Let me get the door.
All right. Good morning, everybody. My name is Brandon Nispel. I cover communication services for KeyBanc. This presentation is a 25-minute fireside chat. With us today, we have American Tower's CTO, Ed Knapp. Ed, thanks for being here.
Thanks, Brandon. It's great to be here.
Maybe just start, what's, what's your background and role at American Tower? It's sort of unique to have a CTO in, at a tower company.
Yeah, I've been with the business six years now. I came from a long generation of wireless businesses, so I was formerly at Qualcomm, working on infrastructure there on a silicon level. When I had the opportunity to join American Tower, you know, the business was looking at new growth areas and innovation, and the job really took on new meaning for me because it broadened my perspective on the wireless business. We looked at things in power and energy space in one area that I work on now. A lot of optical areas where we interconnect towers and data centers. Wireless in general, right? everything from satellites to in-building systems to, you know, let's call it even other parts of the wide area network that, that I'm involved with.
Then, the other part is, is really data center, right? We've done a lot of work recently. We obviously acquired CoreSite about 1.5 years ago. The edge data center business is part of what I've come to, let's say, verticalize these areas in the business and then talk about them with our executive team and the management team, looking at how our regions around the world, you know, will be able to take advantage of that.
Got it. Let's start on the U.S. segment. There's a lot of talk around U.S. capital spending trends from your major customers. What's your view on where your customers are from, like, a 5G deployment standpoint?
We've talked, and I think, you know, when you look at the history of a lot of the wireless build-out, there are multiple phases of the build-out, right? The first phase has always been about getting coverage out there. In the 5G cycle, it's a little bit different maybe, because we started out, you know, maybe five years ago, looking at, millimeter wave in the US. We didn't have the mid-band spectrum. Then, the mid-band spectrum came online, you know, through some of the satellite, changes in terms of reorganizing the spectrum from 3.7 to, say, 3.98. Even the second part of that is just becoming available now. There was this pent-up demand to sort of push that coverage, you know, through in the last year or two.
We've probably got to the point where that has happened, and I think about half of our sites, we'd say, have been touched with 5G. There are different underlying technologies in how people are building out 5G for coverage. That's the first phase, and the second phase would be sort of continuing to build coverage, 'cause you get to this, how many pops can you cover efficiently? Then you have a longer tail of pops you need to cover, and a lot of our sites do, you know, represent suburban and, and rural areas, too. There's a long tail ahead of that for coverage growth, which we hope to see, and that's part of the capital plans that the operators will continue. Then that next middle phase is really about building the demand.
We didn't have necessarily the wide area coverage and mid-band, and then devices have to catch up. Once you get devices in the marketplace and the transformation in the network to standalone, one of the other quirks is when all the other Gs were built, we built new end-to-end solutions, like, day one. Like, we had the radio, and we had the core, and we put new services from the beginning. As devices ramped, those services became part of that. 4G had a brand-new IP. Everything went to IP in 4G, and now in 5G, we use the 4G core to append 5G radios initially. Now you're seeing a standalone core where new services can start to take place. We can look at Voice over NR, you could look at things like slicing.
You could start to really experience the true potential of 5G. It's the coverage build now. The second phase is about putting more services in demand and grooming the network for performance. There's a lot of work that's gone on in the last few years with, like, raised CapEx in the industry, and the second phase is about trying to find new services and applications. The third and, and, and the next phase after that will be once those applications start to drive demand, we'll see densification and fill-in, whether it's in building systems, maybe some smaller cells and hotspots, and also more potentially modifications and, and co-locations as, as densification is required. Then we deployed a lot of what's called MIMO, and so those things have to be tuned to support whether you're doing beamforming for, for link budget gain. ...
In a nutshell, those three phases.
Would you say it's fair to characterize sort of we're getting past phase one, and phase two is sort of at our fingertips?
Correct
... where we're really looking for new applications on the 5G network? Help us understand phase three, the densification and the capacity layer that customers will eventually build. What does that look like and mean to you?
First, if we open up pipes, like one of the applications you start to see in the marketplace today is fixed wireless, right? You start to see demand on these networks grow significantly. Now, we've grown year-over-year at 30%, 40% wireless. I mean, how many exabytes a year are being driven over wireless networks? That continues to grow, you know, regularly, and, and now that you have 5G, we just deployed what was considered what they call best efforts, like eMBB, right? This enhanced mobile broadband. We need to then start to take the services with SA, this, this 5G core, into things like 4K video. We need to look at metaverse applications and AR, VR, MR. Those devices are in the, in the pipeline.
When those happen, then you're going to see site thresholds start to, you know, essentially get reached. The operators will have to then refactor where do they want to place investment from a CapEx perspective to support those demands. There'll be hotspots, you know, that may vary as a function of time of day because if it's a machine to machine, we haven't even seen IoT take off in 5G yet. There's this reduced capability, modems that are coming online now. They call it RedCap for short. Release 17 and Release 18 are really that, the meat of features and capabilities beyond eMBB that really are the three legs of 5G. That should drive a lot more demand, and we should see-...
that, that demand translated to changes, whatever you could program software-wise on the sites, but then also more colos and more modifications to upgrade the capabilities of, of what's on the towers.
You, you mentioned everybody's sort of taking a different approach to 5G in, in terms of deploying from an underlying tech perspective. Can you just sort of help us outline what those different technology deployments look like, whether it's carrier aggregation or you've mentioned MIMO, but, but what are you seeing there?
Well, in general, around the world, and we see it from a global perspective, when 5G came, mid-band was the global piece of spectrum that you needed to get to really have 5G, right? The US, it took a little more time to get that in place. What you really needed to do is you're at a higher frequency, right? When you're dealing with, like, the lower frequency bands, you get a lot of more coverage, the higher frequency bands had to compensate for that. That's where MIMO plays a role, right? Now you can extend essentially and keep. What they wanted to do was keep the same inner site distance.
So you coupled, let's say, the mid-band for a lot of your, let's call it, best efforts data with, with some of the lower frequencies for what was essentially the voice that still impinged on the 4G core. Everybody kind of went through that phase, if they built mid-band from start. In the US, when we had some millimeter wave, we were building small cells in the beginning and doing stadiums, and then all of a sudden, mid-band, and some operators had mid-band, but not the 3.5, they had 2.5. They were able to race ahead and build a lot of the millimeter wave, sorry, a lot of the Massive MIMO capabilities, which, you know, allowed them to take sites that before that maybe weren't able to provide contiguous coverage and use that to provide contiguous coverage.
All the operators started to go down that path. When you had a gap in mid-band, people substituted what was called DSS, right? This was this, sort of dynamic spectrum share, you know, so you could switch between 4G and 5G. That kind of was giving you 5G on your phone, but really wasn't giving you a full 5G experience. We see the beginning of the progression, like mid-band, massive MIMO, try to stay on the same site grid, build that out from city center towards suburban to rural as 5G devices come online. See it in markets like US, Japan, and China, and other markets where device costs are high.
We just see device costs coming way down now to something where you can get 5G market prices. Some devices could be in $150, right? Instead of a $1,000 device. That's gonna push 5G out in more markets around the world. When you get into what's happening under the hood, there are a lot of variations and details around massive MIMO and how people will employ that. When we first saw it, we saw a lot of power increase in the sites, so people had to consume a lot more power because you had a lot more antennas, you had a lot more, quote, unquote, "radios," and now you're trying to rationalize that. The portfolios of the OEMs continue to grow in terms of complexity and also optimization of cost and power.
You'll see the second generation of that build-out for coverage start to fine-tune how coverage will, will be optimized as we move to further and further pop coverage from whereas, let's say, 200 million today in the US to something like more like 300 million, where people will wanna get to.
Right. One of the things that you mentioned there is sort of the OEMs and their portfolios. I know the OEMs are sort of working on trying to develop dual-band radios centered around, you know, 3.45 GHz spectrum and C-band spectrum. What's your view on what that means to your business and sort of the timing aspect of when carriers come back with those types of radios? I would assume that that would sort of go through, you know, the densification phase three process that you were talking about.
Yeah, the business in general is based on the fact that every time you have new spectrum, you have to build new radios, and the physics eliminates the, the opportunity to say, "I can build 1 radio that covers it all," right? You just can't do that. We had low band, and people have been for years trying to package different frequencies that are adjacent. We've seen, like 600, 700, maybe even 800, 900 people trying to repackage that. This is a phenomenon that's been going on for a long time. You took, in the old days, a passive antenna, and you had the radios on the ground, you moved the radios up next to the antenna, we called it an RRH.
You put 2 RRHs there for two bands, then you put them together with the antenna, you put them in the same package. Now, when you talk about 5G and mid-band, you have the DoD spectrum at three, four, or five to three, five ,five, and operators that had that in markets came later, right? Now they, they already deployed the mid-band at 3.7 GHz. The radios that exist to support the two, this is starting to go through their sort of approval process in the FCC for type acceptance, and they have out-of-band emissions and all kinds of regulatory requirements to make sure they're not interfering. In the middle, you have CBRS, which is, it got its own set of technical specifications. Now you're trying to build a radio that's 500 MHz across-
Mm-hmm.
You have to have what they call occupied bandwidth that can operate continuously on both. That's really something that is great technology, and what you'll have to do is go back and do an amendment, go back to the sites and change out those requirements, and there are MLAs will accommodate that according to the agreements we have with our operators. It is a natural process. It's happened historically in the business. We've seen, you know, operators working with OEMs to try to consolidate what they can do with each radio, but there's always gonna be physical limits, right? You can't take all the radios. The more challenges of spectrum band as we move forward, we've expanded spectrum now from what used to be just sub 1 gigahertz, then we went in 4G to sub two gigahertz.
Now we're looking at sub 4 GHz, right? Then in the future, you know, we still have the millimeter wave and other things that are coming.
I guess the question on timing still on, on those pieces of equipment from the OEMs dual band, do we, do we know when those are gonna be coming to market?
I've seen some data recently. You can go on the FCC website, you can see what, what has been approved. There are radios today that are, I think, accepted now that can be deployed. I can't speak specifically to what's been deployed and where by any operator, but the radios do exist, and so that will allow for three, four, and 3.7 GHz to be used, and that'll give carrier aggregation benefits to the operators, and again, a higher performance.
Got it. We'll take a step back from the sort of densification, the next wave of densification, and go back to use cases. What, what's your view on sort of the key use cases that are gonna drive bandwidth on the network? One that I really wanted to ask you about was Apple and, and the Vision Pro, which is obviously coming next year. What, what's your view there?
We talked about this years ago when I said, you asked me what's the, the game changer. I said that, at some point, that smartphone device is gonna change to a wearable. I gave the 2020, this is in 2020, I think I said 2025. I'll, I won't, I, I guess I have to stick to that until we get to 2025. The reality is, the Pro Vision is really based on wireless, today with Wi-Fi, but not cellular. They'll get there. I do think the whole, let's say, mixed reality environment is gonna be transformative, because people will talk to their devices as opposed to spend a lot of time touching it.
That interaction is gonna change, and now when you have large language models and AI, you're gonna have a richer interaction, potentially, that you could prompt things, and you can get specific answers to what you want, and it could be visual if you're wearing something on, on a glass perspective. That phase will start to happen. Apple obviously could be a dramatic transformation in terms of the marketplace. Price points have to come down. Technology's not really ready for prime time yet for mass market. That is gonna be one of the game changers and should drive edge in, in, in some respects as well. Metaverse is still out there. It's gone through its hype cycle. It's now probably in the Trough of Disillusionment, then we get back into, like, it just happens.
Like, Qualcomm's got a plethora of devices in the pipeline with folks around the world. People will start using these things in different ways. There was attempts, obviously, years ago with Google Glass and others to try to do this. I think people are more acceptable of that, and especially in enterprise use cases, maybe not the consumer, but, but private networks, private 5G, things like that.
I think the pushback that I would have come to you with is, you know, things like fixed wireless that would take up more bandwidth on the network or sort of a fixed solution, it's not a mobile solution. Vision Pro, it's probably gonna go over Wi-Fi. What's the true, like, mobile solution? How does the tower benefit?
You'll see, obviously, there's a lot more, let's call it, bits that each device is consuming. Just you have to watch the silicon on the device side. Carrier aggregation and the ability to put more spectrum on a single device and more efficiency in terms of bits per second per hertz, that spectral efficiency is going to drive opportunity for supply. Now, the question you're asking is on the demand side. Today, we've seen people experience more or less, more of the same in 4G. We need new transformative use cases. What are those? Well, first of all, voice. Voice is still primarily a VoLTE, which is a, let's say, guaranteed bit rates around 30-40 kilobits per second. You go to rich, immersive voice, spatial audio, you start to talk about hundreds of kilobits. You're trying to maintain that.
Voice over NR could be a game changer when you have SA, when you have standalone core. That's one thing. Interactive, let's call it, video, relative to what we do today with Zoom and all that, that's gonna be also something that's gonna have a richer experience and capability over a mobile network. Those are practical use cases that are coming. Obviously, you could stream 4K. Obviously, more capable screens, larger foldable devices also drive that behavior. No one has a crystal ball, but, you know, fixed wireless does, you know, consume. I think it's there to stay. Really what we haven't seen is automotive. We haven't seen the heavy use cases of drones and robots and those types of devices.
In San Francisco today, right, you can, you could see Cruise and Waymo do their thing, but they have to upload a ton of data. What they have to do, they go to a parking garage at night and do it over, let's say, a Wi-Fi. That's gonna be more continuous. That's gonna require a lot more updating and information to get that to mass market scale. The wireless operators stand ready. Today, most vehicles have 4G phones, I mean, 4G modems in them. 5G modems are just starting to, to sort of be in the pipeline of production. That'll be another driver. You'll see machine to machine and, as opposed to just consumers consuming what they did in 4G the same way. Those are just examples.
Okay, that was great. Are there any questions from the audience? Does anyone want us to dial this back a little bit?
We can go up, or we can go down.
Uh, let's-
There's a question here.
Oh, we do have a question. Question was power on MIMO antennas.
If you looked at 4G, we had two-layer, four-layer MIMO, they would probably be about 300 watts or something like that. If you look at the Massive MIMO ones that had 64T, 64R, they could be up to over 1 kilowatt. two to three times is what we have. Some of them have backed off and tried to come up. It gets into things like GaN, right? What's your power amplifier? How many watts per PA are you supporting? What's the efficiency of that? That led to a lot of consumption. Frankly, the OEMs have gotten their arms around that, and there are also some things that are occurring on the network side to turn, to tune and turn on and off different parts of the arrays to make them more power efficient.
You'll see the average power come down, but the peak will still be something that will be on the order of one kilowatt per sector.
Is that still power?
That would be one sector. One, yeah, one sector. Three kilowatts to four kilowatts, you could see, and that doesn't include the baseband changes. We've seen, there's data on the web about a four to five-kilowatt lift in the consumption at a site for per operator based on massive MIMO.
What are the different use cases? You know, MIMO, like, four months MIMO, now two. I guess it's a couple of miles. What's the range?
It depends on the area. It might be 500 meters in a city. You could be, like, 1 kilometer in a suburban, you know, maybe four to five, too, in a light suburban, and in a rural, it could be somewhere around 10-20 kilometers. You're trying to use the MIMO in different ways. You also use frequency, right? You want lower frequencies in the more rural. The mid-band frequencies have been in urban to try to support stuff, but they'll progress out. And what you're trying to do with the bandwidth, now you have much bigger bandwidth than we had in the recent years.
We started out in cellular with, like, 30 kilohertz in AMPS, but then we went to megahertz, and then we went to tens of megahertz in 4G, and now we're out to 20 megahertz-100 megahertz channels, and by- combining them, which is back to the carry ag. We now can bump, combine all these things to give even higher data rates. One of the things you're trying to look at and when Massive MIMO is, you put that entire bandwidth up, and you now have to put weights on the antenna to point beams and to make adjustments based on where users are. That gives you some additional capacity gains.
You can get coverage gains, as I mentioned earlier, by being able to use it for, let's call it, the array size, gives you some ability to point energy in a certain direction, and that helps alleviate the coverage gap when you move up in frequency. You also want to add capacity, you can then tune, tune that differently and change the way you use your antennas. You have to have the right number of, of transceivers, and that's where the cost comes in. What we're seeing is, opera is gonna have to rationalize, do I need 64T64R, 32T32R, 8T8R, and back to the 4G, 2 T, you know, 2 R, 4 T, 4 R? You have a more complicated set of CapEx expenses and challenges and power consumption.
This, the whole deployment has become more complicated. Even with 5G and carry ag, you got 4G plus 5G, you got TDD, FDD. You got a lot of different things that, as an operator, you have to think through. From a tower perspective, you know, we obviously stand ready to support them on how they want to deploy and how they want to update, and we have tools to help them know what towers are ready to support their equipment, and we think we're best in class at doing that.
On our side, the requirements are going up. Our electrical grid stinks. What do you guys do about that? Are you trying some solar, and what-
Yeah
what's the model for you guys going forward?
We, we have 20,000 sites in Africa today that are. Now, Africa has a lower power portfolio. They haven't done 5G, but we use solar, and we use lithium-ion battery, and we have DG, right? The generators. These are off-grid sites, a lot of them, right? The grid is unreliable, we are, we believe, world-class when it comes to tuning between those power sources at a tower site using AI as well behind it. We have the ability to take the, when the sun's coming up, we know how much charge is in the battery. We can look at how to switch from solar, how to minimize diesel fuel, and that's part of our ESG report we just released a couple weeks ago.
In the US, because the grid is very reliable, what we're trying to do in our, this is back to my earlier statement about power and energy, is getting deep into, like, behind-the-meter, what can we do from a sustainability perspective? The operators and the OEMs are doing their best to try to reduce that number. The peak number might stay, but the average number might come down. Before, when they put it up, it was like a continuous burn of power because they didn't have the smarts behind it. Now, with, with the peak to average, you still have demand rate pricing, you have utilities looking at their grid, moving to renewables. Renewables are less reliable.
We're trying to also work with our operator partners to say, "What can we do to invest behind-the-meter on shared power?" Shared power would be being able to say the power plant, just like the tower, can be something that people should share because it's more efficient to share than to do it all yourself. That's not happened yet, but it's something we do, and we already do that in markets around the world. We, we believe we're a leader already in that sector. It's a matter of can we get the benefit of doing that in other markets where the grid is reliable? We're looking at that in South America as well.
I have to switch gears a little bit.
Okay.
I have just a few minutes left. I have to ask, ask you about Foresight, but, but mainly I want to just get your thoughts around AI. I think everybody's trying to get investment implications around AI and, and in particular for the data center, you look like you're in a pretty good position to be able to have good visibility in terms of what's going to happen. What are your thoughts?
Obviously, AI's been around for a while. I think you, anybody who saw the panel this morning, I think they rationalized it, quite well in terms of the, the hype versus the reality, and where the large language models have democratized the ability for people to understand, like, what's possible. Dialing that in is something that every corporation or even consumers have to figure out how to do. However, it has really been transformative for a CoreSite business because people were moving already from, like, say, we have the cloud on-ramps or we have interconnections. We're, we're not just four walls and a generator, as I would call it. It's a much more intelligent operation that's behind people wanting to interconnect.
The rise in data, which is going to be generated by these models, gives corporations an opportunity to say: Do I want to do a custom model on-prem? Do I want to use a cloud model? Most of, and we've said this recently, too, in the earnings calls, like, Tom opened up with saying, "Yeah, a lot of this is still in the public cloud. It's, it's people getting to and from that," but that has implications for interconnection in CoreSite. The hybrid multicloud model says you want to be in colo because then you can connect with other data sources, and you can find ways to, to sort of use the large language models in the cloud, maybe lock them or use them as a copilot, but you still build your fine-tune models within colo. You could do them on-prem.
We're seeing that, and it goes into how many billions of parameters you want, what sort of use cases you have in your business, what workflows you want to improve. I think the, the, the panel this morning spoke to some of that pretty well, especially around the risks and the regulation and some of the things to think about when using these models. Human intervention is important, and data cleanliness, and the ability to use data to make decisions is part of that.
From CoreSite perspective, all that has really been a positive, you know, let's call it tailwind to the business, and we see that also moving not just from these centralized platforms, but eventually to tower platforms and near the edge, where, they, we s- I mentioned earlier today to someone, the Hardee's example was in The Wall Street Journal. When you're using natural language processing to do your ordering, that's gonna transform that sort of, let's call it remote office, branch office environments, call it franchises. How does that work when you start to have many of these, and the cost of that is potentially high, the power consumption's high? Can I use that more in a centralized or near-prem environment? We think some of that is opportunity as well to bring more AI computing for inferencing closer to where our towers are at the edge.
Instead of going into another question, I think we're just about out of time. Ed, thank you very much, and thanks, everybody, for being here.
Thank you.