR. So pretty significant growth. And even when we drill into DRAM, DRAM has been underloaded, lower utilization, there's probably only about 5% of DRAM starts that are allocated towards HBM memories, but customers have said that that's growing at 50% to 60%. So today, it's not an overwhelming portion of the wafer starts driven by AI, but it's growing quickly. On the logic side, it will quickly eat up utilization and start pushing for new investment on leading edge because all these devices DPUs, CPUs, accelerators, they're all leading edge devices. And on the DRAM side, we think it will help drive much higher utilizations and more healthy marketplace. And then, of course, for Applied, it will be both the DRAM equipment, the leading-edge equipment and then the HBM equipment that goes along with DRAM.

Great. That 6% number is kind of interesting. So that's 6% of today where you are today.

Yes.

Okay. And that's including both wafers for memory and wafers for logic?

It's AI specific. No. It's just -- I would say that's a logic number. That's a logic number. Yes.

Great. So maybe if we talk about the arms race on Logic. You talked about gate all around, backside power. You've talked about Applied capturing 50% of kind of incremental money. Can you just walk us through the math there and talk about the opportunity.

So for gate all around. So a FinFET transistor, there's a certain amount of equipment that we sell for -- to make a FinFET transistor. As we switch to the gate all around transistor, it's a more complex it requires extra steps. The way we've described that is it's $1 billion incremental equipment requirement on top of what's already required for FinFET. So $1 billion of extra SAM, if you will, available market for Applied. And we should sell 50% or more of that capability to produce that. That's $1 billion per 100,000 wafer start per month capacity, just to orient that from a size perspective. And then it's much the same equipment types that are used for FinFET transistors. It's different epitaxial deposition, it's selective removal, it's materials modifications, and also e-beam technologies to help analyze those devices as they're being made.

Yes, it can be difficult for us to validate this because every company is selling us are going to pick up share. But certainly, some of these epitaxial steps particularly come up again and again as being the most critical. When you talk about -- I think last year, it was right after you had introduced the Sculpta tool. Maybe if you could talk a little bit about that and the opportunities that that provides.

So Sculpta, new tool for Applied, it's a materials modification tool. And if you were looking at a patterning diagram where you saw circles, ovals, lines that represent the patterns to building a transistor. What the Sculpta tool allows you to do is sort of elongate a circle or elongate an oval. Or if you have 2 lines, you can elongate those lines and have them be closer together. And so it actually -- where you don't have the imaging technology to get something closer together or change its shape. This tool helps you change the shape, which can improve quality, and it can improve the density of your design. What we have right now are leading-edge customers experimenting essentially with the tool and finding different ways to use it. And I sort of alluded, I think some customers are finding ways that those tools will improve or the Sculpta tool will improve their quality and others are actual steps in the process where they'll be able to use that to advance the density of the design. And I think Gary said in our earnings call, we expect that to be a $200 million tool sale for us this year. We said conservatively last year when we launched, it would be over $1 billion tool for us on the road map. And I think that was very conservative. We see a lot of experimentation for that equipment going forward and even with the customers today.

Okay. Great. So I have a couple of more questions. I'll open it up for the audience. Maybe you could talk a little bit about the Services business. You guys were the first company to really break this out and talk about it. You had very good numbers last year in the Services business despite some of the utilization headwinds. Can you talk about some of the opportunities there?

Well, we're very excited about the Services business, and I like it financially. I've tried to tie our dividend conceptually to our services business because it's a very stable business. The equipment business grow -- goes up and down with whatever the industry is doing, and we talked about that a lot. The Services business, every day we ship a new tool that grows the installed base for us to sell spares and to offer services. And so that's sort of a constant growing base that you can serve. I'd like to joke that we have more tools in place to potentially serve them McDonald's has franchises. So you have all that in the field. A lot of that business on the recurring side, more than 2/3 of that business is under contract, a subscription contract. Those are long-term contracts. They average over 2 years. There's a high renewal rate. So it has some -- a lot of constancy to it. We signaled last year that we are going to raise our dividend significantly. Our intent is to double it over the 4-year period. That's our intent. We have a Board meeting coming up where we'll decide this next dividend payment. And what I tried to articulate was that the services business, the consistency there, if you look at the profits from that business that those recurring revenues and that stable will kind of conceptually tied to our dividend, not tying it directly but just saying it will cover our dividend and that will provide the payments for that. And then the equipment business will use the buybacks to return and distribute profits to investors as that business goes going forward. And the Services business is very exciting. So the installed base is constantly growing. And then because the tools are more complex, the company is creating new service capabilities for those tools. Of course, some of the most interesting are in the AI space where we're able to use data that we can collect across our entire tool base to help customers ramp those tools quickly and ramp them to high yield. And you have applied offering capabilities that help the customers do that. So it's the installed base, it's the tools becoming more complex. And it's our ability to identify new valuable services to help the customers ramp. Those will grow that business, we think, in the low double digits for the foreseeable future. And that's been our forecast. And so it's just -- it's very exciting for the company and a source of stable growth for us.

And low 20% of sales, I should have mentioned upfront, I mean, and that's a number that should have some utilization, probably tailwinds in some of the more cutting-edge markets over time?

That's right. Great point. If you look at our most recent quarter, that business, I think, grew at 7.8% year-over-year. And this is with sort of globally low utilization levels. As utilization start to pick up, that will be a tailwind for our Services business. That's what we expect to see as we move forward. And we highlighted in the earnings call that we actually see our utilizations across the entire semiconductor ecosystem, ICAPS, NAND, DRAM, leading logic in this quarter that we're in right now. All of those utilizations are starting to improve.

Great. So I'll ask one more and then turn it to the audience. Can you talk about uses of cash? You mentioned a little bit how you think about dividend versus buyback. In the past, you've tried M&A. It's been tricky not for anything that Applied did wrong, but -- so how do you think about the uses of cash going forward?

Yes, I think it probably won't be surprising to anybody. But first, reinvest in the business. We have -- I talked about $3 billion R&D. We think we have a growing road map. We've got growing opportunities for us to invest in new technologies. We've highlighted that we're building the Epic Center here in Sunnyvale. And that is to increase collaboration with our customers. So that will be -- on the outside, it will look like a fab. It's fab technology, a leading-edge facility from that perspective. But on the inside, it's a customer collaboration vehicle. It's where we develop our latest technologies and we work with the customers in order to prove those for their road maps and for their utilizations and to develop new materials innovation. So that's a big investment for the company. But after we invest internally, and it's a high cash flow business, our business averages over 20% of revenue is free cash flow. And free cash flow for the business, I think, grew over 30% over the last 10 years. So it's a significant cash-producing business. We have limited M&A opportunities. So all of that cash -- all of those profits will be distributed to shareholders over time in a limited M&A environment and we'll use the dividends and buybacks to do that.

Okay. Great. Let me pause there and see if we have any questions from the audience?

Microphone coming.

Just on the China strength and sustainability. Do you see risk of any future restrictions? Or do you think that that's now been established in terms of the rolls of the road going forward?

Yes. Thank you. So most of the business for us in China are all of the business has either been the DRAM side or this mature logic capability. And we don't speculate on new rules, but we think that we've already been -- the business is already adjusted to the rules that are put in place. And it's the mature technologies that are driving the business for us in China. So we're not expecting significant changes from that perspective. And we would go back to what we highlighted that we think with the device growth and the road maps that we see from our customers that, that will be a significant market for Applied going forward.

Any other questions?

Can you comment on the evolving capabilities of Chinese semi-cap vendors in this environment?

Yes. I think that's good. So I think we're talking about local Chinese equipment vendors. They are growing share. The way that Applied thinks about it is the TAM is growing. So the way we think about it is we have to innovate at sort of the leading edge of the road map, whether it be ICAPS, DRAM, NAND or leading-edge logic. We think that TAM is growing fast enough that Applied will be able to grow share over the foreseeable future, even when or while some of the local Chinese vendors or other competitors are able to gain share in some of the lesser differentiated longer-term solution components. And that's what's happening. So we do think there are successful vendors in the market there. We think they gained share in some of the places where there's more well-known technologies. And our job is to continue to innovate gate all around or backside power or CFAT would be places where we're pushing the edge of the envelope and it's growing the number of steps and requirements for our business as we go forward.

You have to lose some share just mathematically because there are companies on the entity list that are spending money that you can't do business with.

Well...

We'll see -- something we can do about that.

We'll see. I think it's a matter of what grows faster, competitors share or new markets for us, at least in the current time frame, we see the new TAM, a new SAM that's being created by complexity is growing faster than -- obviously, we've been gaining share. So that's what's allowing us to gain share.

All right. Maybe we have one more.

You've mentioned GAA and CSAT technology. What about hybrid bonding and in particular, your relationship with BESI?

Yes. Thank you. So advanced packaging of all types huge focus for Applied. I kind of described the -- it's a little bit of an arms race going on for customers right now to find ways to build high-performance systems with these different chip types. Hybrid bonding will be a focus for the company. And we expect, in general, for the advanced packaging capabilities that Applied will have significant share in all of those new areas. And we're exploring other areas in packaging too. We have partnerships with customers -- or sorry, partners. BESI is a good example where we have a strong partnership to provide an optimized manufacturing flow. And there are other places where we have partnerships with other companies where we're developing new capabilities for packaging. And so I think we view it as an extension of Moore's Law, if you will. It's another place where there'll be a rapid advances in the capabilities to combine hundreds of chiplets and be able to do that in a high-yield, high-performance way. And it's an important part of our road map. And when we go back to advanced packaging, we talked about it will be a $1.5 billion business. Gary has highlighted in our earnings calls that it should double over several years. So that gives you a sense of the amount of growth that we expect to see in that business.

Great. Any other questions from the audience?

This is a high-level question. So it seems like there is this memory all that the industry is going through that memory bandwidth on these GPUs in this AI cycle are lagging when it comes to feed compared with the process it. So sitting in your vantage point, when you think about high bandwidth memory, over this cycle a few years from now, do you think that memory all can be bridged?

So I'm not a system architect, but I do see road maps that continue to advance both the I/O speeds and the memory capabilities themselves. So I would say that we are not at a memory wall at this point that our road map shows advances across both speed and memory capabilities. I think it is probably -- it's a wall from a system designer perspective, meaning in each generation, you only have so much capability, but I think that's increasing in speed and capability over the next 5 years.

Great. Brice, thank you very much.

Thank you, Joe. I appreciate it. Thank you.