Intel Corporation (INTC) Earnings Call Transcript & Summary

Okay. Let's go ahead and get started. Okay, great. All right. Good afternoon, and thank you for attending our 18th Annual Technology Investor Forum at CES. Very pleased to have Tom Lantzsch, Senior Vice President and General Manager of Intel's Internet of Things Group or IOTG, particularly relevant given the focus here at CES of more compute intelligence, more connectivity in areas like retail, manufacturing, transportation, industrial. I've asked Tom to start us off with a description of his role and responsibilities at Intel and provide a brief background of the Internet of Things Group. But before I turn it over to Tom, let me just read their safe harbor statement. As a reminder, today's discussion may contain forward-looking statements, which are subject to risks and uncertainties. Please refer to Intel's SEC filings for risk factors that could cause actual results to differ materially. So with that, Tom, thank you for joining us. Let me turn it over to you.

Great. Thank you, Harlan, and thank you, everyone, for making the time to spend with us today. Yes, let me give you a little perspective about this because it's probably the most unknown business at Intel, although we're pretty proud of what we're doing over here. I joined Intel about 3 years ago. Prior to joining Intel, I had spent 10 years at ARM. So I think I have pretty good exposure to the entire semiconductor industry in that, and I've been in this business now for, hate to say it, but I think it's my fourth decade. So it's sort of scary. But about 2.5 years ago, we embarked on what we thought was the right strategy for Intel in this IoT space. And at the foundation of that strategy and taking a look at the entire semiconductor world, we believe that our focus needed to be, first of all, on businesses, enterprises, schools, governments, nonconsumers. So that's the first point. As part of that, to deal with what we thought were the compelling events and the compelling applications, we believe that where the data was created from all these connected things, that a solution wasn't only going to be all this data was going to get sent to the cloud. And there are a lot of reasons for that. And you guys have heard about them, latency, the amount of data created, blah, blah, blah, lots of reasons. And so we embarked on a fairly aggressive strategy and focused our efforts on, what we called at that time, consolidation of workloads at the edge of the network. We weren't talking about edge computing then, but it was really about this notion of consolidation of applications closer to where the data was created. And we leveraged a lot of the traditional software resources that were historically in the cloud and brought it down closer to where the data was created. So that, that is at the foundation of our strategy. And we believe that, that required us to do a couple of other things to support that. One was, we now needed to have differentiated silicon to enable that. So no longer was it purely we were taking silicon from businesses like our Client Group or our Data Center Group, but we were also beginning to do differentiated silicon to enable those applications, both at the CPU center and also at the accelerator side of the business. So we embarked and started to do differentiated silicon. That will become more and more apparent. This year actually, you'll see -- we announced our first product -- I'll talk about it briefly in a few minutes -- at the end of last year. The other thing we looked at was, what's the killer application that's going to drive these applications over across this very fragmented vertical space because as you guys know, this IoT space is made up of hundreds of different companies and applications. And so it's how do you deal with this fragmentation issue. And we thought that the killer app to really be the impetus for this change for these customers across these verticals was computer vision applications. Adding cameras changed everything. And it didn't matter if it was in the retail space. It didn't matter if it was in the smart city space, didn't matter if it's manufacturing space, didn't matter if it's education. You name it, cameras, integrating cameras with inference technology at the edge was sort of the killer app. So we doubled down heavily there. So that was sort of the killer app. And then 2 other pieces of supporting technology or activities that we did was we needed to fundamentally change the developer environment around enabling these applications. And so there, we invested pretty heavily initially in computer vision technology applications to enable a developer community around that. We announced this developer environment called OpenVINO. Really it was about democratizing AI and inference technology at the edge of the network and really enabling that for our customers to use AI in these various computer vision applications. Because you can well imagine, you pick whoever you want, your favorite retailer, your favorite fast food restaurant, your favorite manufacturing location, very difficult for our customers to get access to these developers. And so we needed to create this environment to that. And as an outcome of that, we created this thing called OpenVINO, which was this developer community. It actually makes it very easy for customers to develop applications, AI-based computer vision applications. And write the technology once and it determines which hardware it runs best on. It could run on a CPU. It can run on a GPU. It can run on FPGA. It can run on a neural net. So we did that. Long story short on that, we announced it in June, we're adding about 10,000 developers a month. So I mean, it's going great. We're over 150,000 developers worldwide and continuing to grow. And then last but not least, our customers, our end customers needed solution providers, ISVs to help them with their case and solve their problems. And so we created an activity. We call it Market Ready Solutions. It's effectively ISVs around the globe focused on very specific verticals to create applications up on top of these solutions that we've done. And just to put numbers into it, we had 1,000 different applications from ISVs this year to try to be a part of this program. We picked 300. And we scaled out with our end customers, over 10,000 of them year 1. So we got a lot of traction across the board in this. So that's a little bit about the business. A lot of things that again most people don't realize, our average ASP on our products is over $100. We're not cheap and cheerful guys. We're high-performance compute guys.

Great. That was a great snapshot. IOTG is on a $4 billion kind of annualized revenue run rate. It's been growing at a double-digit percentage year-over-year growth clip. Can you shed some light on the sustainability of this growth and what the greatest revenue drivers are for your businesses? And what have been some of the key end markets and applications that have driven this growth?

Yes. Like I said, I think the catalyst of vision systems has been like the application that spans across all of these verticals. I can honestly say, every one of our verticals are growing pretty fast. I mean, they're all growing double digits. I inherited a core business of stuff that clearly is not growing that, but it's traditional, I would call it, traditional embedded compute. But that even, that business is pretty nice. It's grown probably in the high single-digit growth rate. So it's geographically dispersed. Our business is geographically dispersed but everywhere from education to health care to factory automation, it's fairly distributed. I can't say that there's one that's more prevalent than the other. They're all growing fairly strong. The other indication on the sustainability is, as you said, we passed our first $1 billion quarter in the third quarter, which is nice. Obviously, we're pretty proud of that. The question that I asked my team is not that we passed $1 billion quarter, but how fast were we going when we passed it. What's the speed passed it? And we have an activity that we track as most companies on design wins. Through the first 3 quarters of this year, we had more design wins than we did all of last year. So it gives me confidence that we can continue this growth at least for the foreseeable future. Our challenge has been actually this year is, as you know and it's been well-publicized is, unfortunately, we haven't had enough product for our customers, which has been a little bit of a struggle through the first 3 quarters of the year. So that's been a little bit of an inhibitor, even though we've had incredible growth rates year-to-date.

Are customers getting what they need now from you?

I just met with some today. So far, so good, but it's too hard, right? It's hard. It's painful for our sales team. Certainly, we don't like putting our customers in a situation. But so far, we've been able to scrape, claw.

A follow-up question on the growth and the sustainability of the growth. How does Intel's approach for IoT differ from that but the traditional PC, data center markets? You mentioned vertical approach, partnering with market leaders, differentiation through silicon, market ready systems and system-level support. Can you just provide some examples?

Yes. I think maybe what makes it, probably what makes us maybe most unique is the fragmentation of the space versus all the rest of the colleagues of my -- I mean, maybe take our PSG or Altera, what was our Altera business, let's put that aside for a second. With the exception of that business within Intel, our business is really a big channel play. And we've got really strong channel partners around the world that we have developed over the years, both at multiple layers of the value chain, ODMs, distributors, ISVs. What I think we've done a really good job of and one of the reasons that we have to have this vertical focus is that the end customer, the real consumer of this technology, each of these vertical solutions that we go to, to the market, the ISVs are very typically very -- they're not large. They're smaller-sized firms. They're focused on a specific vertical. And they're geographically oriented. So you may have a great manufacturing ISV partner in Western Europe, but it's totally irrelevant to us in Japan. So I think one of the secret weapons we have is our brand and our position globally and the strong channel that we've been able to establish over the years. We just had an event in India, okay? India, I'll pick India. We just had an event and 1,200 customers showed up, 1,200 customers and partners showed up to an event. I mean, it's the kind of scale that we can bring to the partner. I had a similar one in China this year, hundreds, right? We did a webcast, 18,000 people, 18,000 people logged into a 1-day webcast when we were talking about AI and at the edge. I mean, so it's a great advantage, frankly, that we have well and beyond the technology that, obviously, we love and create every day.

Intel, to my knowledge, started talking about edge computing a while back. And more and more of your compute peers are talking about edge computing, but I think you guys defined it about 2 or 3 years ago. Can you tell us about Intel's journey to the edge and how does Intel define edge?

Yes. Well, we sort of define it as not things, right? I mean, we do some. I do some silicon for what I'd call things. A camera would be a thing in my world.

And endpoints. Endpoints, yes.

Yes. A camera would be a thing. I sort of like things that think are sort of interesting. Things that are smart are not so interesting. I think other people can do that better. So we'll do neural network-based technologies on smart cameras because we believe there's a natural loop with the next connection. So where we define the edge is in 2 different dimensions. There's a on-premise edge, which is predominantly where I focus, which is really focused on vertically-specific use cases, factories, could be electrical grids, stores, whatever. That's one activity, which is all these things get connected and then this processing happens. There's another network edge that I collaborate with, which is in our different group. It's our networking group that mostly serves the carriers. That's another network edge, CDN-like technology. We don't define it as far as things. I think it's a thing, not really our sweet spot. We do some things, but that's not really our sweet spot.

So I'm a visually sort of oriented guy. So when I'm thinking about a smart factory floor, where do I see your products?

So you'd see, here would be an example of areas where we would participate and things that we're doing differently. So if you take a look at a discrete manufacturer, a car manufacturing facility and one of their manufacturing modules. Traditionally, they'll have several different controllers running that module. They'll have a programmable, a PLC system. Those PLCs are usually cards, usually have MCUs on. Here's what we're doing differently. We're virtualizing all those PLCs. So what was a microcontroller business, I virtualize them in software and put them on a Xeon server. Why do customers want that? Why is that important? Because they can upgrade these things on the fly. They can have redundancy on the fly. They don't have to have cards to go manage. They can remotely manage these activities. So that would be like one module. But they'd also then have a camera system to do computer inspection. Today, most of that's in a separate box. It's a separate box that they have to manage. It's a separate application that they have to manage. And so what we're doing now is we just make that application and put it on the same basic platform. So if I had to explain this to, I'll say my mother for simplistic words. It was sort of like in 2007 before we all had one of these things. We had a GPS module. We had an MP3 player. We had a camera. We had a phone. We had all these things. And all we did is put them together and made interesting. That's sort of what we're doing. And as these sensors get connected in these various locations, we're putting that all together and increasing the amount of compute and then only sending a subset of that data to the cloud because just like -- I'll use a car as an example, our Mobileye team, just to talk about the amount of data that's getting created. For every kilometer a single camera Mobileye system gets driven, it creates about 4.5 gigabytes worth of data for every kilometer. You can't send all that to the cloud. It's just impossible. And so what we do in our mapping applications that they do is we actually inference that technology. We pick up things like pedestrians and bikes and traffic signs and all sorts of information. We only send 10 kilobytes to the cloud, a 400,000:1 compression of data created versus data that gets sent to the cloud. And the same thing is happening in the industrial space. You just -- they can't handle this much data. They don't need to handle it.

Right. There's someone here.

Go for it.

Wait for the microphone.

Appreciate the time. I'm a little out of my depth here but...

Go for it.

The processing that's happening at the edge when you talked about the incredible amount of data that's being created and the difficulty of managing that. Are you supporting the various big data frameworks at the edge and that's really where the value proposition is, is deploying the chip low power and then supporting TensorFlow or whatever the case might be to do analytics at the point of ingestion?

Yes.

So you still need those frameworks for AI, analytics, data ingestion and analytics on top of the architecture?

So yes, for inference or computer vision inference or inference-based applications like you're talking about, be it TensorFlow, be it PyTorch, be it whatever. Yes. So this OpenVINO development framework basically, what that does is it takes those frameworks and models that were developed for those frameworks. It puts them into a format for our end customers that then when they go to deploy this, depending on the hardware that's available to them, they only have to write once and depending on the hardware that's available to them because a lot of these applications that we're dealing with are constrained. I mean, there isn't an infinite amount of power in some cases. There's not an infinite amount of bandwidth. There may be a rough environment temperature-wise. So there's constrained compute versus like go to the cloud and just spin up another instance. So the answer to the question is, yes, those inference algorithms are all native frameworks, are all running locally at the edge to go do this inference. The training today predominantly happens still in the cloud. Our customers train those inference in the cloud. So that doesn't go away. But yes, the inference actually happens at the edge. So the answer is yes.

That ships with those frameworks?

Yes. And we can do it, like I said, we'll do that on a CPU with accelerators. We'll do that with a GPU if there's a GPU accessible. Sometimes we may use FPGA accelerators. We have another technology that we do a lot of our own, another acceleration technology. There's a company we bought several years ago called Movidius that we use Movidius acceleration, which is neural net technology. So it sort of depends on what the customer is trying to do, and it could be as simple as a PCIe card that's slapped into that same system that the customer then can optimize their solutions.

It seems like this AI at the edge is going to become increasingly more important just because you have this, whether it's factory floor, retail floor, smart building, whatever, you just get inundated with all this data that has to be analyzed and decisions need to be made in real time. And so as you think about your design win pipeline, your engagements, what percentage of your future customer engagements, deployments are going to be at the edge are going to be sort of AI-focused?

I think we will -- 2, 3 years from now, we won't even talk about AI-focused or not AI-focused, they'll all just be part of it. I mean, if you -- I don't know if you went to the press briefing yesterday, we're integrating AI technology in the new CPUs for even laptops. And if you saw the application where we showed how Adobe uses it to do applications and content creation. AI will just be like it's air, right? It's going to be just everywhere. And so I don't think we'll be talking about it much because it's going to be everywhere. That's my take. So big. So the answer is 3 years from now, I don't know, 100%, but vast majority.

So that leads into the next sort of discussion, which is software and firmware become extremely important in this kind of scenario, right? And so how important is it? How does Intel capture mind share with developers and engineers working on these embedded applications, these AI-focused capabilities? Everybody's working with different software architectures. How is Intel helping your partners sort of bring all this together?

Yes. Like I said, I think we're trying to enable sort of a next generation of developers. If you are, I'm going to just pick an age group only because I don't know how else to say it. But if you're in your mid-20s and you want to be a developer, you have a specific developer mindset, which is probably not like the traditional developers that worked on embedded factory automation systems, which are probably my age. So we're trying to modernize that entire development framework and make it very almost cloud-like on how or mobile-like centered on the developer community. Again, we started with computer vision activity because we thought that was the first one. And we sort of, we're internally talking about it as democratizing these applications. We need to make it -- democratize this application and organizing. So it's this combination of democratizing and organizing. We don't create frameworks. Other people create frameworks. We just enable those frameworks to be used. We aren't the best AI, I mean, data scientists. We enable data scientists. And more importantly, we try to enable people that aren't data scientists to be data scientists because there's a finite number. And the appetite for applications is infinite across these various markets. So we want to just create more and democratize more. And I think our scale does that. And just as an example of sort of the progress we've made on that. There's a online training company called Udacity. They do a lot of online training. They're well-known. Microsoft will use them. Amazon uses them. They're well-known. We offered a scholarship for training of AI inference at the edge. 15,000 people signed up in 48 hours. 30,000 people signed up in 2 weeks. It was the #1 training class ever for Udacity versus any application that they've ever done. So the appetite for this stuff is insatiable. And so we're just fueling into enabling this next generation of developers to go do cool stuff. And again, we're Intel. We're fortunate we can recruit some really good interns. And so I get the opportunity to meet with them and some of this bright talent from universities and colleges around the world. And I ask them, do you like working on industrial automation stuff because you would think that they wouldn't. And actually, they love it if you give them the right tools. The problem is people don't give them the right tools and they hate it. But if we give them tools that they're used to, to do cool development and it's more like they're used to do in development and not having to go run around and chase boards, but actually just go to the website and spin stuff up and get instances that they need and do evaluation and provide the right kind of development environments, they love doing those stuff. And so that's really our quest is to just create more and more developers in this democratized world globally.

In addition to AI, the other strong technology adoption curve will be 5G. So how does the 5G impact the direction of your business going forward?

Yes. So if you look at this business probably 10 years from now, I think one of the biggest verticals will certainly be industrial and manufacturing. It's just a great opportunity. In the short term, probably you got to look at this thing over time horizons. But 5G in the short term, probably the biggest impact it's going to make is in -- actually in manufacturing in doing, enabling private networks in manufacturing systems, largely due to the latency capabilities as a lot of these advanced factories want to get to more and more autonomous-based systems. And from traditional factory lines to more 2D manufacturing, where you have AGVs moving around. 5G is great for this because of its, first of all, its reliability, its manageability and also its latency characteristics. So it really enables for these next-generation applications. And I think you'll see more and more of that in the short term and then we'll go from there. But in the short term, that's really where our focus is.

At the silicon level, you mentioned in your opening remarks, the IoT Group will now have to provide -- would now have the capability to provide differentiated silicon to address customer requirements in areas such as acceleration technology, I/O and so on. Can you give us some examples of where Intel has actually delivered more custom-focused silicon to customers?

Yes. So again, we started this a couple of years ago. So it takes a while, unfortunately, for us to get these chips out and show them to everybody, but we announced our first one in November called Keem Bay. It's an accelerator technology. It's a neural net technology. Just to put context, it competes with who you would think that talks a lot about using GPUs for this space. It's 4x the performance of an NVIDIA TX2. It's at par of the performance of the Xavier of 1/5 the power. So that's the first accelerator chip that we've announced. And again, it's all supported by the same software framework called OpenVINO. Again, in this AI world, software is really the big issue. This stuff is hard. And we've got a lot of people working on compiler technologies to make it easier for customers to deploy. So that would be the first example. There'll be more. Tiger Lake was announced as a product yesterday. Well, it was announced as a PC product. There will be a Tiger Lake IOTG product coming down the path. And a lot of that -- we baked in early on into that product a lot of really good features like time-sensitive compute, functional safety capabilities. It's going to be really important for autonomous systems. Neural net, I mean, some acceleration technology like you saw to do more AI. So you'll see more and more of these differentiated products as we go through 2020. We got a couple of them that we'll be announcing. So it's happening at both the accelerator and the IA level.

Okay. And then my last question from a financial perspective. The team is driving high 20s, low 30% operating profitability over the past several years. Given the growth opportunities ahead of you, I assume that near to midterm the team is going to be prioritizing R&D spending just given the big opportunities in front of you. But how should we think about the segment operating profitability profile over the next several years if you continue to drive strong mix shift, you continue to drive double-digits top line growth? Is there the opportunity to drive some OpEx leverage and expand operating margins?

Yes. So our strategy is, if you just sort of look at this, we sort of look at the competitive space in our market. And our goal is basically to grow at least twice as fast as the market. And we're pretty happy where the operating margin is. It was, let's call it, 30%. It's in that range right now. 3 years ago, when I joined, it was 22%. I thought that was too low. I think versus our competitive space, we've now got it at 30%. I think I'm pretty confident with that. And if we can continue to grow twice the rate of the market and hold it at a 30% op margin, I think it's a pretty attractive investment portfolio business for everybody. And so far, we've been able to do that over the last 3 years, and I don't see any reason why we shouldn't continue looking like that. So I think the answer to your question is sort of flatline a little bit on the op margin and more emphasis on growth.

Yes. Great. Well, we're just about out of time. Tom, thank you very much for joining us. Really appreciate it.

You're welcome.

Very insightful.

Thank you.

Thank you.

Thanks for the opportunity.