Intel Corporation (INTC) Earnings Call Transcript & Summary

Let's get started. I'm Vivek Arya from BofA's semiconductor, semi-cap equipment research team. Really delighted to have our friend, John Pitzer, Corporate Vice President and Head of IR at Intel join us this afternoon. I'll go through a few of my questions. But if you have anything you'd like to bring up, please feel free to raise your hand, and I'll be sure to get you in. So warm welcome, John. Before we start the Q&A, I'll have to read a very exciting statement.

Appreciated.

So before we begin, please note that today's discussion may contain forward-looking statements that are subject to various risks and uncertainties and may reference non-GAAP financial measures. Please refer to Intel's most recent earnings release and annual report on Form 10-K and other filings with the SEC for more information on the risk factors that could cause actual results to differ materially. Additional information on non-GAAP financial measures, including reconciliations where appropriate to the corresponding GAAP financial measures.

Perfect.

Do you remember this by now?

I haven't memorized it yet. I probably should.

Excellent. Great. Thank you so much, John.

Thank you, Vivek, for hosting. I appreciate everyone joining us this afternoon. Thanks a lot.

Before we get things started, I know post close Intel made an announcement about Apollo's investment in your Ireland fab. and maybe you could just give an update on that. And importantly, what does that imply for this year's capital expenditure?

Yes, I appreciate the opportunity to talk about it. So yes, after the close today, we did announce what we're calling our SCIP2. This is our second implementation of our SCIP program. Remember, SCIP stands for semiconductor co-investment program. It's part of our smart capital strategy of trying to be smart about sources of capital, tapping financial partners to maintain the strength of our balance sheet, yet still be able to execute and accelerate on the strategy. As you pointed out, the SCIP2 partnership is with Apollo. It creates a joint venture between us and them and their affiliates, where we have about 51% stake, they have a 49% stake. And that JV has certain rights to Fab 34 in Ireland. And Intel corporate will buy wafers from the JV at a cost plus margin structure as a way to give finances into the JV. It's structurally somewhat similar to SCIP1, but it is a little bit different. Remember, SCIP1 was a new capacity build out in Arizona. And the way that worked is for every dollar of CapEx we spent in Arizona as we're building out that facility, we contributed $0.51, Brookville contributed $0.49. This is a little bit different in the sense that Ireland is mostly built out already. And so we'll get the $11 billion upfront, which is what -- and that allows us to kind of take capital out that we've already deployed in Ireland to date. We've spent about $18.4 billion in Ireland. It gives us capital to finish Ireland, but also to use that capital elsewhere in our global footprint. And what we said about the structure is that over the next couple of years, there's not a significant impact to the P&L. There's no change to the capital spending plans. Our capital spending plans already contemplated this. And then as the fab ramps into full capacity, you'll start to see more of an economic benefit flow to the JV. From an accounting standpoint, the way to think about this, given that we have 51% ownership, we will fully consolidate this through all the way to the net income line. And then as you know, we've got that noncontrolling interest line after net income, and that's how we reflect both the Brookfield agreement, now the Apollo agreement. Quite frankly, it also reflects when we sell down Mobileye, when we sell down IMS, everything that we deconsolidated that NCI line before we calculate EPS. I think what's important to note is this is likely the last SCIP deal that we are contemplating. And so we did a SCIP1, we did a SCIP2. It really helps us protect the balance sheet and prosecute the strategy during a period of time where we know we're playing catch-up and our P&L and cash flow is under earning. If you go out over the next several years, as we talked about in the April event, we think we're going to see meaningful improvement to both the P&L and the cash flow statement over time. And so as you think about incremental sources of capital from here, I'd make the argument that one of the largest incremental will be cash flow from operations over time.

Got it. Makes sense. Second thing, John, on Computex. So Pat had the keynote yesterday. If you could maybe give us a summary of the main announcements we heard about Lunar Lake. We heard about Gaudi. We even heard about Gaudi's pricing, which I thought was interesting. So maybe if you could just get us up to speed on it.

Yes. But Pat was in Taipei, his keynote was APM local time last night. There were probably 4 product announcements that sort of came out. Clearly, given that it's a PC show, I'll hit first on Lunar Lake and some of the comments we made there. We're very excited about the AI PC. We think our silicon road map is extremely strong even before you layer on top the ecosystems we've been able to build, both in consumer and in enterprise PCs, which we think is going to advantage us as you see AI PC pick up momentum. With Lunar Lake, as we talked about at the Build conference a couple of weeks ago, it will be a Q3 launch. We talked about it at the MPU level being a 48 TOPS processor. Actually, at the SoC level, it's greater than 100 TOPS, if you include the GPU and other things. And so one of the things that I think is important to note is we don't think -- we're going to have a performance superiority with our silicon road map. We also think we're not going to have a significant sort of battery life deficiency relative to the ARM camp and peers. And I think that's important. We tend to shy away about -- talking about battery life benchmarks in general because there's so much more that goes into battery life of the system than just the CPU that it can be a very misleading benchmark. But one of the things that Pat said about Lunar Lake, it is 40% more power efficient than Meteor Lake. And so we're pretty excited, and we think our road map only gets stronger after Lunar Lake. I think we were the first to talk about the AI PC. I think we're doing a very good job helping to define that category, and we will be the biggest beneficiary as AI PC penetration increases. On top of the client side, we also had some announcements within data center. We talked about our Xeon 6 generation E4, which is Sierra Forest. As we talked about on the earnings call, we released that in Q2. The peak core version of that, Granite Rapids, will be released in Q3. We're excited about both. And I'm sure we'll get into some questions about the server market, so I'll probably leave my comments and expound later on in the conversation. And then lastly, as you pointed out, we talked about Gaudi, some of the pricing dynamics, some of the TCO dynamics. Gaudi 3 is now in the market. As we said on the last earnings call, we're set to have greater than $500 million of revenue from Gaudi this year. So we're excited that we've moved from pipeline to actually now talking about backlog and revenue. It's a good start, but we're not satisfied. It's obviously a much larger market than that, and we want to gain more traction, but we think Gaudi is starting to see some good momentum in the marketplace.

Got it. So let's talk about the PC market first. So during COVID, we got too high of like 340-plus million units. Now I think they are like 260 million, 270 million. What does AI PC mean? Does it mean a positive refresh cycle? Does it mean an ASP upside? When do we start to see it become more tangible?

Yes. That's a really good question. When we first started talking about the AI PC, and I think we were the first to actually brand the AI PC, pat likened it to the Centrino-like moment. It's, I think, the exact words that he said. And we like that comparison for really 2 reasons. One, when you think about Centrino and Wi-Fi, it took the ecosystem 12 to 18 months to figure out the functionality around Wi-Fi and the incremental applications it could drive before it became a tangible driver. And so while we're very positive about the AI PC over a multiyear period, we do think it's going to take some time for the ecosystem to figure out what this thing can actually do. And I think you were at CES earlier in the year. And there was a lot of buzz around AI PCs going into CES. And I think everyone came away going well. The ecosystem hasn't quite figured out what this thing can do yet. Every quarter that goes by, I think you'll see more tangible proof points. I think the Build conference from Microsoft earlier this month -- or last month was really important. I think every quarter that goes by, we're working with more ISVs to bring AI-infused applications onto the PC. But just like Centrino, we think we'll take a little bit of time. And just like Centrino, when it did kick in, it was much more of an ASP uplift than it was necessarily a unit uplift. And we see a similar dynamic with the AI PC. Now as you pointed out, consumption last year was about 270 million units for the market. We're pegged this year to like 276 million to 278 million. We still think, longer term, this PC market is roughly a 300 million plus or minus unit market. I think AI helps on that front, but it's really more of an ASP uplift. And if you're thinking about a more concrete tangible unit driver for this year, it's really as we get into the back half of the year, whether or not enterprises start to think about refreshing their installed base, there's 2 reasons to do that. One, we're 4 years-plus past the COVID peak. The other 1 that's really important is Microsoft will end of life their service level agreement for Windows 10 in the fall of next year. And historically, if you go back and look at how the enterprise reacts to that, about 12 months in advance of that, you start to see some sort of enterprise refresh.

Got it. Sitting here in June, would you already have visibility into that conceptual refresh? Or is it too early to...

Vis-a-vis conversations with customers and OEM partners, absolutely. And you've seen some of our OEM partners make some public comments on recent earnings reports that would suggest they're getting more optimistic about the enterprise refresh in the back half of the year. Whether we have tangible orders or not, that will be a little bit closer to the actual event.

And does it imply better than seasonal? Is that what a refresh implies? Or is it...

That is our expectation. Now remember, we only guide revenue one quarter ahead. But we have kind of given some breadcrumbs for the full year that would suggest at the corporate level that we would have a better than seasonal second half. And there's many sort of bottoms-up drivers for that. But one of them is off of the low end of seasonal first half per PCs. We do think we'll see a little bit better than the second half trends in PCs because of the enterprise refresh.

Got it. Makes sense. On the AI PC side, John, what we hear about is so many TOPS here, so many TOPS there. Is that the right metric? Is that what we should all be chasing, this person of so many TOPS or that's why that silicon is better?

Yes. There's a lot of different ways to try to define what an AI PC actually is. Clearly, a large software partner, Microsoft, has sort of fixated on TOPS. We've kind of moved a little bit away from that. I mean, quite frankly, if you look at Meteor Lake, which is a 30 to 35 TOP sort of system, it can run natively everything in copilot. And so to say that there's some sort of TOPS requirement that you need to hit to be able to do AI on the PC, not probably the right way to think about it. As you know, we've talked about shipping 40-plus million units of AI PCs this year, going to 60-plus million units for next year. We think that will far and away lead the market, and we're going to dominate this category and help define the category as it unfolds.

I see. And can we double-click on whether you think of AI PC as an enterprise play or as a consumer play and what implications that has had? Because you guys dominate enterprise. Consumer, does that become a more competitive environment? What if there used to be Intel Inside, what if there's Snapdragon Inside, right? How does that...

We think that the utility of the PC vis-a-vis bringing AI to the edge will play out both in the enterprise market and in the consumer market. I would remind you that we have pretty healthy market share across the overall market. And we think we've built very strong ecosystems in both of those segments, as you pointed out. I mean, roughly speaking, enterprise is about half the market, consumer is about half the market. In the enterprise market, a key part of our strength are some of the offerings that we provide around vPro, which makes security and manageability of a remote fleet just very seamless for IT departments. And it's a value for us, and it's a value for our customer, and it makes kind of our position in that place relatively sticky. And so to the extent that you think there's an enterprise refresh cycle coming in the back half of the year, we think we're particularly well positioned for that given our historic footprint. Even if you think that there's not a critical mass of AI applications by the back of the second half of the year, enterprises will likely future-proof their purchases by buying up the stack because they know that PC is going to be in service for 4-plus years. On the consumer side, listen, we think we have a fairly strong ecosystem on the consumer side as well. We spend a lot of time and effort to make sure that our OEM and software partners are extremely successful in that part of the market. And we're doing that today with optimization of AI-infused applications with our integrated software vendor partners. I think we talked about 2 quarters ago, working with over 200 ISPs. Today, it's well north of 500 ISVs. And we think that ecosystem footprint helps us be successful because it helps them be successful. Now what I would say, if you think that the AI PC sits on top of existing ecosystems, we've got a pretty strong position. If you think, for some reason, the AI PC creates a brand-new ecosystem, then I would say that ARM perhaps might have a better opportunity within that space. And I want to be clear, we coined the term "only the paranoid survive". We take all competition serious. We actually like competition. It makes us better. But even before you think about the ecosystems we've been able to build around x86, our silicon road map, I think, puts us in a very strong position. You marry that with our ecosystem, and we feel pretty good about how things are going to unfold for us as the AI PC continues to grow.

Right. Makes sense. I'm glad you brought up this ARM versus x86, right, question. So we have Microsoft, who, I think, for the first time, launched a new PC technology, right, and launched Surface with ARM, then at Computex, the ARM CEO, Rene Haas, right, he said that he thinks there will be a 50% conversion to ARM, right, by 2029. What is Intel's view on how that competition, right, is? Obviously, you made it clear that competition, right, is kind of good for both sides. But how seriously should we take those claims?

I mean, listen, again, we take all competition seriously. And so we're not resting on our laurels relative to ARM coming back into the market. And I'll remind people ARM and Windows PC is not a new dynamic. This is something that was a big concern of the investment community as far back as 2011. And so there's been 14, 15 years of trying to break ARM into the Windows PC market with very little success, in large part because we had a very strong road map; in large part because we had a strong ecosystem; and in large part, x86 PCs not only make us profitable, it makes the OEMs profitable as well. And so we kind of feel like the dynamic really hasn't changed all that much from the 2011 time period. Clearly, Microsoft is throwing more weight behind this. They've done an exclusivity with a single vendor in Qualcomm, and that is up at the end of the year. We fully expect to see other potential ARM suppliers come into the market when that exclusivity is up. But in general, there's been one successful ARM PC vendor in the market, and that's been Apple. And they've had 25-plus years in the market, and they've got about a 10% market share. And quite frankly, you could make the argument that the reason why they do what they do in the market has very little to do with which architecture they're using in their silicon. It's the fact that they've built an ecosystem around their silicon, their software and their hardware. And so we very much believe in the power of ecosystem. There are hard things to build and there are hard things to displace. And we think we've got an ecosystem that's going to make us very competitive as the AI PC rolled out.

Got it. John, you've started the market, right, for a long period of time. Technically, is there anything about ARM that gives it a power efficiency advantage over x86? Or is it just a different implementation and that there is no structural advantage that ARM has?

Yes. We don't believe so. Now we've got to go up and execute. But if you think about designing a piece of silicon, there's 3 vectors that you try to optimize to: performance, power and historical compatibility. To the extent that x86 has a long legacy of historical compatibility, that is one dynamic that we have that the ARM camp doesn't have. But we see that as an area of strength not weakness, especially in the enterprise space, because as you know many enterprises out there are still running software that's 15, 20 years old and having that comfort that no matter what you buy, you just know it's going to work with every piece of your software is something that helps to build that ecosystem lock-in that I talked to before. But I'll give you another example with Sierra Forest and our E-cores. We think our E-core performance on the server side is going to be as good, if not better, than ARM on a performance per watt basis with still all the benefits of that ecosystem lock-in. Again, we have to go off and execute, but there's nothing structural about an ARM instruction set. That means it's always going to be power beneficial over other micro architectures.

So why not stress the battery life? Because I would imagine that as a consumer, I would be more attracted to somebody saying the battery life could be, I don't know, 20-plus hours as opposed to some hazy notion of what AI could be.

And we will, but it's hard to stress that until you actually have systems in the marketplace because the CPU is only a small portion of what drives the overall battery life at the system level.

Right. Makes sense. On the data center side, what are you seeing in terms of just kind of near-term second half outlook? What is the launch of Sierra Forest and Granite Rapids mean in terms of how the market share evolves versus AMD? And then if I could just kind of ask the Part C of that, NVIDIA launching, right, Grace Blackwell. For a lot of people who want Blackwell, I guess, they're getting Grace as well with that better performance. So how does that change the dynamic?

Yes, all very good questions. So we only guide revenue one quarter out. I mentioned earlier that we've given some breadcrumbs about the second half being a better-than-seasonal second half. Let me just build on that across all of our businesses and answer your server question as part of it. But if you think about a better than second half -- a better than seasonal second half, there's a couple of different things I could point to. I think first and foremost, we've got a bunch of businesses that are more levered to the industrial auto and infrastructure markets in NEX, Altera and Mobileye. And just like every other semi company that has exposure to those end markets, they're seeing the worst of their inventory cyclical correction in the first half of the year, and they are expecting some sort of recovery in the second half of the year. And so when you think about that grouping of business, not a bad way to think about it. In addition, we talked about now $500 million plus of Gaudi revenue coming into this year, the vast majority of that is coming in the back half of the year. So when you think about half-on-half growth, that's actually a relatively meaningful contributor driving that half-on-half growth. For the PC market, we talked about the enterprise refresh cycle that we expect to start to pick up into the back half of the year. I will remind you that we were at 270 million units of consumption last year. We're sort of pegging the market to 276 million to 278 million this year, but we were at the low end of seasonality in the first half, and so that will help as far as half-on-half growth. On the server market, we've been a little bit less explicit on talking about the inflection point. Having said that, the way that we're kind of thinking about the market in general is the first half of the year, units are likely going to decline low single digits year-over-year. As we move into the second half of the year, we think they're going to likely increase low single digit year-over-year. So not heroic sort of unit acceleration. Having said that, something we have a little bit more visibility on is core count mix and what's happening with core count mix. And clearly, on a unit basis, core counts are going up and that does have a benefit for us on the ASP front. I'll also remind you that what we've talked about from a market share perspective is this year being plus or minus flat. And that's really off of the Q3 '23 order level from last year. If you remember how last year played out, we actually had a little bit better share in the first half of the year than we were expecting. That took a little bit of a step down in Q3. Q3 to Q4 was flat, Q4 to Q1 was flat. And we think plus or minus flat from here is the right way to think about it. That's mostly still on the back of Sapphire Rapids and Emerald Rapids, while Sierra and Granite are launching this year, volumes release is to begin to ramp late this year into next year. We do think that those are the product suites that give us the ability to go back and win share as we go into 2025. And I think the last part of your question was Grace Hopper. Good question. Listen, clearly, given how tight the market is for accelerated compute today, suppliers have more power than buyers right now. In a more steady-state market, though, I would argue that hyperscale companies would prefer to buy a pool of CPU compute, a pool of GPU compute, a pool of connectivity and do the integration themselves and not give that margin away to any one of their suppliers. And so we do think longer term, that's the right way to think about a normal buying power in the market. And quite frankly, we are in a very strong position on the head node for server-accelerated servers, AI servers, because Sapphire and Emerald are hands down the best CPUs out there for AI. And we think that, that's really due to the single-threaded performance that both brings. Now I think that there's obviously advantages to bundling, but on a stand-alone basis, we feel very confident about the performance metrics that Sapphire and Emerald offer, and that only gets better with Granite.

Is the ability for ARM to penetrate the server market different than what it is in the PC market? Or it's the same dynamic just because it does have a lot of champions kind of on the server side in terms of custom silicon Microsoft or Amazon Graviton and Google now with Axion? And...

Yes. The raw data would suggest absolutely, right? Because if you look at the share of ARM in the server market versus the share of ARM in the PC market, they've absolutely done a good job within the server market. Now a lot of that has been, going back to my ecosystem comment, in areas where new ecosystems have developed, namely internal workloads at the hyperscale companies where they're doing their own silicon. And that is a dynamic that we don't think is going to change all that much. And quite frankly, we have an opportunity to capitalize on that with Intel Foundry, both on the advanced packaging side and on the wafer side. And I would be remiss not to remind people about a year ago, we actually made a joint announcement with ARM, an R&D announcement, multiyear co-development announcement, where we were going to optimize their latest cores on Intel 18A. And so there is a part of our organization that is trying extremely hard to make ARM very successful across all the markets that ARM wants to penetrate, and we are happy to do that. And that internal workload for internal silicon is one where ARM has been very successful and will likely continue to be successful. Now having said that, when we think about the core server market, we've talked about unit growth in that market longer term, in that 1%, 2%, 3% range. We think the more important metric to look at is actually core growth, where we think the cores are going to grow probably at a 15% to 20% CAGR. That would be down from what they've grown over the last 5 years, which is closer to 25%. Now it's off of a bigger base, but it is still very healthy growth. The big swing factor that you have to be willing to underwrite is we're arguing that ASP per core, which over the last 4 or 5 years has been going down almost as quickly as number of cores per unit has been going up, is the rate of decline is going to slow dramatically. Now our closest competitor is actually staying flat. If that's true, then we think it's a market that supports probably high single-digit, low double-digit revenue growth over time, even with all these dynamics.

Got it. John, in the few minutes we have, I wanted to touch on 2 topics quickly. First is Huawei restrictions. So Intel had announced back in May that Q2 could be, I think, at the lower end of 12.5% to 13.5%. Just what does it imply for the full year? Because when we look at Huawei, I think they only ship like mid-single-digit million or so, right? So is it really an impact beyond?

No. So just to level set, we put out an 8-K saying that while we're still tracking to the original range of 12.5% to 13.5%, we're below the midpoint. So we didn't say the low end. We said basically the bottom half is sort of the way to think about that. And that's all being driven by the incremental restrictions. We had a license to sell to Huawei that was always up for renewal in Q3. Our sort of working assumption was it was not going to get renewed. But it is having an impact to Q2. We do think that as we go into the back half of the year, if Huawei can't ship, it doesn't change the size of the market in China, and so we'll be able to recapture that revenue with other OEMs, just really difficult to do intra-quarter. So that would be sort of the best way to think about it. We didn't really quantify the magnitude amount, but your unit numbers were about right. And so if you think about it on a quarterly basis, you can kind of back into the right number and that kind of is consistent with what we said in our updated guidance.

Does it conceptually mean because this is only, I don't know, like a halvish quarter impact, that we will see a full quarter impact in Q3? Or.

No. Because I think -- even if we always assumed in Q3, we weren't going to get the renewed license.

Oh, I see. So that's incremental...

And as time goes by, if Huawei can't ship, we can recapture that revenue with other OEMs who will ship in lieu of them not being able to.

I see. Got it. And then the last topic was on Intel Foundry services, right? So now the name of the combined kind of manufacturing, right, plus external foundry. What came as a surprise was just the period at which Intel would break even in that business. So what went into that thinking? Is it just that the TAMs are different than when the original decision was made? Like, talk us through what does that implies for Intel.

Yes. I'm glad you asked the question because it's a really important point of clarification. When we presented the webinar in early April, what we tried to present to the investment community was a target model, not a forecast, but really a target model that was based on what we thought was reasonable to conservative revenue expectations. And so at the endpoint, the 2030 model, we kind of gave at the Intel consolidated level, a 60-40 breakdown between gross margin and operating margin. That's really based on about $100 billion of revenue in that time period. Where if you look at Street numbers today, we're roughly at $55 billion is what the Street's estimating for 2024. If you kind of break down how we get from $55 billion to $100 billion, we told you that $15 billion of that would be external foundry customer by 2030. It's effectively 0 today in Q1. So that takes that $100 billion down to $85 billion. So what we're really talking about is $55 billion today going to $85 billion. The way we built that up bottoms up was just making an assumption that the core business of PCs, data center and NEX grew at a 3% to 5% CAGR. Again, not a forecast as much as kind of an assumption. You layer on top of that some share gains for accelerated compute. And I think out in the 2030 time frame, that share gain would get us to about a $8 billion to $10 billion number, which relative to the size of the overall market doesn't seem to be that challenging. And then if you look at some of our growth of your businesses like Altera and Mobileye, that's more in that 10% sort of CAGR range. And if you do that math, that's what builds up to $100 billion. Now listen, again, it's a target. It's not a forecast. If revenue comes faster in a fixed cost business, we should be able to get to that -- those margin targets more quickly. More importantly, when you think about the breakeven point, midway between now and 2030 for Intel Foundry, which is effectively 2027, a lot of what we need to see for that to happen is somewhat independent of end demand. It really is mostly about mixing from Intel 7, which is a very uneconomical wafer for us today, to Intel 18A and our ability to pull back wafers from external foundry suppliers that will be more reliant on this year and next. And I would tell you that the economics of moving from Intel 7 to Intel 18 at the endpoints, you almost have 3x the ASP per wafer here at only a modest increase in cost. And so there is a significant economic benefit. Now the time to get that mix shift is just going to take some time. I mean one thing I'll tell you to give you some context here. We talked about 40 million AI PCs this year going to 60 million next year. Now again, that's a forecast. We could be undercalling the market. But if that 60 million is right, it means that 2/3 of our units next year are still not AI PC units, which means that 2/3 of our units in client are still on very uneconomical Intel 7. And so we've got a lot of visibility over time that this mix shift will occur and we just need some time for it to happen and to catch up.

Got it. So I think with that, thank you so much, John. Thank you for your perspective.

Thank you.

Thank you, everyone.

Thank you.