Intel Corporation (INTC) Earnings Call Transcript & Summary

All right. Great. We'd like to get started. Thank you so much, everyone, for joining us. My name is Toshiya Hari. I cover the U.S. semiconductor space here at Goldman. Great to have John Pitzer from Intel here with us. He's a Corporate VP, Planning and IR. Before diving into questions, I need to go through disclosures. 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 and additional information on our non-GAAP financial measures, including reconciliations, where appropriate to the corresponding GAAP financial measures. How did I do?

Perfect.

Okay. Great.

Appreciate that. It's a mouthful.

Yes, it is. It is. So we do have 40 minutes plus, John, which I think is a little longer than typical. So I wanted to start off with a high-level kind of softball question. I think you joined Intel roughly 2 years ago from the sell side. What's the experience been like, any surprises in terms of what you see internally, how you think about the markets? Yes...

Yes. It's a good question. I put a lot of thought into this over the last couple of days because to your point, it's almost been 2 years to the day since I joined. And it oftentimes feels a lot longer than that. And as I think about kind of coming into the organization and what's really surprised me, I'm going to kind of turn the question a little bit. It's actually everything that we've been able to accomplish over the last couple of years, quite frankly. And it's clearly not because of me. It's really started when Pat rejoined a little bit more than 3 years ago. But if I kind of had to kind of focus on kind of the key accomplishments and listen, we still have a lot of work that we need to do, a lot of these accomplishments haven't flown through the P&L or the cash flow statement yet. But the 3 that really stand out in my mind is really the execution engine and getting that back on track with 5 nodes in 4 years. I mean, quite frankly, when Pat first came back to the organization, and made that declaration, a lot of people thought he was crazy. We're now fully through the Intel 7 ramp, the Intel 4 ramp. We have PRQed, Sierra Forest, means Intel 3 is now in the process of ramping. We've talked about getting the 1.0 PDK for AT&A out this quarter. We've got products in fabs that will be ramping middle of next year for release in the second half of next year with Clearwater and Panther Lake. And really, the ability to execute on the 5 nodes in 4 years was something that I think a lot of people doubted, and we're kind of seeing the light at the end of the tunnel on that one. The second one really is how much work the organization has done to stand up the Intel Foundry Services business, now being led by Kevin O'Buckley. I mean, I think that really culminated in February at the Intel Direct Connect event. And to me, I think what was the most pleasant to see is how much of an ecosystem development that we've done since the last time we tried the foundry effort pre Pat, and we're not successful. It was great to see accolades from ARM, from Cadence, from Synopsys, from Broadcom on our packaging. And it really speaks to how far we've progressed in building out that ecosystem. We now have 6 Intel 18A external foundry customers. We've got a lifetime deal value of greater than $15 billion. And I think we're well on our way to our aspirations of being the second largest external foundry by 2030. And then the third thing I would kind of point out as a relatively big accomplishment is everything we've done around smart capital to make sure that we can invest in the strategy in a way that is strategic and yet, I think is beneficial to our owner's capital. And it really starts with the CHIPS Act and everything that Pat and the organization did to drive completion of the CHIPS Act. But I'd also bring up the SCIP 1 transaction with Brookfield, what we've done with Mobileye, the path we've set Altera on, the investments that we've gotten with IMS. It really allows us to have multiple pools of capital to really go pursue the strategy, especially during a period of time where the economics of the business are still playing catch up.

That makes sense. On the foundry ecosystem that you talked about, investing in it, building out the ecosystem, you made a significant amount of progress. Is it fair to say that you have the necessary pieces to be effective and competitive in foundry? Or do you feel like you still have a little bit more work to do? And if you do, what are those pieces...

Yes. It's a good question. I mean, I think from a technology perspective, we're clearly on the right track. And like I said, the next big mile is the 1.0 PDK for 18A, which we expect to get out this quarter. That's going to be an important one. I think from a cost structure perspective, we're still at the beginning stages of that journey, and that was really a key part of the rationale of providing Intel foundry with their own P&L for the first time ever. As you know, historically, in the IDM 1.0 model, Intel foundry was really an allocated cost out to the BUs. And they really focused on 2 things: speed of node transition and having available capacity, and they prioritize that over sort of cost and efficiency. And as you know, being a good foundry is not just about the technology. It's also the cost structure. And quite frankly, it's also the services. And I think that one of the things that we've done a good job on the services side is actually augmenting our internal talent with a lot of external hires that have foundry experiences. The most recent being Kevin O'Buckley, with his background at Marvell, prior to that Avera, prior to that GLOBALFOUNDRIES. And we're happy to have him as the leader on that external foundry strategy.

Got it. In terms of customer engagements in the foundry business, you publicly announced many customers, I think MediaTek, Amazon, most recently, Microsoft. What characteristics of IFS drove success with those engagements and those customers? And to the extent you've had instances where customers said, "No, thank you. We'll continue to work with foundry A, B and C." What were some of the concerns that they had?

Yes. So I think it's not hard to imagine given our legacy and high-performance compute transistors. Our foundry efforts probably have been most successful within the HPC market as opposed to the mobile market. I think one of the things that Ann talked about at the Direct Connect event in February is that we're going to bring out subsequent sort of flavors of Intel 18A that might be more mobile applicable. So that is a market that we definitely have in our crosshairs. But if you look at our success to date, it has really been more on the wafer side within HPC. I think on the packaging side, we've been pleasantly surprised. We've clearly gotten significant incoming interest because of the current capacity constraints for CoWoS. I think what's been good to see is that those tactical conversations have quickly morphed into more strategic conversations as customers start to understand our road map in advanced packaging and how that aligns with their product road map over time. And it really is an area that we've got significant differentiation. We'll start to ramp some of the advanced packaging wins later this year more fully in 2026. And as Pat has said, it's a fantastic on-ramp vehicle to bring customers into the Intel foundry on the packaging side, show them our capability, hopefully under -- over-execute relative to our commitments and get them to morph into a wafer customer. So that's been, I think, a pretty powerful dynamic for us to date.

Got it. Conflict of interest is something that comes up -- still comes up in my conversations with investors. I know you guys segmented the 2 businesses, if you will. Is that something that still comes up with customers or potential customers? And I guess, part two is the Intel product team. Are they truly free to work with IFS or TSM or what have you?

Yes. I mean, listen, it's a potential concern. It's not the most significant one. The most significant one is can we execute to our technology road map and to their needs from a product perspective. And I'll remind you to see, and you know this well, coopetition is a very common dynamic across the tech sector. So it's not as if these companies haven't dealt with the coopetition dynamics in other parts of their business. I think that we've taken great pains to kind of separate the 2 operations. Dave has talked about separate legal entities for both, separate ERP instances for both, separate P&Ls for both. And to be clear, the giving Intel foundry P&L was really a way to drive different behaviors internally, and that was the primary reason why we did it. Giving more transparency to our owners, giving more transparencies to potential customer was an added benefit. And so we think we're managing through that fairly well. I would also remind you, we are better together. Having Intel products as a customer 0 helps to derisk the yield curve for subsequent external customers that come on to a node after that. And so there is some comfort level in there. To the second half of your question, how much agency does Intel products have to use external foundries? I mean, listen, I think, if Michelle were on stage right now, she would say, I'd always want TSMC to be a supplier of tiles for me. It's healthy to have that competition between an external foundry and the internal foundry. I think clearly, as we get our roadmap back in shape on 5 nodes in 4 years, we think that what we call the PPA, or performance, power and area, deficit that we're in today goes away. And it just makes a lot of sense from a better product standpoint and a better economics for the overall company to start bringing some of those tiles back. And as we've talked about over the next couple of years, our client strategy has significant tiles going to external foundries. It is a headwind to gross margin improvement in 2024 and 2025. But as we get to Panther Lake and Intel 18A late next year, we think we have a real opportunity to pull wafers back in. And as Pat has mentioned, we think there's about 2 [ mods ] worth of potential incremental wafers that we can pull back in over time to help fill our fabs.

I think you sort of answered my next question, but IFS profitability, you've talked about losses peaking this year and reaching breakeven sometime between now and 2030. And I think you mentioned a couple of things there, but the key drivers that get you there?

Yes. So the breakeven was the midpoint between now and 2030. If you kind of do the math on that, it implies somewhere in 2027, we think we're going to break even. And as we talked about in the event we hosted in early April, we tried to provide a target model. And I'll stress it's a target model, not a forecast that had relatively reasonable to conservative revenue assumptions. And so if you look at the endpoint of sort of a 60-40 model by 2030, that's really based on about $100 billion of revenue in that time frame. The way you get there is $15 billion coming from external foundry, $85 billion coming from Intel products and sort of Intel all other Mobileye and Altera. And the breakdown of that is really an assumption of about a 3% to 5% growth rate in our Core businesses. Some accelerator share gains and Mobileye and Altera, which are historically faster-growing businesses, growing more like a 10% CAGR. And so we think all of that is a relatively reasonable revenue forecast to be able to drive that margin structure. If you look at the point between now and breakeven, I would sort of make the argument that it's even less dependent on demand growth and consumption growth because really what we're counting on is a mix shift of -- in our wafer capacity from uneconomical Intel 7 to very economical Intel 18A and the ability to pull tiles back in, independent of what PC unit growth does. And so a lot of that is in our control. And I think one of the things that we wanted to stress in early April is that if you look at the move from Intel 7 to Intel 18A, the ASP per wafer goes up almost 3 times. The cost per wafer doesn't change that significantly. And it really kind of illustrates how uneconomic the Intel 7 process is without EUV, with all that multi-patterning. And so while it's going to take time for that mix shift to recur, we have a high level of visibility and confidence that we will get there in that time line, somewhat independent about what happens with demand. Now as you think about the 60-40 model out to 2030, clearly, scale will help get us there more quickly, and it's more demand dependent. But we've got a lot of the levers to get to breakeven that are in our control.

Got it. And the tiles coming back in, is that a '26 dynamic? When does that start to kind of go in motion and...

Yes. So this is really the first full year where we're using the tile strategy for our client business, and that's clearly Meteor Lake, Lunar Lake and Arrow Lake. Those units will ramp again next year. And so that exposure goes up next year. And as we get to Panther Lake really at the end of '25 going into '26, we'll have opportunities to pull those tiles back in. But it does take time. I mean one of the contextual points that I try to make is we've given a forecast of 40-plus million AIPCs this year going to 60-plus million units next year. And that's a forecast we could be under calling the market, but that's about in line with what other people are thinking. If that holds, it still means that 2/3 of our units next year in client are not AIPCs to kind of give you some context of how long these transitions sometimes take to flow through the P&L and the cash flow statement.

Got it. Makes sense. Shifting gears a little bit. In the world of AI, obviously, a lot going on. I think there's a CPU to GPU sort of trend currently, and there's one company with a very loud megaphone. We did have Renee James here at the conference yesterday. And she talked about the important role CPUs play. And I think you guys have talked about that, too, particularly within the context of inference. Maybe you can talk a little bit about that and then more importantly, your target, your goals in acceleration as it relates to Gaudi and Falcon Shores.

Yes. Let me flip the order of that question and hit accelerators first and then transition into the importance of CPUs within all of this. Clearly, we're happy from a Gaudi perspective that we've moved from pipeline to backlog in revenue. And we gave out a target of greater than $500 million in revenue this year from Gaudi, which is a decent start. But relative to the overall size of the market, we're not satisfied. We think momentum is growing well. I would point out specifically with our Intel Developer Cloud, we're seeing very high interest, which is really our way to try to bring enterprise customers onto a Gaudi platform in kind of a frictionless, seamless way for them, and we're seeing good traction there. We have said that if we had more supply this year, we'd likely be able to ship more Gaudi. We've secured more supply for next year. I think the important milestone for us is really the move to Falcon Shores, which we talked about as being a late 2025 event and really having our first true GPU offering in that market. But we are very excited with Gaudi as an accelerator for the applications that we focus on. It has superior TCO to many other accelerators that are in the market, and we think it's a compelling product and will be for the next 18 to 24 months and even longer. And as you think about sort of the role of CPUs, one of the things that we try to level set investors on is we don't really think about AI as necessarily being a TAM as much as a workload that's got to span across the compute continuum. You've got to go from data center into network, into enterprise, into client and into edge. Rightfully so, today, all the activity is really hyperscale, training and inference of these ginormous LLMs, where a GPU is particularly well suited from a TCO perspective. But we're in the bottom of the first inning of the AI deployment. And quite frankly, the next, I think, longer and larger phase will be as you deploy AI out from the center, out to the edge with inference on smaller LLMs. And the point that we keep trying to make is, as you think about that compute continuum, depending upon where you are and what application you're running, there's going to be an optimal sort of silicon solution that's going to drive the best TCO. And as you move into networking, enterprise and definitely client and edge, CPU is going to have an important role to play on that. You're seeing that today with all the excitement around the AIPC. We are the first to market with the AI CPU for the PC with integrated NPUs. I love the fact that Microsoft is looking to monetize all their investment in the cloud by pushing AI to the edge with the AIPC because I think it's a good instructive model of how AI ultimately gets monetized over time. And as you move from that center out, you're really going into ecosystems and really silicon product portfolios where we have a lot of strength. And we think that, that's going to bode well and make us very levered to this trend independent of what happens in core acceleration in the cloud.

Got it. And just going back to the accelerator side of the portfolio, Gaudi and Falcon Shores, how do you compete with the incumbent? How do you differentiate vis-a-vis the incumbent?

Well, right now with Gaudi, it's a TCO argument. And some of that is entry price, but some of that is just raw performance as well. And like I said, for applications where we have the ability to go in and optimize for a customer, the Gaudi TCO is competitive with everything in the marketplace right now. And in many instances, more competitive from a TCO perspective. The issue with Gaudi is it's not as programmable as a GPU. And so as you think about trying to have a leverage sort of product portfolio into that accelerator market, Falcon Shores is going to help a lot. And remember, Falcon Shores doesn't get rid of Gaudi. It incorporates the best of Gaudi with our GPU technology. And as I said earlier, we'll have that in market in late 2025.

Got it. Got it. On the server CPU side, I had a couple of questions on the road map. I guess before we go there, maybe on the demand environment, '23 was a very challenging year for the industry. I think Mercury had units down more than 30%. And with AI, CPUs kind of got boxed out, if you will. Both you, your nearest competitor, others have sort of sort of suggested that the environment is bottoming and hopefully improving. What is your outlook into the back half? And as you think about your overall business growing year-over-year for the full year, what's embedded from a server CPU perspective?

Yes. I'll just remind, we only guide revenue one quarter out. Implied in our original guidance was really for our DCAI business to be relatively flattish sequentially Q1 to Q2, which would have been at the low end of seasonality. Now remember, we did inter-quarter update with an 8-K. That was more about consumer and NEX than it was about DCAI. So that flattish for DCAI probably still holds for Q2. As we look into the back half of the year, clearly, some of the statements we have made would suggest a better-than-seasonal back half. I think as we think about server market, in general, that's more about core growth and ASP per unit than it is any significant sort of unit rebound that we're embedding in our forecast. And so roughly speaking, in the first half of the year, units are going to probably be down low single digits year-over-year. Roughly speaking, in the second half of the year, they're probably going to be up low single digits year-over-year. What we can really have some visibility into is what's happening to core count around those units and what's happening to ASP per unit. And so we do expect a revenue uplift that is larger than the expected unit uplift as we go into the back half of the year. But clearly, we're still fighting what a lot of companies in tech are fighting right now, which is accelerators are eating up a lot of the IT budget. You heard that from software companies as late as last night in their earnings, and so we're going through that today, but we do see, as you pointed out, kind of a bottoming process.

Okay. Understood. In terms of the road map and how to think about competition to your prior point, you're launching Sierra Forest, you've got Granite coming in, I believe, in the second half. I guess, first of all, what's been the early feedback from your important customers on both platforms? And how should we think about your server share stabilizing going forward?

Yes. Feedback has been good. As we talked about, we PRQ-ed Sierra Forest this quarter, Q2. Well, PRQ i.e., launch Granite in Q3. Sierra is our first E-core server product, E being our efficient core. All the Rapid lines are our performance core, or RP core. What we've talked about relative to share is from the Q3 level of last year, we see share this year plus or minus flattish. Now I think that's still more driven on the back of Sapphire Rapids and Emerald Rapids because, while Sierra and Granite will be ramping in the back half of the year, they become a much bigger part of the equation more next year than necessarily this year. We do think as we go into next year with that product portfolio that we have the opportunity to win back share in the server market.

Okay. And most of those gains would probably come on the cloud side because enterprise, you never lost...

Yes, it's a very good point. I mean it really speaks volume to our ecosystem strength within enterprise because if you look at the share loss that's incurred, it's been mostly in the cloud, as you pointed out.

Got it. ARM-based custom CPUs. I think Graviton has been pretty successful, obviously, within the context of their own infrastructure. Microsoft, Google have recently come out with announcements. How do you think about that? Again, to your prior point, there's a bit of a coopetition dynamic there. On the foundry side, you could certainly service them. So how do you think about that balance?

Yes, it's a really good point. I mean I think we have a great opportunity on Intel foundries, both on the packaging side and on the wafer side to really capitalize on hyperscale companies that want to do their own internal silicon. And we're going to drive that to the best of our ability to be levered to that. What I'll tell you is we think about the market overall, those internal solutions have been very successful for internal workloads. They've been significantly less successful for external workloads, where we think the x86 ecosystem lock-in is really important. Now clearly, from a performance per watt perspective, both ourselves and AMD need to be competitive. And we've got product portfolios that we'll do that. But we think going forward, the hyperscale companies aren't going to stop doing what they're doing relative to their internal workloads. And we've got a strategy to capitalize on that. But for those external workloads, we think we're very well positioned.