Intel Corporation — 2024 Q1
Transcript
Each turn shows the speaker, their inferred role, the section, and that turn's net sentiment (×1000).
Thank you for standing by, and welcome to the Intel Corporation's First Quarter 2024 Earnings Call. [Operator Instructions] As a reminder, today's program is being recorded.
Thank you, Jonathan. By now, you should have received a copy of the Q1 earnings release and earnings presentation, both of which are available on our Investor Relations website, intc.com. For those joining us online today, the earnings presentation is also available in our webcast window.
Thanks, John, and welcome, everyone.
one, the beginnings of an enterprise refresh cycle and growing momentum for AI PCs; two, a data center recovery with a return to more normal CPU buying patterns and ramping of our accelerator offerings; and three, cyclical recoveries in NEX, Mobileye and Altera.
Thank you, Pat, and good afternoon, everyone.
Thank you, Dave. We will now transition to the Q&A portion of our call. [Operator Instructions] With that, Jonathan, can we take the first question?
And our first question comes from the line of Ross Seymore from Deutsche Bank.
I guess for my first question, I wanted to dive into the demand side of the equation. What was weaker in the near term than you had expected? And much more importantly, it seems like the back half you're going to have double-digit sequential growth in largely both quarters, so that's significantly above seasonal. I know you went through some of the reasons at a high level, but can you dive a little bit deeper into what gives you that level of confidence in the second half ramp?
Yes. Starting out, so Ross, thanks for the question. I'll just say the market was weaker. You've seen that in a number of others that have commented as well. So I'll say somewhat across the board a bit. We've seen that, cloud customers, enterprise, across geos. So I'll just say a bit weaker demand, right, we'll just say at the low end of seasonality Q1 to Q2 that we saw. And as we go into the second half of the year, we're engaging deeply with our customers today, our OEM partners, and we just see strength across the board, right?
Ross, do you have a quick follow-up?
I do. Maybe for Dave, on the gross margin side, nice upside in the first quarter, but the drop in the second quarter is a little bit puzzling with revenues going up. So could you just talk a little bit about that second quarter drop and then the confidence in the rebound in the second half? Is that just revenue driven in the second half? Or what's the key metrics there, please?
Yes. Good. Thanks, Ross. Maybe start with Q1 because it somewhat explains Q2. We had better sell-through of product. I hadn't even mentioned Meteor Lake strength. That better sell-through was on previously reserved material. And so we just saw some upside in gross margins because of that. We had a little bit more of a flattish plan between Q1 and Q2 in terms of how that would flow through. And so it kind of pulled some of the benefit of gross margin improvement we would have seen in Q2 and kind of pulled it into Q1. So that was part of it.
And our next question comes from the line of Ben Reitzes from Melius.
Appreciate the chance to ask a question here. Pat, can you talk a little bit more about servers in the data center? There was talk of a bottom there in previous discussions. How do you see that kind of going throughout the year in light of your 2Q guidance? And what's the catalyst for the pickup there?
Yes. Thank you, Ben. And obviously, as we look at our position in the data center, I'll just say we're stabilizing. And with that, we're improving our competitiveness. We also see, as I mentioned in the comments, that the ASPs are going up comfortably as well. So socket fairly stable through the year, but the ASP per socket with increased core count improves our position. And then new products like Sierra Forest or Xeon Gen 6 product, definitely gives us power performance capabilities. So overall, we're seeing a very healthy growth rate, mid-20s as we go through the year. We're also seeing increasing interest in the AI capabilities of Xeon. And we're winning head node positions, and we're seeing pretty extraordinary performance at Vision.
Ben, do you have a quick follow-up?
Yes. Thanks. Can we just double-click also on Gaudi, $500 million in the back half of the year? I think you previously talked about a couple of billion in the pipeline. What does that say about your yield to revenue on an annualized basis with AI? And is there an update on the pipeline and your confidence there heading into 2025 on the accelerator front?
Yes. Thanks, Ben. And obviously, pipeline converting into revenue, revenue is much more meaningful and as we said, greater than $500 million for the year, and that's obviously quarter-on-quarter accelerating rapidly, which also gives a great indication for the business in '25 as well. At our Vision event, we had over 20 customers publicly describing their embrace of Gaudi 2 and Gaudi 3. And I was super pleased to see the breadth of those customers. It was CSPs like Naver and Ola and IBM Cloud. It was ISVs like Zeekr, right, coming on board, but maybe most importantly, enterprise customers.
And our next question comes from the line of Joe Moore from Morgan Stanley.
I wonder if you could talk to the server road map. It sounds like you're confirming the time frame for both Sierra Forest and Granite Rapids. Can you talk about -- is there demand for the Sierra Forest product as well? Do you expect that to be bifurcated where you see demand for both? And then how quickly will you see those products come to volume?
Yes. So Sierra Forest, our first Xeon 6 product on Intel 3, and I'm super proud, right? Now we have a leadership process technology back on American soil for the first time in a decade. This is really exciting. And Sierra Forest, high core count, 144, 288 core product, very focused on power, performance, efficiency, and we do see a good pipeline of customers and a good pipeline of, I'll say, socket win backs because the area of power performance has been an area that we've been carrying a deficit, being on an older node. And now that we're on leadership nodes, we definitely see share gains for that.
Joe, do you have a quick follow-up?
Yes, I do. Thank you. On the foundry webinar, you had sort of talked about Intel 3 volume being kind of more of an inflection next year. Does that mean it was in server that these Intel 3 products are sort of get to volume crossover kind of some point next year? Or could we see -- obviously, the leadership you just talked about is important. What's kind of keeping you from getting those products ramping in the second half?
Yes, Joe, thank you. And servers always just take a while to ramp. Customers bring them in, they qualify them, they test them because they're generally putting these things at scale. So there's just an adoption cycle for server products. And the numbers that I'm holding my team accountable for are some of the most aggressive volume ramps that we've ever achieved on server products. So we're driving them very hard. That said, in terms of the total wafer volume this year, right, it's dominated by Intel 7. And the Intel 4 and 3 wafer volumes become much more prominent next year, and that's what I was communicating on the webinar.
And our next question comes from the line of Vijay Rakesh from Mizuho Securities.
Just a quick question on the Grand Rapids, any thoughts on the timing? And do you expect to regain some computing share, server share there with those ramps?
Yes. Thanks, Vijay. I'm building a little bit on the last question. Granite Rapids will come in Q3 of this year when we'll have the production release of that product. Same as -- it just takes some time for customers to get comfortable, qualify, and bring those products to marketplace. But Sierra Forest, Granite Rapids, these are much more competitive power performance products on Intel 3. So we see them stabilizing and then giving us opportunity to regain share. And as we go into next year, we expect that we're regaining share as we end this year and go into next year. These are great products and we're going to be ramping them very aggressively with our customers.
Vijay, do you have a follow-up?
Yes. Thanks. Just on the GPU side, on the AI side, any parts on Falcon Shores? Any preliminary takes on that? How do you see that building out into '25?
Yes. And Gaudi 3 announcement this quarter, extremely well received. And as I mentioned already, 20-plus customers for Gaudi 2, 3, so we're seeing that build. Obviously, Falcon Shores will build on that momentum. We'll be bringing that late next year when Falcon Shores when we combine the great systolic performance of Gaudi 3 with a fully programmable architecture, and all of that comes together with Falcon Shores. And then we have a rich -- a very aggressive cadence of Falcon Shores products following that. We also added the Gaudi 3 PCIe card to it. This use case of Xeon plus an accelerator or Gaudi accelerator is getting very good response from customers as well. So we'll be bringing that out later this year. But the real story is delivering the TCO value, delivering the enterprise use cases. Falcon Shores will just build on the momentum that we're establishing with Gaudi 2 and 3.
Our next question comes from the line of Timothy Arcuri from UBS Securities.
Dave, I also wanted to ask about gross margin. You did say it's going to be better next year, but it is really whipping around a lot. And it looks like you sort of have to exit this year at 48 or maybe a little higher, which is already well above the 45.5 that you'll be at this year because you're guiding it up to 100 basis points. So I know you don't want to guide next year, but if you can even qualitatively help us, can you sustain those margins at that level? And I asked because last year, you sort of exited at 49% and then things crashed here during the first half of the year. So can you help us just think about what some of the puts and takes will be next year off of that high base if you're going to exit this year at?
Yes. Good question. So there will be additional start-up costs next year. We do think it on a percent of revenue basis, it will be lower. So that should help lift the margins. Of course, the expectation would be we see growth in revenue. That also should help. On top of that, we already are seeing good decision-making and changing decision-making around how we operate now under this new different business structure that we have at this point. A lot of that stuff doesn't actually show up in the P&L. We have all these decisions get made this year, but a lot of the decisions made -- sorry, a lot of the benefits to those decisions don't show up until next year and the year after. So we should see some benefit from that as well.
Tim, do you have a follow-up question?
I do. I do, yes. So I want to ask about server CPU share. March -- I think the assumption for March was that service share was going to be pretty flat. So the question is, was that the case? And it sounds -- you sound maybe a little bit less optimistic, if I'm sort of reading between the lines, on share into the back half of the year, just given how long it takes these things to sort of impact your share. So your bullish outlook in the second half of the year, it sounds like it's more market-driven versus share driven. Can you just clarify that?
Well, overall, I like to say it's hard to predict, right, exactly how these will play out in light of the overall gen AI surge that we've seen. That said, products are good, right? We came into the year improving our market share position in the first quarter of the year. It does take time to ramp these new products. But better products, rebuilding trust with our customers that we're delivering on these and now hitting the, what we would call the early end of the cycles on these new products is giving us a lot of interest with the market and the customer. New use cases also demonstrated a 70 billion parameter model running natively on Granite Rapids at our Vision event, all of these just make us more and more confident in our business execution.
Our next question comes from the line of Srini Pajjuri from Raymond James.
My question is on the client side. I think, Pat, you mentioned something about supply constraints impacting your 2Q outlook. If you could provide some color as to what's causing those supply constraints and when do you expect those to ease as we, I guess, go into the second half. And then in terms of your AI PCs, I think you've been talking about $40 million or so potentially shipping this year. Could you maybe put that into some context as to how it actually helps Intel? Is it just higher ASPs? Is it higher margin? I would think that these products also come with higher costs. I just want to understand how we should think about the benefit to Intel as these AI PCs ramp?
Yes. Thank you. And overall, as we've seen, this is a hot product. The AI PC category, and we declared this as we finished last year, and we've just been incrementing up our AI PC or the Core Ultra product volumes throughout. We're meeting our customer commitments that we've had, but they've come back and asked for upside on multiple occasions across different submarkets. And we are racing to catch up to those upside request, and the constraint has been on the back end. Wafer-level assembly, one of the new capabilities that are part of Meteor Lake and our subsequent client products. So with that, we're working to catch up and build more wafer-level assembly capacity to meet those.
Srini, do you have a follow-up?
Yes, John. Thank you. And I guess my other question is on the other bucket. I think, Dave, you kind of talked about Altera potentially exiting the year at a $2 billion run rate from current levels. That's a pretty steep ramp. And also, I think you said next growth will accelerate over the next couple of quarters. So -- and given the telecom weakness out there that we're seeing, I'm just curious as to what's giving you that visibility or confidence. I mean is this driven by some new products? Or is it just the market recovering? Any color would be helpful.
Yes. On Altera, and this is not unprecedented when you see a massive work down of inventory, of course, that significantly impacts the revenue. But as that normalizes, then you start shipping to end consumption. So it's actually a pretty easy lift to get to the $2 billion mark once we're through the inventory digestion period. So I think we have high confidence on that.
And others have commented on their inventory cycles as well in the FPGA category. We have good products in the second half of the year, with Agilex starting to ramp as well.
And then on NEX, of course, that business also has gone through its own inventory adjustment. So we have good confidence around that reversing, which will help drive strength. And then some of the products that are more tailored to the AI space, of course, we'll see like, at NEX, for example, we'll see strength through the year. And so that should drive good revenue growth through the year as well.
Yes. And also in NEX, the AI networking products are strong, our IPU products, we're seeing strength in that area. So it's inventory as well as products. Even though, as your question suggests, the communication sector and the service providers, that is weaker through the year, but pretty much every aspect of their business in edge AI, as Dave said, is seeing strength as we go into the second half of the year and into '25.
Our next question comes from the line of Vivek Arya from Bank of America.
Pat, just a conceptual question. In a gen AI server with accelerators, how important is the role of a specific CPU? Or is it easily interchangeable between yours or AMD's or ARM's? I guess the question is that if most of the workload is being done on the accelerator, does it really matter which CPU I use? And can that move towards gen AI servers, essentially shrink the TAM for x86 server CPUs, because a number of your cloud customers have announced ARM-based server alternatives. So I'm just curious, how do you think about that conversion over to gen AI and what that means for x86 server CPU TAM going forward?
Yes. Thanks, Vivek. And we spoke at our Vision event about use cases like RAG, retrievable augmented generation, where the LLMs might run on an accelerator, but all of the real-time data, all of the databases, all of the embedding is running on the CPU. So you're seeing all of these data environments, which are already running on Xeon and x86 being augmented with AI capabilities to feed an LLM. And I believe this whole area of RAG becomes one of the primary use cases for enterprise AI. And if you think about it, an LLM might be trained with 1-, 2-year old data, right? But many of the business processes and environment are real time, right?
Vivek, do you have a quick follow-up?
Yes. Thank you. Maybe one for Dave on the potential operating loss, kind of how do we model that for the foundry business. So let's say, if I exclude the $2 billion in depreciation headwind, which I'm assuming is almost all going to your foundry business. What is the right way, Dave, to think about foundry operating income or loss this year? And how much of external foundry revenue are you expecting this year?
Yes. Good question. The operating losses will pick up. We roughly were at like 2-, 4-ish in the first quarter. It will pick up in the second quarter, given the start-up costs are increasing and I would say, be roughly in that range for the remainder of the year. And then what I said before is we see that improving then going into '25. And Pat's given me the order, he wants to see every quarter some improvement in the operating loss ultimately to get to breakeven midway through the point between now and 2030. And I think that is very achievable. Sorry, Vivek, what was the second question?
Vivek, are you still on?
No, we've moved on.
Why don't we go to the next caller, Jonathan?
Our next question comes from the line of Matt Ramsay from Cowen.
Yes. Pat, one question I've been getting from some folks, and I totally understand the lead times of starting some of these programs to put increased tile volume at external foundry, but you guys have made the progress on the 5 nodes in 4 years, as you highlighted multiple times. Is there any flex at all to bring back some of that external volume earlier? And I think it matters to some folks because it's a demonstration of you guys being able to ramp your own product to volume and to yield and to economics on 18A, which might give some indication to some external customers that are looking at your foundry business. So just any flex at all to pull that timeline in of sort of reshoring some of the external tiles?
Yes. Thanks, Matt. And largely, those decisions are made when the product decisions are made. So there's limited flexibility to move them around. And if you pick a process node for a certain tile, generally, that's the process node that it's on. So there's limited flexibility there. And many of those decisions, as we've highlighted before, Matt were literally made years ago, right? And those choices were made.
Matt, do you have a quick follow-up?
Yes, John. Thank you. I wanted to ask a question about the AI accelerator road map. So you guys have Gaudi 3 that you talked about and Falcon Shores coming next year. And the hardware looks quite good. I wanted to ask a question about the software that goes on top of that for both -- well, really for inference but also for training. How do you feel about the software road map that you guys have in the AI space going forward? And how much compatibility, or uniqueness rather, is there to the software that runs on Gaudi 3 versus what will come on Falcon Shores and the forward road map?
Yes. Maybe 3 different points there. The first one is for inferencing, you need a whole lot less software compatibility, right? And as the market is more focused on inferencing going forward, if you can run the models, right, in the context of the databases and the other, so that portends and that's why we're seeing the strength that we're seeing right now, Matt, in these use cases. And clearly, some of the software compatibility issues of a GPU have led to the training environments that have been challenging for us. But now as customers get much more focused on enterprise use cases, inferencing TCO, we're finding a lot of strength in the offerings that we have. And as we've matured a number of customers now, we've worked through many of those use cases and getting quite a lot of acceptance of the software stack that we have with Gaudi 2 and 3.
Thank you, ladies and gentlemen, for your participation in today's conference. This does conclude the program. You may now disconnect. Good day.