Intel CFO: Our 10nm Will Be Less Profitable than 22nm [Morgan Stanley Transcription]
by Dr. Ian Cutress on March 4, 2020 11:30 AM ESTThis week at Morgan Stanley’s Analyst Conference, Intel’s CFO, George Davis, sat down to discuss the future of where Intel’s profitability lies. No stranger to the odd comments relating to how Intel manages its money, Mr. Davis was on fine form explaining that Intel is going to be in for a rough time as it corresponds to the leading edge. Among the statements made, Mr. Davis confirmed that Intel’s new 10nm node will be less profitable than its 22 nm node, let alone its 14 nm node.
Mr. Davis was quite frank in his comments, perhaps alluding to the fact that Intel might be willing to accept that 10nm development and deployment hasn’t gone quite the way the company had anticipated. In one statement, Mr. Davis said that
“Look, this just isn't going to be the best node that Intel has ever had. It's going to be less productive than 14 [nanometer], less productive than 22 [nanometer] … The fact is, like I said, it isn't going to be as strong a node as people would expect from 14nm or what they'll see in 7nm.”
Intel’s current 10nm portfolio, both released and announced, includes
- Cannon Lake (pre-10nm*, EOL, 2017),
- Ice Lake (10nm, launched 2019),
- Agilex FPGA (10nm, launched 2019),
- Tiger Lake (10nm+, late 2020),
- DG1 (10+, late 2020),
- Snow Ridge (10nm, late 2020),
- Lakefield (10nm/22nm, 2020),
- Ice Lake Xeon (10+, late 2020?),
- and Sapphire Rapids (10++, 2021?).
*Intel still isn’t completely decided if it wants to make Cannon Lake a proper thing in its history books, despite the fact that the company stated at CES 2018 that it had met commitments to ship the product by the end of 2017.
In order for Intel to maintain its market share in key areas, obviously it still needs its 10nm portfolio going forward. The company was keen to wheel out its first Tiger Lake working samples this year at Computex, for example, including a wafer of Tiger Lake silicon.
There was also commentary on how 10nm will be in terms of performance, essentially confirming that Intel isn’t striking its performance targets as required.
“But still, the effect of 10nm in 2021 is just, it's sort of built today because you've got to get through that product cycle and the node. We're excited about the products but you know, the node isn't going to be quite the performer that historically we've had… Interestingly, and indicative of how we're approaching process technology going forward, we also have 10+ coming out this year… it's hard to find a conference [where] we've been able to talk about some of these things”
*Update: Intel reached out to us to say that in this paragraph, even though Mr. Davis continually talks about 'performance', in the financial world this is reference to the gross margin performance, rather than the standard semiconductor metrics we're used to. I've advised Intel to make a note in Mr. Davis' ear that because of the product stack he's dealing with, 'performance' has some very rigorous definitions, and if he means to continually talk about margin, he should be saying 'margin' and not 'performance'.
Of course, we’re happy for Intel to reach out to us to talk about these things.
Intel 10+ Tiger Lake Wafer shown at CES 2020
As part of the comments made at the Morgan Stanley Analyst Conference, they include that Intel will only be in line with its competition at the foundry level by the end of 2021, when the company launches its next generation 7nm node using EUV technology. When this happens, the competing foundries will be at 5nm, however the transistor density is expected to be equivalent between the two for their high density libraries.
"So we bring a lot of capability to the table for our customers, in addition to the CPU, and we feel like we're starting to see the acceleration on the process side that we have been talking about to get back to, you know, parity in the 7nm generation and regain leadership in the 5nm [generation]" said CFO George Davis.
Intel intends to retake its former foundry competitive advantage when it launches its 5nm processor node, which even on the best estimates is 2024 at the earliest (and likely later), when its competitors are talking about 3nm and using the next generation of transistor technology with gate-all-around designs (known as GAAFETs) to replace ‘basic’ FinFETs. At this time we may be exploring High-NA EUV at the leading edge, or at least getting ready to install new optics for it.
With regards Intel’s current portfolio, Mr. Davis stated that Intel will be overlapping its process technologies quite considerably. This echoes statements made by Dr. Murthy Renduchintala, who has stated that Intel will disaggregate the naming conventions of its products from the technology used for future process nodes.
“[In] order to regain process leadership we [had] to accelerate the overlap between 10nm and 7nm, and then 7nm and 5nm. So the cost that you're absorbing, starting in particularly 2021, you’ve got this intersection of the performance of 10nm, the investment in 7nm, and we’re also well into starting the investment in 5nm” said Mr. Davis
Intel has also had slides that suggest that it will create future products with back-porting and back-porting opportunities in mind, should a particular process node not perform as the company has expected. This means that any initial products for a given process node might end up on the more advanced version of an earlier node of a node can’t ramp in performance/power/margin within its expected time frame. As Intel and other will note, back-porting is a difficult thing to do, however it may result in fewer failed product launches as process nodes enter risk production.
Transcription
Following is a transcription of the Morgan Stanley TMT event. The transcription was AI pre-processed with a light manual editing to improve readability. The original audio can be listened to below, or through Intel's website.
Joe Moore 0:00
Good morning, everybody, I'm Joe Moore for Morgan Stanley, and I really wanted to welcome everyone to the Morgan Stanley TMT (Tech, Media and Telecom) conference. Thanks very much for your partnership, and we certainly want to make this the best possible event. [We] appreciate [you] being here, everyone being here under the circumstances. I was told to tell people, you know, shake hands if you want to, but you know, don't feel obligated to. So you know, be safe. Enjoy the conference, and it should be should be great.
Joe Moore 0:27
Very happy to have George Davis here, from Intel. I've actually been asked to read the Intel safe harbor as well: Today's presentation may contain forward looking statements. All statements made that are not historical facts are subject to a number of risks and uncertainties and actual results may differ materially, please refer to the Intel most recent earnings release form 10 Q and 10 K for more information and the special risk factors that could cause actual results to differ.
Joe Moore 0:54
George, thank you so much for coming. I think maybe you want to give a couple of minutes of kind of opening remarks.
George Davis 1:00
Sure, thanks Joe. Good morning to everybody, thanks for coming.
George Davis 1:07
It's been quite a good year for Intel over the last 12 months, we've finished on a very high note, with demand really across the board much stronger in 2019 than we anticipated. Obviously, we'll spend a little bit of time talking about 2020 here in a minute. [I] would say [2019 was] much stronger for PCs than anticipated and for our data center products [were] also much stronger second half. In fact, many of you may remember there was a lot of discussion in our guidance about the strength of the second half [of 2019] that we saw and relative to sort of historical patterns with a swap first half strong second half, but we had seen a very soft first half and then a bounce in the second half, and we were both wrong. We were wrong on how strong it was going to be and turned out that it did turn out to be a much, much better year overall.
George Davis 2:16
I would also add that what we saw is a market that is rapidly moving to more and more processing power, which means that in the PC end we're seeing clients going for higher performance PCs, and on the data center end we're seeing people buying our highest SKUs. So part of the strength that we saw in 2019 was really ASP (average selling price) driven, or 'mixed' driven as we would typically describe it, where you saw people buying a much, much richer mix of products throughout the year. So a lot of momentum in compute and, you know, we talked a lot about the world beginning to look like it's creating compute centers, more and more in more and more places. [People] often refer to this as Internet of Things or the Edge, and we're finding that there's a lot of demand and growth in our edge related businesses, whether it's IoT, whether it's Mobileye, whether it's supporting the network. All these dynamics are part of a very strong push for growth in the data centric world.
George Davis 3:30
Maybe I'll stop there, Joe, and let you ask questions.
Joe Moore 3:33
You know, ordinarily, I like to start with bigger picture, vision questions, but I think at the moment, with all the focus on CoVID19, maybe we could just start there. You know, you made some, some of the issues around that, in late January reported, just maybe, you know, tell us what your what your latest info is in terms of the virus?
George Davis 3:53
Obviously, China, [is] the epicenter of this. I'll briefly mention what we have in China, for those may not know. We have our NAND manufacturing plant is in Dalian. We have a test and packaging facility in Chengdu, and then we have design centers in both Shanghai and Beijing, along with, you know, kind of corporate groups in Beijing. What we've seen, being in areas that weren't the hardest hit initially, we've been [through] this period with some modest disruption as you would expect to maintain relatively normal operations in China. Obviously, our priority through this whole period has been employee safety. We've been able to, to manage through this quite effectively. Obviously, like many companies, we have a lot of protective measures and the facilities for employees. But what we are starting to see obviously, you can see it from the impact that it has on [current] events. It [has] spread around the world. My biggest understatement today is probably this is a developing situation, I'm sure you'll probably hear that from a number of people.
George Davis 5:20
So we're starting to see customers being impacted by it, where their supply chains are being disrupted. And so they're trying to figure out when do they want all their components to arrive, just to make sense with what they're seeing in the rest of their supply chain. You know, that being said, we're coming into this environment from a shortage perspective, and that was probably the negative side of 2019 for us from a performance standpoint - we weren't able to meet all of our customers demand and [that] actually impacted some of their outlooks as a result of that. But going into a situation where supply might be a little bit [of a] challenge, we find ourselves in the ability to operate on a relatively normal basis, I would say. However, you know, we're very, very mindful that the full impact of this could, you know, could expand much more significantly.
George Davis 6:22
We have a 24/7 pandemic team that's been working from the day this happened. You know, a lot of the tech industry went through SARS and other events. So they have actually pretty, pretty good playbooks. And we're certainly going through all the elements of our playbook and we're managing travel around the world. We have the ability to charter flights where we need to to get certain, as part of our supply chain to make sure that we're able to operate so I think there's a lot of things that we're certainly going through the the normal process of evaluating. But ot's, you know, right now for us, we're able to manage at a relatively normal, or on a relatively normal basis.
Joe Moore 7:10
You mentioned these shortages that you had in the fourth quarter that were fairly severe. I mean, I was expected to ease somewhat over the course of Q1 [2020]. [Are] customers in the position of being able to add inventory, up to the levels that they want to hold? Or is it still fairly tight.
George Davis 7:26
You know we go into this year fairly tight. Probably the only disappointment in Q4 was that in the beginning and middle of the quarter, we just weren't able to supply customers in the way that we had hoped to. And, you know, we continue to see capacity growing in our system. You know, our capital spending has increased significantly over the last couple of years, really to add the capacity necessary for our logic customers. We've actually decreased our NAND capital over that same period.
George Davis 8:01
So we have about 25% wafer starts up in 2019, we think will be up about 25% again in 2020. But it's that translates into a smaller number of die out growth because obviously, you know, with 10 nanometers you got, we're in the middle of 10 nanometer, we're still at a fairly large die size. Some of that, as we move to 10, we're absorbing some of that capacity increase. But all in all, we expect in 2020 to be fully recovered in terms of the ability to supply our customers to build inventory, we have not been able to build our own inventory, much less help customers, build their inventory, and we think both of those elements will be able to happen in 2020.
Joe Moore 8:48
Okay, great. So back to the bigger picture of issues facing Intel. So data centric, the kind of overriding philosophy of the company, I sort of made the argument that the company still fairly selective in terms of the types of opportunities that you're going after, maybe relative to the past. But can you just talk about data centric, what it means for Intel and what that is kind of a guiding vision, you know what that entails for you?
George Davis 9:09
Sure. So when we talk about data centric, what we're saying is, we see the company moving from a PC centric world, to a data centric world where more and more [of] the demand for raw compute power [is] particularly in areas like the cloud, on the edge, with the onset of Internet of Things. Bob likes to say, you know, everything's starting to look like a computer. You know, a factory looks a lot like a computer. When you look at what they're doing with robotics, what they're doing with basically the communication between all the devices and the factory. Your car is increasingly looking like a computer, not only from the infotainment but also to the ADAS and other elements of the car. So data centric is really about being positioned with products that are not only there to serve the PC world as we used to, but really this growing data management and increasingly, you know, the increasing need for storage. You know, certainly everything you do in AI is tied to this - 'hey, I'm getting a lot of data, what I'm going to do with it, how do I manage it in a way that I can get to the value in that data quickly'. [This is] opposed to just being able to just store more and more dumb data.
George Davis 10:40
So for us, data centric is sort of everything but the PC that serves those markets. Whether it's on the network side, IoT side, data center, all those elements are a very broad investment in AI. All that fits data centric. We used to be kind of a 70/30 PC [company] then 60/40 in the fourth quarter, we were actually slightly over 50% data centric [now]. So we continue to make progress there and we think that's also a very long, long outlook, part of the market and the growth there is been just terrific for us..
Joe Moore 11:22
And as part of the way you guys think about that you talked about, you know, wanting to be 30% market share and growing in a larger space than the 90% in a smaller space. And I guess I see the positives of that [to] begin looking at attacking new opportunities. But I guess it depends on questions that are raised is this, you know, are you willing to yield share within that area where you've been traditionally dominant?
George Davis 11:46
I think that's a fair question. The impulse is to say, 'hey, they're talking about this larger TAM, so they're going "hey, look over here, don't look over here"'. I think if you take that position, you're missing the bigger picture, which is being able to focus on a much broader TAM [for us] exercises different muscles. Then if you're focusing on preserving a very high market share in two core businesses - don't mistake me we're we take a very competitive view on maintaining market share and fighting for market share and our core businesses - but this data centric commentary that we're going to is real, the growth is real. You look at the strong double digit growth across all of our data centric businesses. Those are real opportunities. Those are real opportunities that come with attractive markets, margins, they're attractive markets and that the margins are strong and sustainable. You have the ability to differentiate yourself. So what we wanted to do is to get away from being kind of focused as a 90% market share player, where your whole focus is defending market share, to how do you compete more effectively? How do you develop the muscles to compete more effectively? How do you become more nimble and aggressive to enter new markets where you're not the large market shareholder but you have tremendous IP assets that should position you quite well.
George Davis 13:27
So for me, it's a very important distinction. It's one of the things that really attracted me to Intel was this remarkable set of opportunities that it has at this time, coupled with an extraordinary balance of IP to address that market. So if you saw the NYT this morning, there's really a nice piece on what Bob is doing to change the culture (Bob Swan, our CEO is going to change the culture as a company) to make it more focused on how do we execute and compete in a rapid, transparent one Intel way bringing all of our capabilities to tackle these new markets. That's a mindset that really comes from looking at these new markets and what it's going to take to win in these markets. You're seeing us take very strong positions in these new markets as they develop. It's really from that change in thinking of ourselves as a PC data center, you know, monolith to really this more agile competitive entity entering some new markets where, quite frankly, we're not the leader. And so we've got to come with solutions that drive customers to come to us.
Joe Moore 14:47
Great. That makes sense. I guess as you think about this wider scope for Intel, how does it inform your M&A (merger and acquisition) strategy? And I guess part of, you know, my concern is always I feel like the core business is inexpensive and maybe undervalued. You know, how do you measure accretion of something that you acquire versus the option of, you know, buying your own stock, which is fairly expensive?
George Davis 15:06
Okay. So, you know, as we think about M&A, it's always in the context of our total. Not only our capital structure philosophy, but you know, how do we think about investments in general. Obviously, you know, we have the good fortune of generating tremendous cash flow as a company. And so the question is, what do you do with that free cash flow, and for us, it's been, you know, [to] stay on the leading end of technology. So [a] very strong focus on R&D. As we have driven our costs down, and we'll talk about that later, but R&D has really been sustained at high and record levels throughout this whole time period, because we think that's still a very fundamental competitive strength of the company. So [we] invest in the company, [we] invest organically in the company, to the extent that we're entering these new markets.
George Davis 16:01
One of the things that happens as you enter a new market is [that] you don't have all the learnings. You don't have all the answers. You're not the expert in that space. Other people have experiences, very valuable [experiences], [that] can make you faster, and a better competitor. It doesn't have to be a large acquisition, it can actually be, you know, picking up the pieces that really makes a difference. So we'll continue to do those elements, and those are actually quite important.
George Davis 16:29
So M&A is in my mind really all about investing, still investing in the business. We think, even with the M&A that we're looking at, that supports that, we have tremendous cash flow that we can return to investors and you've seen that over the last 10 years. The company [has] returned 100%, we were at 113% of free capital last year, and we announced a $20 billion programme for share repurchase at the end of last year, of which we got about two done last year. And we said we have about another 13/14 months to get another $18 billion done. So 20 billion of share purchase, we grew the dividend by 5%. So M&A fits into the overall picture of how we think about managing our free cash flow. But, you know, we're not you know, anybody that's followed the either Bob or myself from an M&A standpoint, we're super focused on cash returns, we're not romantic about M&A. It's really got to drive the business. And we, you know, I think that'll be how we look at it going forward.
Joe Moore 17:49
So before we dive into some of the businesses, maybe a couple of other just bigger picture questions, about process technology. I know you're you guys aren't thrilled with the way you performed last year. So can you talk about the progress of 10 nanometer and 7 nanometer and your confidence level going forward?
George Davis 18:04
Sure. Well, we're definitely in the 10 nanometers era. We launched Ice Lake client at the end of last year. We have GPUs coming out - a discrete GPU coming out this year, we have networking ASIC, we have obviously, [at] the end of the year we have the server SKU coming on on 10 [nanometer]. Interestingly, and indicative of how we're approaching process technology going forward, we also have 10+ coming out this year.
George Davis 18:44
What we've said is, it's as important to launch a new node as it is to launch intra node improvement every year.
George Davis 18:55
We'll have we have our Tiger Lake client product coming out. It's going to be on our 10+ [nanometer] node. I've been told by our client team that I'm not allowed to talk about how much incremental performance is going to come out of that. But the idea is to have a step function move without having to wait for 7 [nanometer] in between nodes, and we'll be able to talk about that. It's been hard to find a conference [where] we've been able to talk about some of these things, which we were planning on doing, but we'll get the information out. They're not gonna let me talk about it first! It's got to be the right people, the people doing the hard work on the products to talk about it.
George Davis 19:41
But so, you know, I feel like we're in the 10 nanometer node. It's important that we're continuing to see yield improvements ratably over the over the time period. But as we said back in our Analyst Day in May of 2019 - this just isn't going to be the best node that Intel has ever had. It's going to be less productive than 14 [nanometer], less productive than 22 [nanometer]. But we're excited about the improvements that we're seeing. And we expect to start the 7 nanometer period, with a much better profile of performance over that, starting at the end of 2021.
Joe Moore 20:26
It seems like 10+ is really important, and we can't really get into the details of it. But it seems like the biggest thing for me is, from a planning process, to like, Intel's on a pretty firm foundation and you didn't, when 10 was delayed, it obviously means you don't have your best foot forward product wise either. It seems right now, at the very least, the next couple years, there's a very clear stepping stone roadmap for where we're going to go.
George Davis 20:46
Yeah, I think we feel very good about where the roadmap is going. I think one of the things that I think gets underestimated, and we're probably as much as at fault for this as anybody because we've always said that it's the CPU, speed and power is all that matters, and the fact of the matter is, you know, if you look at our products today, more and more, pick the data center as an example, more and more workloads are very customized to specific requirements with specific cloud players.
George Davis 21:22
So you have to pull together - it's not just a matter of the CPU that you have, it's all the other elements, from software, to whether you can bring AI into the equation, to you know, some of the architectural elements, [of] how you use memory in the device. All of these things are becoming super important. They allow you, even if you might not have the CPU positioning you would like at that particular time, you know, for instance the software on Xeon, coupled with our AI, we get like 10x the performance to the competition on AI related requirements, even with differences in the CPU. So we bring a lot of capability to the table for our customers, in addition to the CPU, and we feel like we're starting to see the acceleration on the process side that we have been talking about to get back to, you know, parity in the 7nm generation and regain leadership in the 5nm.
Joe Moore 22:30
And then, [for the] last 'bigger picture' question. As you think about all these process technology issues, one of the things from the analysts day was [that] it's going to have some negative impact on gross margin next year, because you're ramping 10nm, and you're also introducing 7nm. Can you just talk about, I know you're not going to get into it updating the three year roadmap, but you talked about that [and] the gross margin degradation is hard for people to get past.
George Davis 22:54
Sure. I [had] just joined [Intel] so my first presentation was saying that, you know, margins would be 57% in 2021. [They said] 'is that okay? Maybe a different CFO would be better' [laughs]. But the fact is I wanted to be clear what was happening during the 10 nanometer generation. The fact is, like I said, it isn't going to be as strong a node as people would expect from 14nm or what they'll see in 7nm. Also, you know, we were at a time when in order to regain process leadership, we had to accelerate the overlap between 10nm, 7nm, and then 7nm and 5nm.
George Davis 23:42
So the cost that you're absorbing, starting in particularly 2021, you've got this intersection of the performance of 10nm, the investment in 7nm, and we're also well into starting the investment of 5nm. All of those elements just combined to impact gross margin. And what we didn't want to do is just go out five years and say, 'hey, we're going to be an attractive margin pitcher five years out', and then the whole time people [would be] going 'well, I don't understand what's happening in 2020 and 2021 if this is where you say you're going to go'. It really is that curve we had to get through and we just felt was very important to just talk about what we saw happening. And, you know, we've been able to deliver a little bit better than what we've seen that what we thought going into that meeting so far. But still, the effect of 10nm in 2021 is just, it's sort of built today because you've got to get through that product cycle and the node. We're excited about the products but you know, the node isn't going to be quite the performer that historically we've had.
Joe Moore 24:56
I want to ask a little bit about just the shape of next year, or this year I should say, that you talked about in the last earnings call. So starting maybe with PCs, you know, [you've had a] very strong start to the year, then gear up to mid-point like around 10% year on year. You know, but you implied in your guidance of significant deceleration over the year. Can you talk about you know where that's coming from?
George Davis 25:20
What we said was that we are seeing in unseasonably strong first quarter. Part of that, I think, on the PC side certainly was that we were unable to fill all the demand that our customers had in the fourth quarter, and you were seeing demand continuing. But also, just in general, demand that had remains stronger than expected in PCs and we're starting 2020 with a overall demand level above expectations. But we know you know, the Windows 10 Refresh is, at some point, going to start to wind down. We think we're well into the refresh. So it's the expectation that some of the momentum from that will start to wane in the second half. We also have our modem business in our CCG business and we would expect that to decrease a little bit in the second half as we start to see competitive units come in.
George Davis 26:21
So the whole shape of that gave us a little different look than we would normally have. And, you know, the patterns have been, the seasonal patterns have been pretty strong. And, you know, you could kind of in many ways, forecast the full year based on what you would see in the first quarter. And this is just going to be a year that if you did that, you know it was would just give a much stronger outlook than we had.
George Davis 26:49
The other half of this then of course is data center, which again has been hot for three straight quarters and there's a very clear pattern that, you know, you get very strong buy and then you get a digestion period. Now, this is not a customer base that provides a lot of forecasts, because they're also you know, they're very demand sensitive too, and want to add capacity as it makes sense. So the ability to forecast when that nose is over the demand nose is over is, is quite hard, and not only for analysts, but also for us. So our view is it you could only have so many strong quarters in a row before its nose and often it looked like that would be in the second half. So we would have this odd, again, an odd pattern and we just wanted to call it out. Because it wouldn't be the natural expectation, otherwise.
Joe Moore 27:53
It's not like you're really seeing that digestion period coming so much. It's just more just pragmatically saying this is not a wrap.
George Davis 28:00
That's exactly it. It's our experience that, you know, once you get, you know, a number of strong quarters in a row it starts to nose off. But now that the pattern there has even been a little bit uncertain. You can have four quarter runs. They've had a, you know, three [quarter run], they've had a seven [quarter run]. So there's, there's, you know, it's not perfectly knowable, but again trying to just get the outlook was already higher than expectation for us. And we just wanted to have it, you know, make sure that people didn't get too far out based on something that looks pretty uncertain to us.
Joe Moore 28:42
That makes sense. So within the PC business over the course of this year, how are you thinking about market share? You know, your competitor has a good high end product, but you also have, you're coming out of an allocation situation over the course of the year. What do you think happens to your unit share over the year?
George Davis 28:58
So probably our biggest challenge in market share in PC has really been our ability to supply the market. So, yes, we have strong competition, but we have a client platform that has a superb track record for our customers and the performance of the whole system provides a very strong base for our market share. We have, you know, when we're short, we asked customers, you know, which products you want us to make for you guys, because they have a full range of products and they obviously they want to make sure they're full at the high end. Which means that we tend to miss the lower end of the market where we sell what we call our small poor products. So we've been unable to effectively compete in that space during this period of shortage because we just didn't have a supply. Again, [we're] working with customers to make sure they had their preferences first. So we think as supply normalises, we're going to be able to go back and compete and take some share back in that smaller core market. Yes, we'll have some, you know, some competitive dynamics in the higher end, but, you know, we feel good about being able to go back and get some share back there.
Joe Moore 30:27
Then probably the most asked question for Intel is same question but for servers. It seems like they're forecasting some share loss [for Intel] over the course of this year. You know, where do you see that going?
George Davis 30:38
Yeah, you know, I think we, we've said we expect to see stronger competitive dynamics in the second half of this year. But what we've seen, we actually thought we would see some of that a little bit sooner. And what we've seen is, again, very strong demand for our leading edge products, not only because many of those products are customised for a specific workloads, but you know, our AI capability is particularly strong. So we do expect there to continue to be strong demand for our products throughout this period. But we do see rising competition, we plan for that, as we talked about our forecast, that they'll be share impacts that we'll have to absorb within that. And then as we look at our product roadmap over time, we think again, we start to present an even more compelling competitive position as we go into 7nm and 5nm.
Joe Moore 31:43
Within data center, also you've you've looked at a lot of different ways of managing these new workloads. Other than CPUs and you've made acquisitions in graphics and FPGAs. You just bought Habana. Can you talk about that overall AI/machine learning and how much of it is CPU centric, and how much of it is driven by these new IPs?
George Davis 32:03
So AI is a very interesting potential and market. We think in 2024 it will be on the order of magnitude of a $25 billion market. We have close to $4 billion in AI related revenue today. I would say over half of that is related to our, our big Xeon processors, and then we have Mobileye's extraordinary success story there. The demand for their products keeps growing substantially. And then we have various ASICs and FPGA that support as well. We believe it's really early days in AI and there's not going to be one technology that is the winner. In fact, the proliferation of AI workloads is growing fairly rapidly. And so what we've seen is in order to engage with customers effectively, not only do you traditionally have to have the traditional way that people think about CPU and GPU is kind of the the foundational technologies but we're seeing, like with Havana, you know, the ASIC strategy where you can basically have a programmable ASIC really focused on delivering either inference type solutions or training solutions, but in a way that gives the customer more control over how they approach that. Then there's a number of GPU applications, [and] we also see the FPGA continuing to be an important device, particularly early on as customers are thinking how they want to approach the specific AI issues for their particular needs.
George Davis 34:24
As you think about AI to, I guess, if you thought about it only as a cloud activity, you would definitely be focused on the CPU and GPU. But more and more, it's moving out to the Edge. [Customers] are getting all this data and have got to figure out what's the data that [they're] going to need for [the] particular application and [they] don't want to bring that up into the cloud to do it, [they] need to do it more quickly, and screen it more effectively. So you're seeing applications like that, that just are proliferating, the types of solutions that you have to bring in. We think we're really well positioned across that array, and we're going to continue to invest in AI. We think that's an area where we can integrate solutions in a way that's quite attractive to customers.
Joe Moore 35:12
I did want to ask you about a couple of other parts of the portfolio. On NAND you've talked about the importance of Optane storage class memory. You guys have also acknowledged some of the challenges of the commodity side of that business. How are you thinking about that going forward? Obviously, the commodity environment is improving. But are you still trying to find ways to spend less in the commodity part of the market?
George Davis 35:33
Yeah, you know, one of the things that we've said for any of the big bets that we've taken and certainly memory and in particular NAND falls into the big bet category is that in addition to having differentiated technology and a product that will play a bigger part in our customer success, you know, we have to have profitability, long term profitability and attractive returns as a element of that. I would say, you know, for NAND, it really hits the first two very well, you know, NAND in the data center is just becoming a more and more important element. And so, but we haven't been able to generate the profits out of that to get the kind of returns that we would like to see. And in fact, obviously 2019 was a super difficult year.
George Davis 36:26
Oddly one of the tailwinds for 2020 is the improvement in memory year over year. And we think when we look at the demand picture, there's still critical shortages on NAND. So I think, you know, that's that's got a good tailwind with it. But we've said we're going to look at ways of improving profitability, not only in terms of how we manage the business every day, but also in looking at partnerships and other things where we can perhaps improve the overall economics of the investment. We're in the middle of that process, and we'll keep people up to date as we as we make progress. But you know, we're active and looking at those elements.
Joe Moore 37:09
Then especially for me, on the Mobileye business. [There was a] really good Mobileye analyst day in Jerusalem, and it seems clear that Mobileye as a company is hitting on all cylinders. How do you think about that opportunity? It seems like a lot of the opportunities are, are sort of less around semiconductors and more around systems and software, but it's obviously very valuable asset growing very quickly. How are you thinking about how that fits into the portfolio?
George Davis 37:31
Yeah, I mean there's a hardware and a software element in the EyeQ technology. But more importantly, we think, you know, the reason we have the analysts day is that I think people really weren't seeing, particularly with all the hype around ADAS in other parts of the technology world getting a lot of attention, Mobileye in the meantime was basically taking more and more share and becoming the standard for ADAS in the automotive industry. So [we have a] very strong position, [with] growing share, growing technology leadership, and the gapping out [of] the competition in the eyes of our OEMs. So we had the analysts day to get people out to 'hey, let's show you what we can do', which if you've never taken a ride and an autonomous vehicle in Jerusalem in rush hour traffic, you spend the first part of it in care because you can't believe this car is can be able to navigate these cars coming at you from all angles - nobody seems to follow any of the [road] rules. You've got to merge into traffic and nobody leaves a lane for you and it's just spectacular to see this car drive, as my friends in Israel say, drives like an Israeli! So the first 15 or 20 minutes of it, you're, you know, kind of in horror, and then the last 10 minutes, as one of the analysts said to me, 'you know I was kind of bored at the last 10 minutes', which is the highest compliment I can pay this technology. So they've really come, you know, a very long way. But that's really in the traditional markets. They've also looked at, because they have these vehicles on the road with this capability, they're getting tremendous amount of data, and so they're also a huge user of AI to process data. They've come up with mapping capabilities that [act like] kind of a feedback loop to the car itself as it's driving, so it improves the quality of the driving experience because it's got its own set of real time maps that it's working off of at the same time that it's got all the other measurement elements.
George Davis 40:01
Plus, we're finding that you can sell the data on these maps because we can tell, we have a case in the UK where utility holders are having trouble actually finding where they need to make repairs, but if you have all the mapping data, you can actually tell where utility lines are running where pipes come underneath. And so we're there's businesses like this popping up all over where it's for us, it's, it's just using the data that we already have. For them, it's remarkable timing and cost saver. We also know where every pothole is effectively in the areas and where every pothole is forming, and so cities are now seeing this as an opportunity to buy data that will allow them to go get.
George Davis 40:49
So there's the near term system stuff and then as you look at mobility as a service, where the driverless cars, not only will that be incredibly important, but our technology and our investments in the stack all the way up the stack and mobility as a service, are another area where we think we're going to be highly differentiated as we move to the robo taxi world, which will be the world that you have to go through before you actually have fully autonomous cars. Why is that? Because Robo taxis are what cities are going to use to determine whether autonomous cars are actually really safe, they can really control [and] understand what parameters are important. We're going to be we're going to be the leader in supporting that - we've already announced some partnerships in that regard already. So Mobileye is probably the best proof case for AI, and we're able to help them both from a software and an engineering standpoint, [to] keep this incredible pace that they're going. But it's a really remarkable team and we feel great about the investment.
Joe Moore 41:58
Okay, great job. Unfortunately we have to wrap up there.
92 Comments
View All Comments
name99 - Wednesday, March 4, 2020 - link
👍👏milkywayer - Saturday, March 7, 2020 - link
Appropriate joke on the confusing af xLake naming there. Intel needs to fire whatever monkey(s) are still pushing these - lake names. No sane person can jeep track of them anymore considering they're churning out a new lake every 6 months. At this point it'll be mentally better for all parties involved to just reset the naming and stick to simple numbers like the olden days 6700, 6750k, 6800.Spunjji - Friday, March 13, 2020 - link
🤣👏JayNor - Wednesday, March 4, 2020 - link
Likely 10nm was necessary to squeeze in the big gen11 and gen12 integrated GPUs. The backports rumored for Rocket Lake are reduced execution units vs the 10nm.It appears the pcie4 support required 10nm, since it was removed from the Cooper Lake design... maybe a thermal issue. I'd guess PCIE5/CXL requires 10nm for the same reason.
Intel appears to have made some nice upgrades to Agilex with 10nm.
Intel's P5900 family is 10nm, and supports extended temperature range -40 to 85c. I wonder if the power efficiency could have been met for that family without 10nm.
So, while their yields may not be great, I'm guessing Intel needs 10nm for the new features.
I suspect we'll see EUV brought in to improve 10nm yields, in addition to its use in Intel's 7nm.
Guspaz - Wednesday, March 4, 2020 - link
AMD: *laughs in TSMC*edzieba - Wednesday, March 4, 2020 - link
No $#!& Sherlock, price/transistor and price/unit area have been rising since 22nm. That goes for everyone, not just Intel: nanoscale semiconductor physics is a bitch. More radical changes than just switching to EUV or using Cobalt metal layers are going to be needed, and GAAFETs and TFETs are only going to make the problem worse.Ian Cutress - Wednesday, March 4, 2020 - link
He's not talking per wafer here, he's talking revenue and gross over the lifetime of the node.eva02langley - Wednesday, March 4, 2020 - link
Elementary Watson when 14nm lasted... 6 years?name99 - Wednesday, March 4, 2020 - link
Oh, snap!edzieba - Thursday, March 5, 2020 - link
The latter follows the former when per-chip prices need to remain stable or minimally increase to meet market demand.