Jon Atkin with RBC. I cover the communications infrastructure stocks, and with me is Sunit Patel, Executive Vice President and Chief Financial Officer of Crown Castle. Welcome, Sunit.
Thank you.
Just had a couple of panels arrive, so that's kind of top of mind. Anything that you're kind of seeing, what your advisors are telling you perhaps to expect in the future around the impact of AI on mobile traffic and what implications that might have for the mobile infrastructure sector?
Sure. Yeah, I think we spend time with both manufacturers of antennas, radio gear, with our clients. What I would say is AI is a clear positive for mobile data demand. It'll manifest in several ways, and many of you have heard about Physical AI. A lot of Physical AI will rest on mobile data traffic transmission. Also with AI, there's a lot more information being gathered or loaded up, if you like, for inference purposes. It's going to change the downlink-uplink ratio of traffic. So historically, most of the traffic, whether it's to the home or even on your mobile phones, there's been traffic coming in. So it's more like 80% download, 20% upload. That ratio is expected to change. There are quite a few research reports out on that, but could be as much as 60% now or 40% upload.
We are still at the beginning. It is a really good, I think it will be a good demand driver for mobile data traffic over the next few years that is positive for us and Mobile Network Operators.
In the network, or maybe even at the equipment level, does that mean additional antenna arrays, new radios? What is your perspective there?
Yeah. The one thing, if you look at the last 20, 30 years, as you push more and more bits out or in, it requires energy to move bits, so you need more power. More power means bigger antennas, whether it is you are delivering more speed, throughput volume or traffic, speed or traffic. While in the background there is continued to be shrinkage in radios and antennas, but the traffic growth in general has been, when you look at our business from 10 or 20 years ago, Mobile Network Operators generally occupy more space on our towers than they did five, 10, 20 years ago. I think that will continue.
We noticed that maybe edge computing, actually going back, so in terms of footprint, and kind of the uplink mix that you have just discussed, thoughts on whether that expands to the majority of the tower footprint, or top 100 markets, or anything around topology that you can divine at this point?
Sure. I mean, a lot of mobile data traffic is driven by people directly or indirectly. It's either people or devices supporting people, or infrastructure supporting people. I think that the larger urban, suburban areas would be earlier, bigger beneficiaries than more rural areas, but the opportunity is across the board.
Edge compute, what's the place of towers in the broader value chain versus, say, carrier COs or purpose-built data centers?
Yeah. Now they're all distinct. There is still a big supply-demand gap in terms of power infrastructure needed for AI data center fabric. Whether you're playing in sort of that 100 MW-1,000 MW range or 10-100, 1-10, different companies are targeting. I think with us, we're focused more on the sub-one megawatts level, but we have 40,000 towers. Each of them have fiber connection and a power connection. And we're a real estate company. We have space on our every tower location, horizontal space, not vertical space on towers. And so we can provide space or rent that space out. The way to think about it is it's having one more tenant like we have on the towers. That's the way to think about it. We provide space. They pay for their own power.
They pay for their own memory compute, whatever they put inside that space, whether it's on an existing shelf or containerized. But it's still early days. I think we described as a trial just to see how that's growing. The applications for that low level of power are more different things, inference, quantum, cyber, but time will tell how that develops.
Talking about kind of the core U.S. site rental leasing outlook, industry seems to be going through a trough in leasing in 2026. Who knows what 2027 might bring? Perhaps it is broadly similar to this year. Has anything structurally changed in your view, maybe driven by the returns on incremental investments in 5G? How do you expect carriers to approach 6G, other mid-band, and other potential drivers to think about next year and beyond?
We said at the beginning of this year that 2026 will be a low point for us, and we expect it to do better than that going forward. Clearly, we have visibility into our MLAs and what is happening with our clients. Back to your point about the Industry sector question. Look, you had some spectrum change hands this year. AT&T spent a lot of money to buy Spectrum, both at the low band 600 MHz level frequency and the 3.45 GHz. Verizon bought some AWS spectrum earlier this year. Spectrum purchases are generally good for us. Lower frequency spectrum is better. For example, for 600 MHz, the radios and antennas are 6 plus foot in height. We think that the industry will benefit from the AT&T 600 MHz.
There will be some benefit just with the mid-band spectrum acquisitions, depending on what Verizon and others want to do with that spectrum, meaning deploy to existing towers or new towers. You have upper C-band auctions coming up next year at close to the 4 GHz ± level, 165 MHz. That should be concluded by July of next year. That is also a positive for us as a tower sector, as that gets translated. As you go up in spectrum, what that means is you either need more densification or, because the propagation is a little less, or you need more powerful, bigger radio antenna structures within the existing macro tower infrastructure you have. As you see, whether it is the upper C-band and then the year after that, some of the 6G application spectrum with the 600, 700 GHz, that should also be a positive for us.
Your MLAs, and you have different arrangements with various of your customers. Broadly speaking, the MLAs in place present and the timing around perhaps the roll-off, does that present an opportunity or an obstacle?
Generally, we have been more in the MLA front. At any point in time, you might be in or out of an MLA depending on the contract. But yeah, we think our MLAs are structured in a way that they generally benefit us as people occupy new towers or want to make amendments on existing towers. But it's particular depending on the client and what specific commercial parameters we've negotiated with them.
Actually, if there's any questions, feel free to just make yourself visible. We should have time for several. But as you think about margin expansion from this point forward, I can imagine a number of drivers, ground lease buyouts, which I think you talked about on your most recent call, transformation initiatives, and just operating leverage as you grow revenues. But what are your ambitions there and what are the most important drivers to keep in mind?
Yeah. So we have fairly aggressive ambitions there that we have been executing on and feel confident we'll do well on that front over the next few years. That's in our control. At the beginning of this year, we reduced our workforce by 20% on the tower side. Last year, we started a set of initiatives to modernize our IT infrastructure. We've deployed two operations systems this year. Next year, we'll deploy a customer lifecycle management and HCM system. The year after that, we'll deploy an ERP system. All of those implementations will serve to both improve our efficiency, productivity, cycle time, number of people needed to get things done, and also how much we spend on operating costs for antiquated platforms. So Mark Lennon, our CIO, is leading that, so excited about what that'll do.
We also have put in place a transformation program that is covering all facets of the company, from head count, non-head count expenses and have a lot of initiatives that we are executing on to drive further expense or margin improvement there. The other big program we have is ground lease buyouts. We have 70% of the ground underneath our 40,000 towers. 70% of the towers have landlords underneath. So we have stepped up our initiative to buy ground leases out. Power is becoming important, as I said. So we're pushing that more aggressively. So I think when you can buy leases out well above your cost of capital, we think that's advantageous for our shareholders. So we're pushing on all three fronts, and I think we've said we'll be able to improve our margins by 200 basis points from where we've guided to this year to 2030.
We are pushing aggressively on that.
Questions. Maybe talk about satellites, use cases, or maybe just more broadly non-big three use cases and demand, and what you have been seeing this year and what you think the future might bring.
Yeah. Let us talk about satellites. I think many of you have read all kinds of research reports that have come out in this space. I think our summary, having looked at those doing our own research, our summary is that satellite is actually an opportunity for us. SpaceX is a good example, but if they really want to be comparable from a service provider similar to MNOs, I think that because of issues with the in-building coverage or in-vehicle coverage even, I think that they will have to use macro towers, especially to target urban and suburban areas with higher population densities. They will probably also need more spectrum over time. But we think that it will be tough to get to where most of the people live, with just satellite infrastructure as currently deployed. We actually think it is an opportunity for us over time.
Capital allocation, maybe just take us through the blueprint. You got a really high dividend yield compared to even some of the carriers and how does that affect your thinking?
Yeah. Our dividend is at $4.25 a share. The yield is very attractive right now. Our capital allocation strategy rests on a few planks. First is leverage ratio. We want to be at 6%- 6.5%. Investment grade, very focused and committed on that. The last quarter that we reported, our leverage was 6.3x. We want to pay out dividends at a 75%-80% coverage ratio of our AFFO. Obviously, we are above that now, but over time, as you will see our AFFO grow, we should be able to grow our dividend at some point. Then we will keep focused on what we can do to further enhance shareholder value, but also keeping in mind our leverage targets.
Esther?
[inaudible] mentioned that, like, getting to trial 5G edge computing now, but maybe starting to look into AI RAN and how would AI RAN change your operating model versus your 5G model. The second question is about connectivity between your towers, like with edge computing and AI RAN, and how would that connect between your towers as well as connect back to all the data centers? What needs to change for the opportunity impact?
Sure. Your first question, if I understand it right, with the AI RAN. From the carriers' perspective, they have several objectives. One is, can we improve our efficiency with respect to spectrum utilization with AI? On the way to doing the 6G standard, they are now calling it the 6G AI standard. So they want to put that in the standard, with equipment providers, say, "How can we optimize that?" I think the other part of the 6G standard, which was supposed to be with 5G, is how many So today, compared to the number of people that live in the U.S., you want to support 10, 20, 100x number of non-human subscribers, and do the standards accommodate that? So that is another area of focus. Your question on network topology. Generally, as you know, our 40,000 towers are connected by fiber. Where does that fiber go?
Usually it is any of the fiber providers that you would think of, telephone companies, cable companies, specialized fiber providers. That fiber usually connects back to some internet access point. So it could go through a junction point. This is a business that I have been in for a long time. But to an internet access point, and then from there, it can be routed anywhere. So once you have that fiber infrastructure coming into a tower, it has plenty of capacity to handle large bandwidth loads. You can put a lot of wavelengths of light through one strand of fiber. So I do not think that will be a constraint per se, but that does, to your point, that does become very important, especially with some of these inference loads and depending on what people want to do at the edge.
I think that was the last question. I appreciate your time.
All right. Thank you.