Let's get started. Good morning. Welcome to day one of TD Cowen's 53rd Annual TMT Conference. My name is Gregory Williams. I cover cable, wireless, and telco here at TD Cowen. I'm joined for this session by Kenny Gunderman, President and CEO of Uniti. So, Kenny, thanks for joining us.
It's a pleasure to be here, Greg. Thanks for having us, as always.
Sure. I wanted to start with your stock price. On earnings day, shares were down on what I thought was an otherwise decent print, and so did a lot of the investors we spoke to. We wrote that it was possibly the optics of missing guidance, including mine, because we had the Windstream numbers in there, and your numbers, at least optically, do not show that. There could have been some algo-driven selling. Have you heard any updates to this theory or any other theories? The 13Fs were published, and we did not see anything glaring.
Yeah, thanks for the question, Greg. We've heard lots of theories. I think the underlying issue, candidly, is a symptom of one of our biggest challenges right now, but also one of our biggest opportunities for investors. What I mean by that is there's just a lot of technical noise around our stock right now. It could be the perceived misguidance when that obviously wasn't true. We had a strong quarter, strong demand. We reiterated guidance for the company. We reiterated guidance for the combined company. We think the trends and the underlying business are very, very strong. That wasn't an issue. We've heard theories about algo trading. We've heard theories about dilution when the deal closes. We've heard theories about dereading and transitioning to C Corp. Candidly, I think all of those are probably true on some level.
I think on top of all of that, Windstream, the business that we're merging with, is private. There is asymmetrical information with the marketplace. The good thing about that, and look, as a large shareholder, I hate to see that. On the other hand, the good thing about that is it's all technical. I do think at closing and not long after closing, many of those issues will resolve themselves. We'll get back to focusing on the fundamentals, which for me is what we focus on every day. The fundamentals of the business couldn't be stronger, not just the performance of the business, but also the overarching themes that are happening in the industry, whether it be generative AI demand, hyperscaler demand, the convergence themes around fiber to the home. All of those things are really pushing the business forward in an exciting way.
Right. You mentioned one of the potential technical noise issues would be the dilution that's coming up. When the deal actually transacts, shareholders would get like a 0.6 to one dilution. The theory is there might not be bid support. Why buy the stock now? Wait till after the deal. Of course, you'd be part of a much larger company, much larger EV.
Yeah, that's right. I think from a fundamental point of view, that should be a non-issue. I also think because there's some noise related to that, because there's some noise related to the dereading and refunds that are required to sell and some of the other issues that you mentioned, there could be a view, let's just wait for closing, let all of that wash its way through the stock, and then come in. We've had terrific engagement with both the bond market and the equity market over the past number of months. Our perception is that people are doing their work, and they're digging into the fundamentals. Yeah, there could be some folks who are sitting on the sidelines waiting for closing. It's hard to say. We don't have perfect information on that.
When we do our best to try to surmise what's going on, that's our perception.
I want to switch gears and actually talk about the business bookings and the bookings cadence.
Thank you.
One source of potential disappointment could have just been with all that Gen AI hype, perhaps there's an expectation of these big hyperscale deals that would just keep coming. Though you've informed us multiple times there's going to be lumpiness in these deals. Can we see million-dollar-plus quarters, or is that sort of behind us? There's some thought that there's hyperscaler pullback. They might have plans for moving on to inference, so maybe the training's sort of done. So help us with all of that. I mean, what are you seeing? Is it still lumpy? And is there still a lot of fervor in the training connectivity?
Yeah. Short version, we have not seen a pullback at all. In fact, the opposite. We've got a terrific funnel of opportunities, and I'll come back to that in a second. That's the short version. Longer version, I've said that you should expect lumpy bookings, but I've also said that you shouldn't necessarily expect to see these learning deals flow through the vanity metrics, your revenue, EBITDA, and bookings necessarily, because the nature of those deals don't impact revenue and EBITDA immediately. They don't impact necessarily bookings, given the nature of those deals. I'll come back to that in a second, too. That doesn't mean there's not a terrific opportunity and a terrific amount of demand that's growing. For example, a couple of years ago, the TAM around hyperscalers in the fiber industry was de minimis.
Now we estimate the TAM to be $15 billion. In a very short period of time, a year, year and a half, two years, you've gone from de minimis to $15 billion. We've also estimated that over the next several years, we think that TAM is going to be close to $50 billion. Greg, you know, when you compare that $50 billion, the wireless carriers today, on a collective basis, spend $30 billion of CapEx. Just point of reference, that is a tremendous amount of growth that's ahead of us. We're in the very early stages of it. When you look at the types of deals that we were doing a couple of years ago with the hyperscalers, we were selling six strands, 12 strands, maybe doing some lit waves deals.
Today, we're selling fiber that's 30 times that amount, 40 times, 50 times that amount. They're coming back and buying more fiber on top of those original deals. Just massive amounts of capacity that's being bought. Again, a lot of those types of deals don't show up in a big way on bookings, and they don't show up in a big way on revenue and EBITDA today. Today, hyperscalers represent probably less than 5% of our revenue and EBITDA. On a go-forward basis, that's going to really change because when these deals start to—when we start to see the economic benefits of these deals, like for example, you'll continue to see capital intensity be lower because NRCs on those deals are higher, as you know.
We believe the yields that we're going to be able to get on these deals are going to be terrific after lease-up. We have started to give some data on that already. Over time, and this kind of gets into your inference question, Greg, but over time, as we move away from more of these bespoke, highly structured-type learning deals and into more of the inference phase of AI, you are really then going to start to see a consistent uptick in bookings and a consistent uptick in recurring revenue and EBITDA from an MRR perspective. All that to say, million-dollar bookings quarters are definitely not off the table. We used to get asked about $800,000 bookings quarters. That has become sort of the norm.
Very greedy.
That is OK. We are the same way. We love to see bookings growth. I also remind folks, we do not need bookings growth to show mid-single-digit top-line growth. We need steady bookings and very, very low churn. We have very, very low churn. That is one of the benefits of our business that I do not think the market really focuses on. We have incredibly low churn. Just with consistent flat bookings, we are going to show mid-single-digit top-line growth. We are greedy, too. We want to show bookings growth because we want to get our fair share of growth in the industry. We definitely want to play defense in our footprint and prevent others from coming in and winning these big deals.
Right. Box them out, if you will.
Right.
You mentioned that the hyperscaler is about 5% of your revenue. What percentage of the bookings? Give us a flavor of how many Gen AI bookings are in the funnel or deals are in the funnel and the size of these typical deals.
We've given a lot of metrics on this. Greg, you know this. A couple of years ago, the number of generative AI deals in our funnel was less than five. Now it's close to 100. That's one metric. When you look at percentage of the funnel, it was less than 1% a couple of years ago. Now it's close to 20%. It's kind of consistently been around that 20% number over the past year or so, which, by the way, is good. We don't want any customer segment to represent 40% or 50% or 60% of our funnel. We want to be diversified across all the different customer segments that we support. 20% is good. That means new deals are coming in while we print existing deals.
When we look out over the rest of this year, I'd say the number of bigger deals that we see with the hyperscalers is better than it's ever been. I mentioned that on our last quarterly call. That doesn't mean they're all going to materialize. They're not. That's the definition of the funnel, right? Only a certain percentage of those print. With respect to the opportunity there, it's terrific. I've also said, and will reiterate again, the most exciting part about the hyperscalers to us is not the period of time that we're in today building these learning models and building these sort of bespoke network solutions for them. It's the inference phase that's going to kick in when the edge of the network really matters, right? That's when you're going to see edge data centers being built by the hyperscalers.
That's when you're going to see them coming back for more fiber to support the learning models, which is going to add to lease-up on the deals we're doing today. Importantly, you're going to see other customer segments that are going to need to support the inference models that they're using, whether it's large enterprises, health care campuses, schools, college campuses, et cetera. I think that opportunity is going to be a really big opportunity for us, given that we've got a very robust edge network.
Got it. Do you want to speak a little bit about inference? Before I do, I just wanted to talk about the current deal constructs. You mentioned high NRCs or upfront fees has been the norm with some of these big hyperscaler deals. Yields should be north of the typical 5%-10% range. Are those sort of deal constructs still holding?
Yes. Very consistent deals. I try to remind folks that we do all types of deals with the hyperscalers. We sell traditional RUs. We sell dark fiber leases. We sell waves. We sell even Ethernet DIA. Those are like regular way business. What I think you're asking about, Greg, and what gets a lot of focus is, what do these large learning model deals look like? These are more of the deals that are exciting, and these are the larger deals. And we at Uniti view those as anchor deals. We have always said that we want to do anchor deals in the 5%-10% cash flow range. Every dollar that goes into the ground immediately has a 5% to 10% cash flow yield generating on it over a 10 or 20-year contract.
We add the second, third, fifth, tenth customer on top of that to really get well north of 10% yields. Historically, we have executed on that to the point where our traditional anchor deals are now approaching 30% blended yields. Over the past number of years, we're really executing on that model. What we said with these hyperscaler deals is we're overperforming on that. Regardless of what you characterize as the anchor yield versus lease-up, right now, we're approaching 20% blended yields on a combined basis.
There was a fear that you're building out to "the middle of nowhere" because that's where the power is and how you're going to do lease-ups. You're saying if you're getting 30% blended rates, lease-ups don't seem to be a problem.
There are those deals out in the middle of nowhere. We're not doing those deals. We're doing deals in Mobile, Alabama, and Jackson, Mississippi, and Little Rock, Arkansas, and Memphis, Tennessee. We're connecting those markets with some long-haul routes, which, by the way, is also a new sort of thing. It's been a while since the fiber industry has built long-haul routes. We're now building long-haul routes with hyperscalers as anchors. We're connecting these core markets that are highly strategic to our network. Yes, we're getting that lease-up. Long-winded answer to your question, Greg. Yeah, I think we're overperforming on our traditional metrics of 5% to 10% anchor 10%+ for lease-up because within a year and a half, we're already at nearly 20% blended yields.
That's great. We were at ConnectX, the conference a few weeks ago. We spoke to a couple of fiber providers. It seems to me that when it comes to the RFP process, training and inference is getting a little more nebulous, meaning one provider saying, we are doing a long-haul route for many miles connected to one location, but then we're doing lit services for four or five data centers in an availability zone. That tells me it's a little bit of training with inference sort of blended. Is that the way? Are you seeing that? Is that the way it's going to maybe we're going to see this unfold from phase one training to phase two inference, more of a nebulous sort of mix of offers?
I like the word nebulous because I always try to translate what we think about every day with the business and what we talk about with the team and our customers and then translate that to conferences and investors. The reality is, on a day-to-day basis, we do not talk about inference. We do not talk about inference. We do not talk about large learning models. We just talk about building fiber. I think that is proof that what you are saying is true, that it is nebulous, that right now, we estimate 80/20 learning versus inference, and it is going to be 80/20 the other way around three or four years from now. I cannot tell you how much time we spent trying to figure that out to give a number that we could stand behind because the reality is we do not know.
All I know is the hyperscalers are building a boatload of data center capacity and fiber capacity. In some cases today, they're building it in locations that are not at the edge of the network. We do think over time, they're going to build more and more at the edge for inference. To support those inference models, they're going to need more fiber back to the learning models. More and more is going to be built around the edge by non-hyperscalers to support their own inference models. All that to say, yeah, it's nebulous. I think when you listen to what our customers say publicly and what they say to us, they also view it as nebulous. They've said repeatedly that they're buying this infrastructure, and they may use it for AI, or they may use it for something else in the future.
They do not care. Microsoft really got pushed on the Azure growth that they had last quarter, and they were pushed on how much of that is AI versus Cloud. I could listen to them struggling to answer that question. Eventually, they just said, we do not know. It is becoming so intertwined that it does not matter. I think, Greg, that a couple of years from now, we are not going to talk about AI anymore. I think it is just going to be part of the accepted use of broadband because it is going to be infused in virtually every broadband use case that we have.
Right. Right. Good point. Have you experienced any challenges in the service delivery and meeting the demand, whether it's labor, supply, equipment?
No, we really haven't. I think, and I honestly have not heard of any of our peers struggling with that either. At least with Uniti, and this kind of goes back to your earlier question about building in some of the remote locations, we've stayed disciplined about where we're building. We're building in our footprint, and we're building on the edge of our footprint to expand our network. We're not building in remote locations that are locations for us where we don't know the permitting environment, or we don't know the construction environment, or the geology of the ground that we're going to be building in, or where we're taking a flyer on, can we get good contractors, or can we get our crews to those locations?
Those are all variables that you can control when you're building in your footprint, and you know those variables within a plus or minus. As a result, we're delivering on time or ahead of schedule, and we're delivering on budget or ahead of budget. I accentuate that point because we sometimes get the question about, are these hyperscaler deals RFPs, or are they bespoke? The reality is they're not bespoke, but they're not robust RFP processes either because I think the hyperscalers really want to work with partners who can deliver on time because time is critically important to them and can deliver on budget. They're not pricing every deal to the last penny or to perfection, I guess.
Right. The amount they spend on GPUs and data centers, the network is.
That's a terrific point. When you look at what they spend on power and what they spend on data centers and compare that to what they spend on fiber, I hate to say it's a rounding error relative to the overall spend because that's.
You'd rather spend a couple of bucks more and get it done and get it done yesterday.
That's a better way of saying what I was trying to say.
Sure. Maybe talk about bookings that are not Gen AI related. Like you mentioned, AI is probably 20%, but you want that diversity. Are we still seeing a typical Cloud on-ramping, which is, I guess, in the middle of the later innings? Is that still strong? You noted wireless is up year over year. Help us with that. Is that fiber to the tower? What's happening there?
Yeah. So yeah, no customer segment represents more than 20% of revenue for us and does not represent more than 20% of bookings or 20% of the funnel. Excuse me. We often talk about the benefit of the wholesale fiber model is that we are diversified across all the different use cases for broadband, whether it is fixed wireless, mobile wireless, fixed broadband, cable even, or big customers, and certainly generative AI and others. Depending upon the theme of the day or the trend of the day, we benefit. There are times when certain customer segments are down. Everyone in this room knows that the wireless carriers have spent less over the past year or so. In 2024, they were down. We saw that in our business, but our bookings were almost a record year of bookings because we offset it with other opportunities.
We said for wireless in particular that this year we would start to see some growth. We saw the early signs of that last year, and that's what we're seeing. I don't have a percentage that I'm comfortable sharing yet, but I think there's going to be some solid wireless growth in 2025 over 2024. Network infilling, backhaul purchases, just some things that had been deferred, I'd say, for the past 12.
Sort of sell splitting after taking that 5G breather.
Exactly. Exactly. Nice growth in wireless this year. The customer segment that continues to outperform is just fiber to the home providers buying backhaul. That includes the wireless carriers, by the way. Some of that wireless demand I just mentioned is actually those customers buying backhaul. The fiber to the home carriers this year, I think, last year for sure, and even the year before, was probably our best customer segment. You would not guess that, right? We talk about fiber to the home, and we talk about that last mile, and we are excited about that given our Kinetic asset. One of the great parts of fiber to the home for us is it has been our biggest customer segment in wholesale. We have talked about how backhaul is critically important to providing fiber to the home.
It represents roughly 10% to 20% of the cost of providing fiber to the home. We have a big, robust backhaul network that is going to be very synergistic to Kinetic once our deal with Windstream closes.
Right. You saw the AT&T/Lumen deal. There's possibly more backhaul opportunity for providers since that deal was very unique.
I think that's right. I think, as you know, AT&T is buying the last mile.
That last half mile.
Everything from the node back is a wholesale arrangement.
Right. Before we talk about some M&A opportunity, I just wanted to ask a question about AI in general. Are you seeing Gen AI benefits internally with your business or any opportunities to come in terms of taking advantage of it?
Excuse me. For sure. Early stages on implementing it internally. Our business today at Uniti is really simple. We like to keep it simple, not a lot of moving parts. That is one of the benefits of a wholesale business model. I would say close to 80% of our costing and pricing is automated now. That is a trend that is going to continue. It is never going to get to 100% because we have a lot of bespoke-type complex network solutions. Automating day-to-day, site-to-site type costing and pricing is a huge benefit to our model. We are test casing a lot of other things with our business today. To your question about on the come, obviously, post-merger with Windstream, the business becomes a lot more complicated. You are going from thousands of customers to millions of customers.
Customer care and automating a lot of that, plus automating costing and pricing and service delivery and other things is going to be critically important. Excuse me, Windstream today has a dedicated team within their IT group that focuses only on AI and innovating AI into the business. We have been collaborating with that team and working on some test cases. I think there is more to come there. As we say internally, we want to be a leader in AI when it comes to network infrastructure and deploying infrastructure because we are set up to do that. We are going to be a fast follower when it comes to implementing AI into our business. We are going to follow the big guys. We are going to follow the tech companies.
We're going to follow our vendors like Oracle and Salesforce and ADP and others who are spending billions of dollars to develop AI for their customers, who we are, obviously. We'll benefit from that over time.
Sure. I did want to talk about M&A. One would have thought with the tariffs and macro concerns, we would have been pencils down, but couldn't be more wrong. Charter, Cox, AT&T/Lumen, Crown Castle. You're in the late innings of the Windstream merger. There's folks out there thinking that it'd be right for some sort of transaction, whether it's on the Uniti Fiber side or doing something with the Windstream side. A lot of M&A opportunities on the common and just help us with the M&A scenarios that would be most attractive to you, if any.
Yeah. So Greg, you know our history. M&A is always top of mind for us. I mean, it's just a big dominant gene in our DNA. Right now, when we look at our use of capital, putting fiber in the ground is a terrific return on capital, especially for Kinetic. When we think about the yields that we're getting on these hyperscaler deals, we think about the returns that we get there over a period of time. When you do the inverse of a 20% yield or a 30% yield, that's a five-times multiple. If there's stuff out there that we can buy for five or six times, then yeah, we may look at it. Right now, putting dollars in the ground is our best use of capital. With that said, given the assets that we have, we're getting lots of ideas presented to us.
I mean, there's bolt-on ideas for both businesses, both Kinetic and the commercial fiber business. There's a lot of opportunity out there. We're definitely having real conversations about joint ventures, joint ventures to not only build hyperscaler opportunities like you're seeing in the data center world, and we're also getting opportunities to supercharge the build at Kinetic with joint venture money.
Sure. That seems to be a hot one too. AT&T just.
There's just a lot of capital.
Another JV with.
There's a lot of capital out there. When you've got investment-grade companies like AT&T and T-Mobile and Bell Canada tapping into that capital, we would be remiss by not focusing on it ourselves. We certainly are, and we're excited about what those opportunities could bring. I'd say those are some of the ideas that we're looking at. Obviously, we're always looking at non-core assets as divestiture opportunities. I think we've been very good at that over the past number of years in monetizing things at either good multiples or premium multiples. I consider all that sort of tactical M&A, just advancing the business and finding ways to optimize cost of capital.
Overarching strategic transformative type M&A is out there, and it's on the come, I think, especially when you see the convergence trend continuing with consolidation in that fiber to the home space. I fully expect Kinetic to be a player in that when the time comes.
Back on the fiber, Uniti Fiber valuation, one that's sort of come to market or in print is the Crown Castle sale and valuation. I don't expect you to talk about Crown's valuation, but maybe talk about how you're different, what you're doing differently as folks sort of make the, I guess, erroneous read-throughs between what they're being sold for and where you are.
Yeah. We obviously follow that deal closely. Knew, know a lot of the assets that Crown acquired over the years to build that business. We looked at a lot of them ourselves. Very high-quality set of assets. I think when you look at the deal that they did, it's a very unique deal for a lot of reasons. One, I don't think Crown was trying to maximize value on the transaction. I think they had other goals they were trying to accomplish, right? That's one very important point. Two, it's a carve-out. On top of that, they're splitting the business apart after the carve-out. It's hard to get a perfect read, and it's certainly hard to translate that to our business because there's just a lot of differences. I will say Crown was very public about focusing on small cells over the past number of years.
That was their focus. The rest of the business, less of a focus, I'll say it nicely. We're the other way around. We focus on all the other things. We focus on wholesale and enterprise and hyperscaler demand. Small cells is a very small part of our business. When you look at the valuation differential, they're getting roughly 15 times multiple on small cells, which is the business they focused on, and a smaller multiple on the business they did not focus on. That's not a surprise, right? We're the other way around and continue to think that our business that we focus on day to day that we talked about earlier does deserve those premium multiples as a result. Again, not a perfect analogy, hard to play translator for the read-through, but that's how I look at it.
I think when I saw the values, I thought, wow, 15 times for small cells? No, no, no, no. It should be the other way around. I do think that's the case for our business.
Got it. With the little time we have, just last topic would be on Kinetic and Windstream. You're on the cusp of closing that deal. When you do, just help us with the next steps for execution, biggest challenges and biggest opportunities with the Kinetic assets.
Yeah. Very, very, very excited about that business. One of the really big reasons for doing the deal with Windstream was to bring Kinetic on net. That is the way we characterize basically eliminating the MLA and eliminating the complicated OpCo/PropCo relationship and now having one fully constituted fiber to the home business, which we think is one of the last remaining large fiber to the home businesses and independent fiber to the home businesses in the country. Execution has already started. I mean, we brought in John Harabin a month or so ago. John is joining us from Frontier. He had a front row seat at Frontier in helping them transition out of bankruptcy and really supercharging their copper to fiber conversion, obviously now to a great conclusion with their transaction with Verizon.
He had a front row seat on the go-to-market and a front row seat on the build. We are excited to have him on board and immersed with the Windstream team at Kinetic. We are really focused on the build. We have talked publicly about the four and a half million homes that Windstream has. We are going to be able to get to three and a half million of those homes with fiber by 2029. We are not going to stop there. There is more to do. That is roughly three and a half million homes without BEAD and other things. We think on a, and we are going to talk more about this in the coming weeks, Greg, especially as we get past closing, we have an opportunity to get to 80% to 85% fiber coverage in the footprint. That is without.
Bead or any government subsidy.
Right. Without the numerator shrinking because when I say that, there is also a terrific backdrop right now for decommissioning copper that is unaffordable. You see AT&T doing it, and you see the FCC.
Yeah. The letter a few weeks ago.
Exactly. There is a terrific backdrop there to really maximize fiber penetration from the denominator perspective and to minimize uneconomical copper from the numerator perspective.
Were those contemplated in your cost synergies? Yeah. Right.
No. The copper decommissioning was not actually. We are going to talk more about that in the coming weeks too. But so very.