American Tower Corporation (AMT) Earnings Call Transcript & Summary

Good morning. I'm Spencer Kurn, and I cover U.S. communications infrastructure at New Street Research. I'm joined here by Ed Knapp, the Chief Technology Officer of American Tower. Many of you may remember Ed from our conference last year. But as a CTO of AMT, Ed is responsible for positioning American Tower for growth as infrastructure needs for wireless communications continuously evolve. Before we get started, I just want to mention, if anybody in the audience would like to ask a question, you can type it into the top -- into the box at the top left of your screen.

So Ed, at our conference a year ago, we talked about your expectations for how infrastructure will evolve to support 5G. I think a good place to start this year would be an update on your view of how 5G will drive network transformation. And really, it'd be interesting to focus in on how your perspective has changed over the past year.

Yes. Thanks, Spencer. It's great to be here at New Street. I wish we were physical in New York City, but this will do for now, at least, for the last 8 months as we deal with the current situation. So regarding infrastructure, it's actually a pretty exciting time relative to this transformation to 5G. And some of the macro trends, as you would imagine, are really around things like cloud-native networking and being able to transform the network to a software-driven model. We also see a lot of convergence opportunities. If you think about wireline and wireless networks coming together as the networks become more seamless in terms of disaggregating hardware and software and turning them into more of an automated environment. The other parts that we're looking at are really across the radio network and how that's transforming with O-RAN. We're looking at ways in which the transport networks have to be modified to support the types of data rates, whether you're looking at front-haul, mid-haul and even backhaul-type deployments. Those are really transforming ultimately where we go with the edge of the network. And it's really about this first mile because a lot of the 5G is taking data that's normally been sort of Southbound from the cloud to the user, and we're starting to see a lot more endpoint generated with respect to IoT and some of the other newer applications that are coming. So transport is changing. Obviously, the access networks in terms of fiber and also the radio and then edge compute are really key infrastructure trends that are really 5G is going to help drive. And we believe, as a neutral host supplier of infrastructure to create those efficiencies, we're at the center of that, and we've been doing a lot of work in our innovation program to look at those business models and try to extend our core platform of macro towers.

Right. And so to dive into some of what you just touched on, a lot of the network transformation, it sounds like it's taking place not necessarily on top of the tower where AMT has really excelled over the past couple of decades. But as part of your innovation program, what are the main opportunities that you're exploring for sort of broadening AMT's core area of expertise?

Yes. So what we're looking at is really 4 elements to our program, and really, we look at it as extending our core tower business, which is still 97% of our revenue today. And when we look at it, there's 4 areas: the power and the ability to look at sustainability, renewables and shared power. So just powering the sites is something that we're doing some really, I would say, advanced work in our teams in Africa, where we're doing a lot of data analytics. We're moving away from unreliable grid and into sustainable models of lithium-ion storage and solar. So we think even that trend could move to other parts of the world, depending on the models for how we interact with our operators. So that's a great opportunity for transformation and growth. The other part of it is in the transport side, we look at transport in 2 ways. We look at it as connecting, let's say, towers or small cells or even edge data centers. So that's kind of just point-to-point type networking. But there's also other businesses that we have in Latin America, and those are looking at open network architectures for fiber-to-the-home or fiber-to-the-neighborhood. So those things are really also changing the landscape of how one can put in infrastructure that's open neutral host and create new value and new business models. And then furthermore, that will be used to support small cells and fiber-to-the tower expansions as we've done in those markets today. The other part of it is edge compute. So when you look at, ultimately, the network will have to transform from what is essentially a packet core that's centralized at a region and to get low latency applications, we need to start to migrate processing and applications closer to the edge, and our towers are best positioned to support that as aggregation points naturally for edge processing. And also on-prem, if you look at venues and other types of in-building systems, those are also opportunities for low latency application processing. So transport through the fiber networks, compute through edge processing, power is also the other key element. And the fourth one is just wireless connectivity. As we look at neutral host and share networks, some of which are really based on CBRS in the U.S., location spectrum in Germany, and how those inside networks become more of a NaaS, a network-as-a-service opportunity, and that also extends into other markets like in ITS band for CBX in automotive, which also ties back into edge compute. It all really stems from -- we can offer a more efficient infrastructure model to our tenants or customers because of the efficiencies of sharing as opposed to doing things and self-performing vertically.

And so you mentioned, I think, the tower -- towers account for about 97% of your revenue today. As you look across these 4 innovation opportunities that you just highlighted, collectively, how big do you think that could be over the next 5 to 10 years? And -- from where you sit today, which of those opportunities seems like it could contribute the greatest amount of revenue?

So all of them, today, we're seeing sort of, obviously, AFFO growth and margin benefits from investing in power and being able to see that capital investment perform reasonably well today. Now we would like to transform that model and extend it to other markets, but we're still early days in being able to look at how do you do that in a reliable grid environment, but that could be significant, and it also could grow into -- you think about Tesla Powerwall in home, imagine our distributed real estate could also perform, in some point in the future, distributed storage for energy, especially as you move to renewables. The -- let's say, the source of wind and solar are unreliable. And as utilities move in that direction, they're also going to need to look at massive storage and distributed storage, and so that's an opportunity we're looking at and doing some research with MIT on. The TAM for that could be significant, but you can't put a number on it today because there's a lot of regulatory and other transformations that have to occur, but I think it's an important thing to keep an eye on from a distributed real estate perspective. Another area in terms of transport is if you look at fiber to the home and how those networks are converging, wireless operators want to do broadband sales potentially as a quad play, so they may have just spectrum but not infrastructure. And other types of telcos would like to be working as MVNOs in a 5G slice network on top of wireless spectrum, so that convergence leads to an opportunity of saying, can you take existing, let's say, vertically owned fiber networks and open them up to multi-tenant and virtual network operations. So that would be based on a homes passed model. So we have some deployments in South Africa and in Latin America, where we're working with partners to be able to share that same network infrastructure, and it can also support small cells and towers. So that's an opportunity. It's a fairly significant TAM because you're looking at homes passed and homes connected and being able to get a wholesale component of that with a shared infrastructure. The other part of it, we're looking at, obviously, is edge data. There's a lot of numbers on edge data. It depends on how you partition it. But those numbers -- they're all sort of $1 billion-plus opportunities in TAM and, in some cases, more significant depending on how many regions in the world where we operate, we can take advantage of it. But in the U.S., you're looking at single-digit billions, at least, for the edge part, and you're not looking at all of the staff. You're just looking at basically the infrastructure and the ability to support space and power and networking to a lot of sites that are needed to be distributed throughout the market. Relative to in-building systems, we had DAS -- we have DAS today, but I think the market is expanding as operators would look for ways to enter other buildings and properties that normally would have been tried to cover from the outside. But as we move up in frequency in 5G, we can start to penetrate those buildings from inside and be able to share those networks because they naturally want to be neutral host shared networks. You don't want to have 2 or 3 networks in the same property. So we see a lot of opportunity in commercial real estate, in different types of verticals, hospitality and so forth, where DAS systems were, maybe historically, too expensive and that becomes an opportunity for in-building systems, and that TAM can be significant as well. And a lot of people are looking at that today in the U.S. with CBRS and in Germany with location inspection.

And just to follow up on a couple of things you mentioned. So what was somewhat new to me was that you're building or you're exploring the opportunity for neutral host, fiber-to-the-home models in some markets. Is that something that you see an opportunity to do in the U.S. or in other developed markets?

I think it depends. Each market is different in how the industry has evolved. A lot of it comes down to is there overbuild, there's dominant players in the U.S. where cable really has the largest foothold, and most of the telcos have tried to and have invested themselves. So unless that market changes, it may not be attractive. But in other markets where folks are more open to sharing, it does become an opportunity for -- we're looking at really 2 things. ROIC, can we get the ROICs that are extending from our tower environment to other adjacencies? And is AFFO growths going to be a part of that because you're looking at things like maintenance CapEx and a lot of other SG&A requirements? So we see opportunities there. It's early days. This is just exploratory in some sense relative to our core business. But as part of our program of looking for ways to innovate and potentially deploy capital in the future, these are all aspects of what we wake up every day and think about, at least from my part of the business, which is innovation.

Got it. And then another -- just to follow up on another one of your comments. I think you said the TAM for edge data centers in the U.S. was about somewhere in the single-digits-billions range. Could you just unpack what you're including in that TAM? Is that sort of as a tower company become a potential landlord for edge colocation facilities? Or is that what you see as the overall TAM for you to fully own and operate edge data centers at your towers?

So the way we look at it is there's like 4 buckets of the edge, right? There's the device itself. You can call that automotive platform or maybe even a drone. There's inside a venue, right, which is something we can support today with local breakout. And then you get into the tower edge and then the metro edge, right? So when we're looking at that number, we're looking at the tower edge and the metro edge. Tower being an aggregation point of some scale, right, so not just a single tower, but in the limit, you can think of every tower of having essentially some local processing to pull out data, like, for example, on a rural highway, where we're doing truck platooning or automotive applications, and you want a mobile edge compute for lower latency processing. You can envision almost every site having some form of compute based on, again, the transport architecture and whether or not it's appropriate to support bits back to some other locations so then you start building horizontal networks as opposed to just North-South networks. But I would say, on the order of hundreds or thousands of locations to really take the cloud to the edge and to work in connection with the mobile operators to support that, that there will be those opportunities for, let's say, 100-kilowatt to 250-kilowatt centers at tower sites or in clusters. And then, ultimately, when you look at the metro edge today, it's really where things are happening. So if you look at some of the early deployments in 5G on the edge, they're really at regional or metro level. So you're talking about millisecond levels that are 20, maybe plus 30 milliseconds, not that different than what you could see today in 4G if you were really at those positions. Now the 5G AirLink improves that, right? But at the end of the day, to get down to that sub-10 or 5 or even single-digit milliseconds for the future applications in AR/VR, haptics, maybe other types of super-low latency for automotive, those types of requirements push you out to the tower edge. So the several billion dollars of TAM split between the tower edge and the metro edge, depending on application, and when we get 5G networks fully ubiquitous and built out in mid-band, and you have the ability to have super-low latency on the AirLink, then those applications will drive closer and closer to the edge, and there will be ubiquitous performance capabilities. Otherwise, you end up with kind of erratic or, let's say, a distribution of latencies. And we really want to have a tightly bound and control latencies to have uniform performance and user experience. And the operators are all working towards transforming their architectures to support that. So the TAM consists of those 2 components, the power edge and it's really infrastructure at that point. I'm not talking about server hardware. I'm not talking about software stacks or SaaS applications. That could really explode the numbers a lot more in terms of where cloud is today and where the growth is in CAGR. It's going to be more towards these edge applications.

Right. And just to dive in a little bit deeper on the edge and how you're attacking the opportunity. You acquired Colo ATL last year, and as you sort of -- just update us on your learnings from -- of the opportunity from that acquisition. What are you considering as -- like what does the economic and operational model look like for you? Specifically, does it make sense for you to be a landlord to edge data center providers? Or is it feasible for you to move up the stack and operate the data center yourself?

Yes. So we have a couple of experiments and some ongoing operational businesses, right? So we purchased the Colo ATL facility in Atlanta late last -- or early last year, and that facility has done phenomenally well. It's more of a telco meet-me room, so it's like part of that IX foundation that all data centers build around. And within that facility, we have space and power to support various compute applications, so from CDNs being present there to potentially we're looking at C-RAN hubs and O-RAN locations that can be supported out of that facility. So economically, it's been really attractive. It meets a lot of our criteria financially. And we're in a position where we're saying, does that business at the metro level extend or are we really tying it to our towers. So we've been doing some experiments. As you know, we've announced 6 facilities that we deployed in different cities. So one is in Atlanta, so we can tether it to the Colo ATL facility, one in Austin, one in Denver, one in Boulder and one in Pittsburgh. And those facilities are -- and one in Jacksonville. So those are the 6. Jacksonville was our first one, and we've learned in each one of those steps what it takes to perform the requirements to support it. We essentially distributed compute, as Tom talked about. So these are distributed workloads that people would be happy to have locally near or around their franchise or facilities that they operate, and those can be orchestrated from a partnership with, let's call it, classic colo data center partners, so that as those sales teams look for ways to extend their footprint, we are a ready and willing partner to host that. And this would be a way to support the deployment of those sites in advance of the 5G demand when, again, we just talked about low latency and that wave of, let's say, 5G maturing, where the development environment in software and applications will start to embrace lower latency applications. So we're in that early stage now of exploring what markets, what locations. But our value chain performance is really simple. We have a good operating point with our distributed real estate of towers. So we have a good -- great cost position. We also have zoning and permitting that's already set up, so we can move really quickly in deploying compute. So we have a really fast time to market, low-cost point. And we can offer networking and space and power because we already have power at the site, and there's already fiber providers. So all of those things add up to be -- and essentially, if you need smaller bite-sized chunks of compute, we're in a great position to offer that, and it's very synergistic with our tower business because we already have all those sites. We have a lot of sites that we've been looking and studying at for how we filter them for attractiveness. So that's what we've been doing in edge compute, and there's more to come, so we believe the opportunity for us to participate in that market is real and can be significant.

It seems like edge compute has been talked about for decades, maybe a decade, edge data centers at the towers. But it really feels -- and it was always something that made sense, but never really manifested. Over the past year, it feels like it got real. And based on your experience in watching technology unfolds, how do you think -- like at what point do you think we'll really be at a time when edge facilities really ramp up and sort of towards that TAM that you've laid out already of a few billion dollars?

So I think part of the problem was there's always a pendulum of centralized and distributed in compute. It's gone on, whether it's mainframes to PCs to cloud. So we've gone through those. But now we're at the point where there's a lot of concentration in hyperscale. Hybrid cloud became something really important in the last few years. We're seeing all the major players, whether it's Azure Stack or Outposts. Those are all things that are trends that say, how do I bring that, let's call it, multi-tenant, efficient compute closer to the edge. So that trend is happening. That's the distributed compute. But the more important, let's call it, sort of catalyst will be the actual telco networks converting to cloud. So as you cloudify the network and the network sits at all of these sites, that's when you're already starting to position servers and compute. So if you have distributed compute opportunities, those are kind of based on customer demand and the market evolving in hybrid ways, like there are metal players and others that want to be closer to the edge or CDNs. That's been happening, but the real transformation will be when you move to a 5G network and have local breakout. A lot of the early adopters in the last few years ran out early and said, "We're going to put compute edge towers and we're going to do low latency." Well, guess what? They didn't understand that the backhaul network doesn't break out the traffic locally at a tower yet, right? That architecture took encrypted, let's call it, tunnels from the mobile device all the way back to the core network before you hit the packet gateway to get to the Internet. So there wasn't a place to tap that. And so now, when we move to -- they went to a CUPS architecture in Release 14 of 4G and now in the, let's call it, 5G architecture, you can put your exit point anywhere. So you can build what, I would say, local train stations to let traffic off as opposed to have the express train. And because of that shift, and we move to 5G, that really opens up the opportunity to really see an explosion in the edge compute and the applications that will drive that. So we're on the cusp of that. We're still a few years away from that 5G explosion, but the early days have been sort of slower because people thought it was available, and it wasn't. So if you came from outside the industry from IT or compute, you didn't understand the wireless space. Now they realize what it takes to get the wireless space ready. And in the interim, there's a lot of these hybrid and distributed compute models that are taking shape, setting the table and then we'll be ready for that explosion when 5G low latency applications actually occur.