Bora Lee. I'm a Communications Infrastructure Analyst here at RBC. I'm pleased to welcome Dave Schaeffer, CEO of Cogent Communications. Welcome, Dave.
Hey, thank you, Bora. Thanks for hosting me. Thank RBC for a great venue, and most importantly, thank all the investors for taking time out of their busy day and joining us.
Great. So I guess we'll start high level. What are some of the major business themes so far in 2024 in NetCentric and corporate?
You know, Cogent is primarily an internet service provider, and the primary theme has been continued growth in internet traffic. So for our corporate customers, it's the continued proliferation of SaaS services, the migration of computing off-site, and still there is a lot of uncertainty around the exact configurations of their offices and what return to work post-COVID, will ultimately look like. I know we're a couple of years past the pandemic, but companies are still not fully back. Now, we have seen our corporate business recover. You know, pre-pandemic, it had grown for fifteen years at a compounded organic growth rate of 10-point, or almost 11% year over year. That, growth rate decelerated during the pandemic to -9%, and today we're back to about a 4% positive growth rate on revenue in the corporate business.
I think, that will continue to slowly improve as companies continue to follow the trend of getting employees back in the office, and they stabilize their work architectures. In our NetCentric business, where we sell bulk internet connectivity to other service providers, whether they be content companies or access companies. We're the largest carrier of internet traffic in the world. We have a presence in 1,680 data centers and 54 countries around the world, and we saw a huge surge during the pandemic in that portion of our business. The growth rate went from 3% before the pandemic, the quarter immediately precedent, Q1 of 2020, up to 26% year-over-year, the fastest growth rate we've ever experienced, and it's somewhat moderated to about a 9%-10% growth rate.
You know, traffic continues to grow in the mid- to low-20% range, and prices continue to fall. We have benefited from some of that growth being more global and from some smaller providers, which has allowed our average price per megabit to come up. But I think the long-term trends of internet pricing coming down, volumes growing, and having a ubiquitous reach, have allowed us to capture market share, and I think that will continue.
Okay. So AI obviously is a topic on a lot of people's minds and a common area of interest for investors. So how do you see AI playing out for fiber infrastructure trends along the path from training to inference, and how can Cogent actually benefit from it?
So I think AI is real, it is transformational, and it will pervade every portion of our society, and probably every job will be touched by AI. I think it's unfortunately somewhat misnamed. It is not really artificial intelligence, it is pattern recognition and pattern projection, which is a portion of what people think of as intelligence, but it is not cognitive thinking. So unfortunately, that's the name it's been given, and someday it will be true cognitive thinking, but we are still decades away from having that kind of intelligence built into software. I think secondly, AI is possible because of two things: the massive amounts of data that have been collected over the Internet at virtually no cost.
For years, people hoarded this data and didn't know what the economic value of that data would be, and now using it for AI training creates value out of something that did not have value previously. Secondly, toolkits have been developed that allow us to mine that data and ascertain patterns very cost-effectively. Now, I think there are two things still missing from AI. One, there is not a clear business model for AI. In which , what is the marketable value of that output? And then secondly, will the cost of creating that inference be less than the value you can charge a customer for that?
So I kind of go back to the internet, and the internet has transformed the world and has been tremendously impactful to everyone, but it destroyed the telecom industry, and it is hard to find a clear business model that says it is a pure internet service provider that has been economically viable. You know, Cogent is, I think, unique in that position. Maybe if you go around the world, you know, someone like an Iliad in France has been able to build an internet-focused business profitably, but for the most part, the internet has been very deflationary to the telecom ecosystem, and the true beneficiary has been the consumer. I think where we sit today in AI, the world is heading down that same path, which is tremendous value creation, but no real ability to monetize that by the people who are spending the money to create that value.
So in terms of your fiber, in terms of your network and assets that you got from Sprint, as well as the data centers that you're building from the switch, from the Sprint switch sites, do you think that dispersion is related to the needs of AI training being in more secondary markets or the need to actually transport data to and from those secondary markets?
Yeah. So I think for AI, Cogent will benefit three ways. First, on the internet side of our business, there will be more data collected than ever before because that data now has value. Secondly, as the outputs of the training models get embedded in existing software, it is likely that that software will be both more bit intensive and will be more Cloud or SaaS-based, meaning needing more IO to and from the internet. The second thing that is helpful to Cogent is the increase in demand for transport services. It's really the question you asked, which is the fact that the locations of the data that are being used for these training models don't sit co-resident with the actual computing.
There is today a shortage of space and, more importantly, power to perform that function, and oftentimes, the training has to be placed in a market where power is available and the datasets reside in another market. So the cheapest, easiest way to move information is over the public internet. It's the most ubiquitous, it's the easiest to use, and it is the lowest cost per bit mile. And moving up the hierarchy, you will spend about two and a half times as much per bit mile to rent a wavelength, but a wavelength or any dedicated service has three attributes that certain customers value. One is defined latency, two is security, and three, no restrictions on packet sizes.
So if you are transporting information for a training program, and you're in the middle of that program, and all of a sudden, the latency varies, you may be forced to restart that program, and it would be highly inefficient. So you will use a wavelength to do that. If you are even less price sensitive, you will buy dark fiber, and probably at the very top of that cost pyramid would be build your own fiber. And we've seen that from some of the hyperscalers and subsea situations and very select terrestrial locations. For Cogent, we acquired a former long-distance voice network. It was actually the first nationwide fiber optic network. It was built between 1982 and 1991 at a capital cost of $20.5 billion, solely to carry long-distance voice.
That network terminated in 482 proprietary switch sites, and the purpose of that network was solely to deliver LD voice. Well, that application and that market disappeared. That asset became redundant for its owner, T-Mobile, and we acquired that asset for $1. We are in the process of repurposing those assets. What that, in fact, means is taking the fiber, physically extending it to connect to metro networks in 110 markets around the country, reconfiguring those metro markets to support wavelength services, deploying transponder shelves and ROADMs along that network, so we can then provision rapidly wavelengths. We will enter a new market for Cogent. It is a $2 billion addressable market.
We have some competitive advantages: a network with no cost basis, a greater ubiquity of footprint, a network that is purpose-built and architected now for wavelengths, the ability to provision quickly, routes that are unique, and the flexibility to price aggressively. We have an existing sales force calling on that market. We have existing customer relationships with most of the buyers. So we see that as a significant opportunity. We have projected that within five years of closing the acquisition of Sprint, we will go from zero to 500 million in wavelength revenue. That traffic will be on- net and carry virtually 100% gross margin contribution and 95% EBITDA margin contribution. We also realized that these 482 buildings have economic value beyond being former telephone switch sites. Now, many of them are too small or too remote.
There was 1.9 million sq ft of aggregate space and 230 megawatts of in-place existing power. We identified initially 45 of these facilities for conversion into data centers. We've actually increased that to 48. We initially thought we would build data centers that looked just like the ones Cogent already operated: smaller, lower power density, and retail-focused. We had, prior to the acquisition, 634,000 sq ft of raised floor, 69 megawatts of power, and 53 of the 55 facilities were in leaseholds, two were fee simple owned. Fast-forward to the acquisition, we're adding 48 owned facilities. We are adding roughly 160 megawatts of incremental power.
We will continue to operate the kind of traditional Cogent type of data center, but with the acute demand of AI, we've realized that a higher and better use of these assets is to use them for AI training or other more power-intensive applications. We will be either selling or leasing on a triple net basis, one million sq ft across 21 of the largest facilities with 100 megawatts of installed power. We began the marketing of that last April, April of this year, and none of these facilities are today completely converted. One of the first tasks we had to do was connect them to our metro networks. The second task was when we took over, there were 22,500 cabinets of telephone equipment in place.
That would be equivalent to about thirteen miles of telephone switches if you lined them up side by side, and we had to pull those out of those facilities. We're rapidly doing that, and we think AI will be a catalyst for us being able to monetize that portion of our asset base.
Can you give us a sense of the size of those converted data centers, as well as what markets, regions they're located in?
They're all in the U.S. They are geographically spread out. There's one here in the Greater Chicago market. They typically tend to be about 15 miles from downtown. The facility here, which by the way, I've never been to, but I'm going to go see tomorrow, is about 4 miles south of Midway Airport in an industrial park. These are typically on 6 or 7 acres. They would have dual substation feeds, generators, fire suppression systems, and they're initially configured to negative 48-volt DC power, so we need to put inverters in to convert these to AC 120. The buildings probably average about 40,000 sq ft. They're probably about 5-6 megawatts of power. You know, they're in places like Charlotte and Nashville and Stockton, California, and Oakland, California, and Phoenix and Dallas-Fort Worth. They are truly all over the country.
You know, I know some investors have toured a couple or at least one of the facilities in Springfield, Mass. We have a couple in, you know, industrial parts of New Jersey, and we have teams of people working on the conversions.
Great. And just to go back to wavelengths, you've mentioned in the past that you have a backlog of about two thousand seven hundred wavelengths, maybe now it's more. But, can you give us a sense of the customer diversity as well as what sort of verticals that you're seeing demand from?
Yeah, so we've sold about 800 wavelengths. We have about 2,700 in our backlog that we have not provisioned, and we are continuing to build a funnel of additional orders. The customer base is divided into four groups: two large, two small. The two large groups are content-generating companies that need to move information from one data center to another. These include hyperscalers, but they also include CDNs, hosting companies, or any other content-generating business. The second group are access networks. They could be subsea networks, they could be international carriers, or they can be islands of traffic that are aggregated between different metro markets from overbuilders. So in each of these segments, there's probably about a 100 significant buyers, so about 200 significant buyers in total. There are then two much smaller segments. One is the government applications that value security.
Typically, they go to proprietary government facilities, and the other are, you know, usually educational applications. And then there's a very de minimis enterprise customer base, but it's not even material compared to the other four. You know, there's probably less than 50 enterprises that justify a global fiber-based wavelength network.
So you've talked about wanting, targeting, basically being able to offer wavelengths in about eight hundred locations by year-end within a two-week window. If everything goes to your plan, how should we think about the pacing of fulfillment in the new year?
Yeah. So today, our competitors in wavelengths sell them on networks that were not initially designed and optimized for wavelengths. They are a bit of an afterthought of selling excess capacity. They are typically sold on a custom basis that requires custom engineering and usually about six field dispatches to complete one wavelength installation. Typically, the installation windows will be between 90 and 120 days, and they're only available in a subset of those facilities. We architected the former Sprint fiber to connect to 800 data centers, allowing any-to-any connectivity with an installation of a wavelength only requiring two field dispatches, one at each end, and no custom engineering. We have sufficient field resources and provisioning to probably add about 500 wavelengths per month to our provisioning process. Today, we provision about 3,000 IP ports a month on average.
Those ports require one field dispatcher. It's a single-site sale, and that business will continue for Cogent. But as we wind down the rearchitecture of the network, we'll be able to divert those field resources to help with the wavelength provisioning. If the demand exceeds that, we would then add additional resources to both field services and provisioning to accommodate more. You know, on the existing backlog, we should be able to eat into it starting in the first quarter. We also know that some of that backlog probably will fall out because they've been waiting a long time. We sold those orders with an indeterminate installation window. We expect that when we go to market and actually deliver on the promises we made, the backlog should build more rapidly. So, you know, we think that there will be a self-fulfilling prophecy of fast installs creating more sales.
And is there any lag between when it's installed versus when you start to see the impact on the P&L in terms of billing?
So we always recognize revenue for all of our services upon installation. So as soon as that is installed on a pro rata basis for that quarter, we will recognize the revenue.
Great. So, there are a lot of moving pieces around the acquisition related to pieces that are trending down over time. You have the severance payments, you have the favorable lease, which has been settled. There's still the off-net services, the services that you're trying to end of life, the CSA. Can you just sort of summarize for us all these different parts, as well as kind of the timeline of when they can start to trend down, including the Sprint wireline employees and anything else related?
Yeah, so it is a highly unusual set of transactions when the seller pays the buyer $700 million in cash, indemnifies them, and then also pays employee severances above that, coupled with selling you a $20 billion asset for a dollar. But this was a business that was burning $1 million a day. There were really two different assets that Cogent acquired. They acquired a physical network that had been dormant for a decade, and they acquired a customer base of predominantly large enterprises where there was no product discipline, there was no location discipline, there was no margin discipline, and that business was shrinking. When we signed the deal, there was $565 million of revenue, 93% of it was off-net.
By the time we closed, T-Mobile had instituted a number of initiatives that brought the revenue down to $490 million, and the burn from $300 million of negative EBITDA to negative $190 million of negative EBITDA. We have accelerated those grooming exercises. We outlined $220 million in specific cost savings that we could achieve. We achieved 62% of those savings in the first 15 months. We had outlined a 36-month plan to achieve 100% of those savings. We will probably beat the time frame and possibly raise the total target. We also are grooming away non-core products and certain locations, intentionally shrinking the revenue base and increasing the margins of that base. We have some unusual accounting associated with the payment stream from T-Mobile, the accounting for the severance of the employees. So initially, this business had 1,800 employees.
By the time we had signed our purchase agreement, they were down to 1,320. By closing, they were down to 941. At the end of last quarter, there were still 651 of those employees employed in a combined company. You know, we've basically right-sized the organization, or maybe another 100 or so employees that are shed over the next year or two, but we feel pretty good about the total headcount of the company at about 1,900 employees on a revenue base of about $1.1 billion. We also know that the acquired employees were mostly operational, the Sprint employees were mostly sales. We are continuing to grow the sales portion of the combined company and shrink the operations portion of it.
You know, we saw a material improvement, both in gross margin and a reduction in SG&A expenses last quarter. And I think as some of the transaction accounting complexity starts to roll off, that improvement in margin will become more obvious to investors. We anticipate that the combined company will be at about $1.5 billion of revenue scale by May of 2028, with a $500 million EBITDA run rate and a top-line growth rate of between 5% and 7%, with about 100 basis points a year of continued margin expansion.
Split between EBITDA: split between gross margin versus SG&A within that margin expansion?
Yeah, I would say more of the EBITDA margin came from gross margin savings in the last quarter, 260 basis points sequentially. I think in general, going forward, investors should expect a roughly 50-50 savings in each of the two areas.
Great. I think we'll end it there, but thanks very much for your time, Dave.
Hey, thank you, Bora, and thank you all for taking the time.
Good morning, and welcome to the second day of the RBC Global Communication Infrastructure Conference. I'm Shelby Tucker, a power and utility research analyst at RBC, and for this panel, we really tried to to to come together with the members of both the energy community and the data center community, and to try to get a better understanding of what the power needs are, how and how we solve the need for more power, so rather than go through introducing each person myself, I'm going to just turn to Gil, have him start, introduce himself, then Marie, then Amanda, and then Tag. So
Good morning. My name is Gil Quiniones. I'm the President and CEO of Commonwealth Edison, also known as ComEd. We are the local utility here in Chicago and Northern Illinois. We are part of the Exelon utility. Exelon, for those of you who don't know, has the largest platform of utilities in the United States when it comes to number of customers. We serve over eleven million customers. We own and operate the utilities in Washington, DC, Baltimore, Maryland, Wilmington, Delaware, Atlantic City, New Jersey, Philadelphia, Pennsylvania, and right here in Chicago and Northern Illinois. We are known to be excellent operators in terms of reliability and resiliency. We do it with high customer service and very competitive and affordable rates. Let me tell you a little bit about ComEd.
ComEd serves about 4.3 million customers in Northern Illinois, about over 9 million residents, 70% of the total population of our state. We are known to be the most reliable electric utility in the United States. PA Consulting, they do this for a while now, for many, many years, benchmarking of all utilities across the United States, and last year, we were named as the most reliable utility in the United States. Site Selection Magazine, the group of site selectors, real estate site selectors, also named us as a top economic development utility in the United States, so we're very proud of, of that, this recognition.
We are a partner of the state here for an orderly and equitable clean energy transition under a state law passed in September of twenty twenty-one called Climate Equitable Jobs Act under Governor Pritzker, and so that is our the path that our state is taking on. One thing that you will hear, and probably that's why Shelby invited me here, is that we're seeing a lot of growth-
... of high density load customers, especially data centers in our service territory, and I'm excited to share that story with you and have a really robust conversation with my colleagues here.
Great, thanks.
My name is Marie Lee, and I oversee the global data center portfolio at Oracle Cloud Infrastructure. Previously joining Oracle, I've only been with Oracle for nine months. Before that, I come from a Canadian pension, overseeing their international commercial real estate portfolio. I'm pleased to be here today. Thank you for having me.
Hi, I'm Amanda Rome. So I'm with Xcel Energy. I'm the Group President of Utilities and Chief Customer Officer, which, you know, if you know Xcel, we're a vertically integrated, pure-play utility, and we're in eight states, and we have four operating companies. So we're Upper Midwest, and then we're the largest in Colorado, and then we have Texas and New Mexico. And we have four operating company presidents who run those, and they report to me. On the customer side, you have everything from the customer call center to our EV programs, to our DSM and energy efficiency programs, to relevant today, today economic development arm. Yeah, I mean, I would say Xcel is really known, for those of you who are less familiar, we've really been kind of the clean energy transition, I would say, standard bearer for the industry.
We were the first utility to come out and say: We're gonna be 80% carbon-free by 2030 and 100% by 2050, and that is a true 100% goal, so that's a little bit about me and the company, and I think we'll have a great discussion today.
Awesome, Amanda. I'm Tag Greason. I'm from QTS. I'm our Chief Growth Officer. I've been with the company for 14 years. QTS has three customer sets that we service, the federal TS business, enterprise, think Fortune 1000, and then we have the hyperscale business that we started in 2018, late 2017, to focus more specifically on large-scale deployments. In that 14 years, I've had the opportunity to be a private company, a public company, a private company. You guys decide what's next. Don't know. We'll decide, but a super exciting time to be in the industry. I've said for many, many years, it's better to be busy than bored, and therefore, we are busy. So super excited to be here, Shelby. Thank you.
Great. Thanks, Tag.
Yeah.
Let me start with the first topic, which comes up often in meetings we usually have on the utility side, demand, real versus, versus not. And what we try to figure out is there's clearly a lot of demand, there's no doubt about that. But some of the numbers we hear sometimes, I was talking to one colleague, for those who were at a panel we had in June, where they talked for Columbus, Ohio, alone, going, the real demand was probably around 3-5 gigs, but the indicative number could be as much as 100 or 110 in a market that's 1,200 gigs total. Clearly, I mean, a market, I mean, the entire U.S. production is 1,200.
Maybe starting with with you, Tag, just 'cause we start with you. As you think about that growth potential, how do you determine what is a true growth number versus or what is aspirations, but just not a real reality?
You know, it's it's everybody's got a number, everybody's got an opinion. I'll tell you from my seat, which I actually love the seat that I have the honor to sit in, I get a chance to talk to every one of the hyperscalers and all the large customers around the globe about what they're doing, and I'm gonna boil it down for you in four trends that are driving demand. The first trend is AI is additive. Lots of people pick up a newspaper, every other sentence is AI, AI, AI. But don't forget that the underlying portions of these businesses is still growing at 20% or 30%.
And so if you think about the growth pattern of any one of these hyperscalers, it's 20% or 30% before you read The Wall Street Journal yesterday that said, "Add AI on top of that." So first trend, AI is additive. Second trend in social media, it's not about adding more users. My mom can only post so many cat pictures. It's not gonna drive growth. It's just not gonna happen. It's machine-to-machine learning inside the data centers that is driving the product set of every one of the hyperscalers and is driving demand. So second trend is, it's not users, it's machine to machine. The third trend is you used to have to be in an availability zone. If you are not in this three-legged stool in Frankfurt, Germany, or Ashburn, Virginia, you are not signing a lease with Microsoft. You're just not gonna do it.
The scale has been broken because AI is independent of that availability zone strategy. So now you're talking about Cedar Rapids, Iowa. North Dakota came up yesterday. There are areas of the country that actually have not heard of a data center, know what we do, not sure how they're gonna plan. And so if you think about this availability zone and the AI, that trend has opened up literally gigawatts of capacity that need to be served. And then the fourth trend is super interesting. The conversation used to be, should the hyperscaler insource or outsource? It was an or conversation. The conversation is now an and conversation. The hyperscaler is going to insource and outsource.
So now there are two players entering in anywhere USA, asking for the same amount of capacity, and the first reaction might be, "it's still just one gig." No, no, no, it is now insourcing and outsourcing. Microsoft's gonna build something, and I'm gonna build something, and that is double gig. So if you take those four trends, you can easily see that the numbers are up and to the right. I'm not the one to tell you that Cleveland is exactly 1.2 gigawatts. I just don't think that's how this conversation is gonna shape out. But ultimately, those trends support this is not a blip. This is real demand that is driving this conversation.
But obviously, there are bottlenecks in that dream or, in that in that in that goal. Yeah.
I guess maybe turning to some of the power folks here on the panel, when you receive bids or request, how do you process that information? How do you prioritize the request? And how do you verify that it is actually information that needs to be, or that it's an order that needs to be filled?
Yeah.
And maybe go start with you.
Yeah, here at ComEd, we have close to six gigawatts of applications that we call our engineering phase. When we say engineering phase, say 29-30 applications, close to six gigawatts, that they have paid us a deposit, a significant deposit, so that we can start the engineering process of the electrical service infrastructure. And in some cases, they have paid us money to reserve and pre-order long lead time critical equipment, such as transformer and circuit breakers. So that's. And in our history so far, those customers who have gone through that phase will proceed. We probably have about a 70-75% hit rate that they will proceed to actually building the data center.
We have in our prospects phase, you know, they have not paid a deposit, they have not paid money to order the equipment. We have 13 gigawatts in our pipeline right now. So, we will see how many of those will go through the engineering phase. But that's kinda how we process and predict demand. But as you can see, they're substantial, and I agree with Tag that that you know, it's no longer very location-based. Meta is building five data centers in DeKalb, Illinois, which is kind of far west of Chicago, where football fields equivalent to the size of an aircraft carrier, you know, those five data centers. Compass is building a data center in the old Sears headquarters in Hoffman Estates, which is northwest of Illinois.
And so it's a nice kind of symbolic from the old, you know, Sears headquarters to now the new, which are, which are going to be hyperscale data centers. So that's kinda how we predict the pipeline.
Got it. And Amanda, in your case?
Yeah, I would say it's not, it's not that different. You know, part of what you're doing, it sort of reminds me how people work through transmission queues. Like, you know, you have to prove that you have some credibility, like site control or, you know, if it's QTS, we want to be talking about who's the end user? Like, do you have someone? Is it real? Because there's the pipeline is significant.
Mm.
So you really and we're not going to do all of those. So you have to figure out what's real, not only, you know, so we can marshal our resources and figure out, you know, what what to put in the transmission study and that sort of thing, but as we go out and talk to our regulators, our policymakers, you know, the crisper we can be about what load is real and what we need to plan to from a policy standpoint, the better off we'll be. 'Cause, you know, we've talked about this, you know, QTS has. You know, we have a deal in Colorado that's that's been public.
I mean, we have to provide a narrative to them around why it makes sense for the customer, why it makes sense for the company, and the crisper we can be on what that opportunity looks like, the more effective we can be there.
Shelby, there's one component of this that Val Walsh from Microsoft mentioned yesterday, and I've had this conversation many times. When a hyperscaler comes out into the market and says they have a 500-megawatt need, then me, Compass, Digital, Equinix, Vantage, we're all gonna then put in an application for a 500-megawatt need, and it's not a 3-gigawatt need. But there's no way in my mind. I'd love to hear, and I'm not trying to be the moderator here, but I'm gonna ask the question. Ultimately, how do we help you guys sort through that? Because I'm gonna tell you, it's a legitimate need.
I have land, I have. I'm gonna buy equipment, I'm gonna lean in, and all my competitors are gonna do the same thing, but it's still only five hundred megawatts, and I'm not sure if we have a solve for that as a-
Mm.
- as an industry to figure out.
I would posit you solve it on your end and come in with a 500 megawatt. I mean, it's kinda hard for me to solve it on my end, like, truthfully.
Yeah.
Right?
Yeah.
Um, so...
All right, I'll find someone else to solve it.
Do you still wanna be the moderator?
Yeah.
In that case, let me switch to Marie. No, actually, but a little bit along this, the total process, when Oracle goes through and tries, they've made a determination they need a new facility, and there are many you probably need, what process do you go through when you try to determine the right site?
Yes, so when we look at site selection and when we try to seek a capacity to fulfill our demand, we ultimately balance three factors. The first one is obviously the power availability, and the second thing is the speed to market, and I believe, again, Val from Microsoft has mentioned it yesterday as well. I think that's partly true for all of the cloud operators. And that speed to market also includes not just power, but the operator's track record, and their access to supply chain, whether there's a predictability to that, and the last factor is actually our network setup that's specific to Oracle, so for any availability regions that we already have investments in from a network infrastructure standpoint, we will definitely prioritize those.
And then, all these three factors then translate into ultimately the total cost of ownership, and we seek to minimize that. And then finally, what I wanna say is these three factors, it's a continuous kind of optimization problem for us, because depending on the use, and the type of demand that we're trying to work through, the importance of these three factors will change. So, for example, if we're looking at a kind of AI contract, so for training demand, then latency is not as critical, so network becomes secondary, and we will prioritize kind of speed to market, versus the network needs.
Got it.
Yeah.
And then just to kinda stay on that, on the topic of kind of development, so from the beginning, from the time that you determine you need something, how long does it take till the point you have construction? I realize it depends on the size of the center, but kind of give us a sense of that timeline and how it's gonna tie in with the power side, where it gives them a little lead time. But so we'll start with you, and then I'll go to Gil Quiniones. When you hear that, what do you do on your end?
Yeah, that's a great question. So, I think based on my experience out of OCI, definitely in the last 12 months, we've seen a huge kind of uptick across the organization on data center contracting and CapEx, and I think these kind of break down into 2 portions, so the first part is around organic demand, and we have a better idea of how these will grow, and we have a better handle in terms of forecasting that, let's say, a year out or or or maybe a bit longer than that.
And then, with the dynamic kind of shifts in the industry right now, with what Tag has just mentioned on AI and, you know, ChatGPT has gone through how many iterations in the last 12 to 18 months, right? So every time when there's iteration, these demands change, and a lot of times we find ourselves in a situation with the customer saying that: "Hey, we need something in this location in six months." When that happens, we go out to the market, and really, whoever can deliver the fastest wins. And I think this has been a common theme, kind of from the capital side as well, where, you know, whoever has access to dry powder, who can be nimble, who can move fast, wins.
I'll just stay on you just for a second on that one. So when you actually decide, okay, let's say QTS or something, to what extent do you go in to verify they actually have the power contracts, or they have all the other as- ... 'cause six months still means that they have to continue to develop-
Yeah
... whatever they're working on, how, what's the level of due diligence that you make on that level, on that front?
Yeah, for sure. That's a great question, so we, of course, we do site visits, and you can physically verify the progress of the construction. Secondly, we've also asked operators to provide us with signed PPAs-
Mm-hmm
or ESAs. In that way, at least we verify that the operator does have a contractual arrangement with the utility, that there is, at least contractually speaking, they're obligated to deliver, you know, a substation or some kind of connection, and there have been situations where we do also meet with the utility companies directly.
Mm.
So all these things become a part of our due diligence efforts.
Got it. Okay. Gil, switching to you, so one of the things about your company, which may not be clear to everybody here, you're in the transmission distribution business-
Yep.
You're not in the generation business, so you don't control that part of the product. You still have an obligation to serve, at least if if if paid appropriately. How do you actually drive the decisions on building out transmission for the customer, but not necessarily knowing where the generation is gonna come from?
Yeah. So in our case, when we look at the applications that are coming into us, you know, the first thing that we do in our service territory, at least, we are fortunate that we have a lot of base load generation here at the moment.
Yeah.
Right? But growth is happening, and so really, we track that. You're right, we're a pure delivery company. We're a transmission and distribution company. We do not own generation, and we rely on the regional grid. In our case, we call it PJM Interconnection, that operates and manages all the generation in that region. So we rely on what they can supply to our transmission and distribution system. And we're very active in the, in leadership and influencing issues. Like in the utilities, people like me, you'll see words like resource adequacy and transmission security, meaning, do we have enough capacity? Do we have enough transmission to be able to supply our customers? So so far, we're good here in Illinois.
And in our case, historically, pre-COVID, it was a lot faster because there was not a big supply chain issue for transformers and breakers. For example, the Meta example that I mentioned, we built a substation in 23 months for their development in DeKalb, Illinois. Post-COVID, I would say roughly data centers 300 megawatts and smaller, we can, we see about three years, plus or minus, and greater than 300 megawatts, about five years. Now, ComEd, we have a lot of history and experience and success with various types of data centers. We have been dealing with data centers for the last 25 years. You know, the Chicago Board of Trade, Mercantile, and the corporate environment here, we have—there are about 75 data centers. Well, I call them legacy data centers here, about 500 megawatts.
Now, we're seeing the newer data centers and hyperscalers. So ComEd has a lot of experience. The owners and the developers are very familiar with our process. We've actually curated sites in our grid. We said that here are the places in our grid where we have hosting capacity for large loads. So that is something that we communicate. We work very closely with the governor's office and the economic development agencies both at the state and regional levels here. So it's a very coordinated process, concierge-like service, if you may, in dealing with data centers. And so with that, data centers have confidence in the planning process with us to be able to deliver-
Got it.
the data centers that they need.
Got it.
I mean, we're obviously different, right? Because we're vertically integrated, and and we view that as an advantage in a lot of ways, because we actually control more directly the generation. And we, when we go through resource plans, and we look at, all right, what's the fifteen-year look? But really, the action period is five years. So if you think about just, like, the high probability demand that we have now, that's captured within those forecasts in our resource plans. Now, we know there's significant upside above that, so we have to think about, you know, there's always the next resource planning cycle, but also creatively around load flexibility. I don't think we've talked enough about that.
I know that there's some needs, and these are, like, high, high load factor customers, but we have tons of excess energy because we are so carbon-free, we're, we have so many intermittent resources, that there's opportunity there, I think, to creatively say, "All right, we can serve you 95% of the time." For that, that's above and beyond, like, the capacity we're already building to through our resource plans. Like, we could serve you 95% of the time. For that last 5%, what's something we could do that's, like, behind the meter, where we could lean on it most of the time, but, but as a system resource, but for the 5% you need it, it's yours?
Like, there are ways to solve, I think, for these issues, but we do have to partner together, be educating our regulators and policymakers about it, and make them believe that this is sort of not just a positive story for data centers and the company, but for other customers. And I think it is, and I think we can actually tell that story effectively. But part of what I'm really happy to see as an evolution over the last, call it three years, is the desire for the QTSs and hyperscalers of the world to step forward and say: Let us go in with you and tell that story together-
Mm-hmm.
-versus, you know, not wanting to do that, 'cause that's a huge unlock, I think, for us in our jurisdictions.
Yeah, just to be clear, that's the unlock with the regulators, not with the investors. But so you actually gave a good point here about, and a good segue into also the mix of generation. You know, there used to be a very dogmatic view about generation had to be renewables, 100% decarbonized. There seems to be a bit of an evolution in that thought process. So maybe, Tag, I'll start with you, maybe, and Marie, right after that, on how do you think about the need for generation versus the type of generation?
Yeah, it's a great question. So, I'll take you on a journey. Eight years ago, when a hyperscaler sent out an RFP, Oracle might be different, but I'll use this example anyway. Check the box: Do you have a renewable energy program? Check it, yes or no. That's the RFP response. Then, the next year was, can you describe your renewable energy program? And then, the next year, the question was, not only can you describe it, but can you tell us if there's a nexus to the generation or if it's just offsets? And then, the next year was, okay, it can't be offsets, it's got to be generation, and we wanna be part of that discussion. And then, the next year, it was, we wanna not only know what the generation is, we want this specific mix in this area.
And then, there's an end to this story. And now, the RFP is, just tell me where I can get a gig. Right? And so you go through this-
Full circle.
You go through this journey, and then you're like: Yeah, but just tell me where I can have a gig of power. And that journey is fascinating to me because there are so many intersections of the entire supply chain in that journey that we need to kinda tie together and figure out what is the the the thing that we want. What Amanda just said, it is super creative. I mean, super interesting: people are now willing to come to the table and have creative conversations that didn't exist two years ago or three years ago. And more importantly, the lease that I sign, which is legally binding with termination rights and fees and all kinds of stuff with Oracle, also can now evolve into being a little bit more creative.
That creativity is what we need to try to figure out how do we solve these problems?
Yeah.
So.
I mean, one of the arguments that we we we hear is that obviously renewables does not provide the twenty-four/seven load. There are other resources, combined cycle gas plants, nuclear, that do. Coal also does. However, there are restrictions, environmental restrictions, just natural restrictions of wanting to build out. It does seem to me, and I'm sure Amanda will confirm this, that coal is clearly no longer part of the equation. It may. Retirements may be delayed, but they're at least no longer being considered to be added unless we get to a point of carbon capture that's extraordinary.
What are, what are the prospects of seeing more gas plants being built in anticipation of all this need and kind of bypass some of the measures that have been passed over the last ten years?
Yeah, I mean, it's it's a complexity for sure, depending on what jurisdiction you're talking about. I think that there is certainly a conversation to be had around this as a bridge, because while at least we're hearing from the hyperscalers, that same thing, right? It used to be cost, clean energy, and then, you know, maybe speed.
Availability, yeah.
Right? It's, like, completely changed, and we know that. But I think what we're hearing is long term, they still have those goals, and frankly, medium to the long term. And so I think we can use that, use where they're at, help them talk to our, you know, regulators and policymakers with us through that transition, right? Like, yes, for a period of time, we are going to need natural gas generation. And by the way, in our eight thousand megawatts of RFPs that are currently out there, there's dispatchable generation, natural gas generation as a part of that, because there's already an acknowledgment outside of the data center world that that's really critical to the generation mix in our jurisdiction. So I think it's a further conversation around that, that tracks a little bit this journey that you talked through, that this is a moment in time.
It's good for our customers, actually, to lock in this load growth, but then there's a path through that if you have jurisdictions that care and customers that care about that carbon reduction story, that we get to fill in the blank. SMRs, you know, geothermal in some places, whatever, carbon capture. But so we, we think about it as kind of a transition story that we can help tell.
I totally agree with that.
Yeah.
Totally agree. Yeah.
I think I agree with Amanda. I think it's regional. I know there are conversations in Pennsylvania, for example, whether you know new generation regulated generation should be built to address growing load. But I like to focus on transmission. Our federal regulators, Federal Energy Regulatory Commission, they issued a you know FERC 1920 , which is encouraging long-term planning to build a transmission grid, and we think that that is a very important thing to do. We support that. They also have a an Order 2023 , which is how do you interconnect generation in a more streamlined and faster way to the grid?
So we've been supportive of those, and we think that in addition to the question of whether gas should be back in the mix, we are really, you know, pushing for long-term planning, for faster interconnection of generation, especially clean generation and dispatchable generation to the grid, because the load growth is real.
One of the topics that also has come up is, and you actually just very briefly mentioned about Pennsylvania and some conversations about re-regulation, is the frustration with, and we heard this yesterday in a number of panels, about just power not coming to market fast enough. And some have been talking about, well, in the deregulated markets, we have to find the right incentives to build new generation. The existing generations may not see that quite yet because the mechanisms are not very well structured. What are the possibilities of re-regulation, whether that makes sense? Also, and maybe, Gil, looking at your past position as head of New York Power Authority, what are the possibilities that the state may decide to create its own power authority and start building generation?
What are the where are those conversations, if there are any, along those lines?
It's been a while, but I used to be the president and CEO of the New York Power Authority for for eleven years. For those of you who are familiar, there are really three types of utilities: investor-owned utilities, Amanda's Xcel, and ours, Exelon ComEd, were investor-owned utilities. There are public power utilities, state-owned utilities, municipally-owned utilities, and then rural cooperative systems, and mostly in in rural areas, in remote areas. I wouldn't be surprised if those, you know, public power and rural cooperatives are thinking about getting back into the mix, especially those that are not part of regional transmission organizations. So it's it's really everything has to be done since we're seeing this load growth, and not only generation, but also transmission, because especially the large public power, they also own transmission systems.
... Particularly when you have, you know, federal agencies talking about as a national security imperative.
Yeah.
I think that throws sort of a policy piece into the mix that opens up the aperture a little bit.
One that maybe last topic I'm going to hit before I open up, so for any questions, emerging technologies. Clearly, we're relying very heavily on our existing fleet and our existing technology, whether it's transmission. There are a number of technologies that are evolving. Some have already matured, like renewables, wind, solar, battery storage is coming into its own. What about nuclear? There's, of course, existing, but SMR in particular, does have, does that have a role to play? It probably does have a role to play, but there needs to be a commitment. Where is that commitment going to come from, in your opinion?
Can you just start with Marie, start with you just briefly in terms of, as you think about investing in the data centers, how willing do you think Oracle would be in co-investing in power facilities to facilitate more power?
That's a great question, because I was just thinking about that, the other day, when we were talking about it seems like the agreement globally is that power is constrained. And the natural kind of reaction to to that is, okay, can the end users contribute towards the generation efforts and and and kind of put up the cost, for the time being and then maybe recoup it later? There are creative ways that you can financially engineer to arrive at a solution, but based on kind of what I the people that I've spoken to, it sounds like it doesn't really add to the it doesn't improve the time, the delivery timeline of the overall project.
And I think that's critical to our consideration and decision around whether we are willing to do it or not. It's not really about the money, but even if we spend the money, we're not going to speed it up by that much. So that majority of the benefit was, is kind of taken away.
Got it.
So that's, yeah.
Yeah, I wonder, like, sort of going back to the previous point, I think you have to do both things at once because your speed is so important. So that's where, like, the natural gas generation comes in or whatever you need for the short term, but, like, you have to get to something else, right? Because all of those clean energy ambitions, that's going to come back. It's gonna come back around, and we need a longer-term solution for it. And so I think there's a couple parts. There's, like, the financial engineering part around, like, how can we get a creative deal that sort of helps us as a vertically integrated utility who can't always take big bet risks on new technology, just given our business model?
We've been pretty successful at it at Xcel Energy, but like, thinking of it at that scale, like, we probably need a partnership there. But there's a partnership on the policy side that's super important, too. Like, how do we work together to, you know, talk federally about, to the NRC, about how do we move the permitting along? Like, how do we think about, you know, all of the issues that have, like, stuck nuclear in a place for a while and, like, get that unstuck together? And it does truly feel like a moment for me where we could get some muscle behind that, that has been unwilling to step to the fore on an issue like nuclear. Do it now, and let's unstick some of that, you know, federal and state policy stuff.
There's one other component, which I totally agree with. The technology, I don't think that's going to be the problem.
Mm.
I don't think the willingness of the operators and the customers is going to be the problem. I don't think the capital markets are going to have the problem. I think the policy has got a long runway, but I also think there's the public perception. We built, we're in the process of developing some land in Manassas, Virginia, which is next to the Civil War battlefield called Manassas. If you want to get opposition to a project, build a data center near a historic battlefield, and then add nuclear to that project. And let me just tell you how that's going to go. Probably not well. So I think there's, to Amanda's point, there's a broad spectrum of of of all five of these.
I had a chance to talk with Governor Youngkin in Virginia, and he said, "Tag, you're absolutely right." And then you need a governor, you need an executive of a state to say: I'm in. I will partner with you to make this happen.
Mm-hmm.
If you line all that up, you're good. But to Marie's point, it's not going to help you today, but it's going to help you in twenty thirty.
Yeah.
You better, we all better be thinking about that.
What's interesting is all of a sudden, we're seeing, at least in our jurisdictions, a number of our politicians at the governor level sort of stepping forward to say, "I'm going to say, like, the word nuclear out loud, and talk about that in a positive way," which is a real change.
Yeah, it is a change.
It's a real change.
Yeah, and it's being tried now. If you talk to our folks at the National Labs, if you talk to the Electric Power Research Institute, they they believe that you have to at least build ten of these SMRs, for example, before you can really say, "Okay, it's a mature technology," or, "You can really for deployment." Ontario Power Generation is doing it now, TVA, Tennessee Valley Authority. But that's just two, right? So you need eight more to really prove it out. Long-duration storage is also just starting. So all of this, not going to help today, but I think we ought to pursue. But I do like the concept, and I'm going to introduce this probably in this conference, the bring your own generation, BYOG, for the data centers and the data center developers.
I think that they, from a policy perspective, we got to work with the federal regulators because that's a long lead time, state regulators, but this partnership that Amanda is talking about, the creativity- ... that's happening amongst data center developers working closely with us, in finding solutions and bringing new capacity to address this high-density load is gonna be critical.
Great. Well, with that, we actually, I'll take one question.
Yeah, take one.
Go ahead.
Really, for Tag, it doesn't relate to energy generation, but we think about Scope 3 emissions. Precast concrete, that doesn't get a lot of play in the press, but it's a big emitter.
Yeah.
You, you're deep in that world, so just curious what you're seeing there.
Yeah, and just repeat the question.
Precast.
Precast. Precast and the Scope 3 emissions problem around precast in the data center industry. We we actually are looking at every aspect of the supply chain around the emissions and what we're doing about it and what we can do to be more effective. Certainly part of the problem, I wouldn't say it's leading the train, but it is part of the discussion, Jonathan.
Yeah.
Yeah. Well, with that, we've run out of time. So thank you so much for joining us this morning. Please join in thanking our panelists.
Good morning. My name is Bora Lee. I'm a communications infrastructure analyst here at RBC. Pleased to welcome Kenny Gunderman, the CEO of Uniti, so welcome, Kenny.
It's great to be here, Bora. Good morning, everybody.
Great! So I guess the big news is the impending merger with Windstream. Can you give us an update on the merger timeline? Any sort of regulatory updates?
Yeah. So we're. We announced our merger with Windstream a couple of months ago, and we, at the time, announced that we expect that transaction to close in the second half of twenty twenty-five. That's still our expectation, but I'll come back to that, timeline in a second. We were very excited about the strategic merit of that transaction when we announced it. I'd say in the couple, three months since then, we've become even more excited and emboldened, by the rationale. I think there's a couple of really important themes in the fiber space right now, that we've been talking about a lot today, and I'm sure you have as well, Bora.
But, the generative AI hyperscaler push in the wholesale fiber space, as well as the convergence trend in in in telecom land around fiber to the home, and mobile broadband. And I would say the new Uniti is uniquely positioned to take advantage of both of those things. And I think we're one of the few fiber companies, pure-play fiber companies, that's uniquely positioned at the epicenter of both. So very excited about it. I think with our fiber infrastructure business, there's a tremendous amount of upside, and expanded TAM for us in the hyperscaler arena, and I'm sure we'll talk about that today. And on the fiber to the home business, we couldn't be more excited about the progress there.
Not only the industry's recognition of the value of that, of that asset, but also the progress that we're making with Kinetic, and so to your question, we're progressing well towards closing of our merger. We've got seven of the 16 required regulatory approvals. Don't see any issues with achieving the others, and we're also making progress on the initiatives we talked about achieving between signing and closing. In addition to getting regulatory approval, we talked about the two businesses continuing to execute, and we're executing well. Both businesses are putting up solid growth, and we think looking in next year, we'll see solid growth at both companies. We also talked about our balance sheet and how we would focus on the balance sheet between signing and closing.
We started out with two separate debt silos when we announced the transaction, but we talked about having a clear path to collapsing those silos. And we're working actively on that and expect to have that buttoned up, frankly, by closing. And we'll talk more about that as well, I'm sure. We're also working actively on the build plan at Kinetic. You know, right now, Windstream has a plan to build to roughly 45% of the footprint, and we've talked about expanding that to close to 65% with the, with the expanded build. And so we're very excited about the progress we're making there. So 65% fiber coverage of the footprint.
And of course, we're we're we're, we never we never sleep on M&A, so we're always thinking about M&A. So all that to say, long-winded answer, Bora, but we're making great progress towards closing.
And speaking of M&A, Verizon announced the acquisition of Frontier. By our math, at an EBITDA multiple of roughly nine times or so. Any thoughts on the transaction, and any sort of read-through to new Uniti?
Yeah, we were excited to see it. You know, I think it's sort of what I've alluded to at the beginning about the strategic merit of this, of this combination and the strategic value of the fiber to the home asset. I think it's when you talk about convergence, which the carriers are all talking about, you've got the three big wireless carriers all investing heavily in fiber to the home. That, in and of itself, says something, right? And I think that we can debate the merits of convergence, whether, whether it's actually needed or not, but the reality is that the three carriers seem to believe that it has merit, and therefore are proceeding down that path.
And when you think about getting fiber to the home coverage around the country, there aren't very many scale platforms out there in order to achieve that. Of course, Kinetic is one of them, and so we view it as an increasingly strategic asset as a result. And with respect to the valuation, you know, our combined business trades roughly in a six to six-and-a-half times range right now, which we've always said is is really low compared to what we think the strategic and intrinsic value of the business is. And we've even put out our own view of what a sum of the parts analysis could look like for the new Uniti.
And we sort of said, you know, in the seven to 10 times range for a fiber to the home business, and with one that's more fiberized, like Kinetic will be over the next several years, it's more in the eight to 12 times range. And so that 9 times multiple is not surprising to us and maybe even on the low side compared to what we, what we think true intrinsic value is.
Okay. So we made it to question number three before talking about AI. So how do you see AI playing out for fiber infrastructure needs along the path from training to inference, and how can you capitalize on that opportunity?
This will be repetitive to all the folks we've met with this morning talking about AI. But, yeah, lots of questions about it, lots of enthusiasm and excitement about it. And we're trying to be balanced, but we also do view it as a sizable opportunity, and we're working on a way to sort of size that for investors better. But if you think about the CapEx that the hyperscalers are spending is over $200 billion a year, which is up 20%-30% from last year, and we think that growth rate is gonna continue based upon everything we're hearing. And as I've said to folks, those businesses all run on the internet. We don't have to talk about just AI. It's really a broadband-driven strategy.
It's about eyeballs and clicks, and so the vast majority of the capital that they're spending is to enhance broadband. It's to enhance eyeballs and clicks on in their businesses, and yes, AI is one of those use cases, and we're big believers in AI as a use case, but there's also lots of other use cases, to the point where many of the CEOs from the tech companies have even said: "If we don't use this capacity for AI, we're going to use it for other things." As an infrastructure company, we're excited to hear that, and so you think about that $200 billion, and you work down, you put a percentage of that is related to digital infrastructure, and a percentage of that is related to fiber.
No matter how you cut it, fiber companies, including and especially companies with the scale of Uniti, are gonna benefit, we think, materially from that. And so right now, we're in this building out the learning phase, as you said. It's hard to point to how the learning phase impacts the the vanity metrics like revenue and EBITDA and bookings, as I say, because those deals are much more bespoke. And there is a construction element to a lot of those deals. But we treat those deals like anchor awards, where we're using hyperscalers to help us build out strategic components of our network.
And when you're able to do that, and you're able to allow for lease-up capacity on top of those builds, it's really a terrific economic outcome for companies like us. And so we're very excited about this learning phase of AI, 'cause I think it's gonna give us the ability to use those anchors as ways to expand the strategic value of our network. The really exciting phase, I think, is inference. And that's when I think more and more people will be using AI, and that's when I think you'll start to see the effect on MRR.
And that's when having the roughly five million connected endpoints that we have in our network, whether it be through fiber to the home, or connected towers, or connected data centers, or connected small cells. That's when you know, inference is being used at the edge. I think that's when network providers that have those edge connections are gonna truly benefit. I think that'll be in the next several three, four, or five years.
Okay. From an operator's perspective, do you have a preference for IRUs versus dark fiber leasing on?
Yeah. I have a strong preference for dark fiber. I think dark fiber is a, is a sizable business in the U.S. It's growing at somewhere between 5% and 10%. Uniti is a small participant in that or has a small share of the dark fiber market today, but our our share is growing at 20%-25% a year, so we're a share taker, and in order to be a share taker in dark fiber, you have to have a large embedded network, which we have.
And and so to that point, a lot of the dark fiber that we sell today is existing capacity, and so we're selling it with very little capital required to to to turn it up, and also very little OpEx associated with with it. So comes with high margins, low CapEx. So I love dark fiber and think it's a terrific opportunity for us on a go-forward basis. But we also sell lit capacity, and when customers are are are desirous of lit capacity, we have the ability to do that. And one of the things that's gonna be a big synergy of our combination with Windstream is not just bringing together the OpCo/PropCo of Kinetic. I mean, I think that's very, very widely understood.
But one of the parts of that transaction that's not fully understood yet and doesn't get enough attention is bringing together Windstream's wholesale business, carrier-like wholesale, with our fiber business, because you're putting together two very complementary businesses. Ours is mostly dark fiber today, and Windstream's is mostly lit fiber, and so those two businesses will complement each other and I think expand the market opportunity for both businesses, on a combined basis.
Okay, you just answered one of my questions about revenue synergies. Okay.
I'm happy to come back to that, so.
No, that's great. So, you've talked about a sales funnel. Can you talk a little bit about the customer diversity, range of use cases, perhaps specific verticals that dominate?
Yeah. So we're we're-- Uniti is predominantly a wholesale provider, so 90-plus% of our business is wholesale. We talk a lot about enterprise because it's a good, growing business. It grows at 10 to 15% a year for us. It's really an important part of our lease-up strategy. It's how you drive incremental economics on a network. It's how you sweat the asset, as we say. But really, our bread and butter is wholesale. And again, that's why the Windstream wholesale business is such a good fit with our strategy. So 90-plus% wholesale, and so when you look at the funnel, one of the reasons we really like the wholesale business is because we're diversified across all the different use cases for fiber.
Back to what I was saying earlier, whether it's AI, or mobile broadband, or fixed wireless, or fiber to the home, or frankly, even cable, you know, those last mile delivery mechanisms, in many ways, we're agnostic as to which of those are successful and which of those are the most most dominant going forward, because as a wholesale provider, we benefit regardless, right? Because all that traffic gets backhauled onto a wholesale network. And so as a result of that, to your question about our funnel, you would expect our funnel to be diversified across all those different use cases, and that's the case.
So it ebbs and flows, but if at any point in time, roughly 20% of our funnel is made up of wireless carriers, roughly 20% is hyperscalers, roughly 20% is fiber to the home providers who are procuring backhaul to support their their fiber to the home business, and roughly 20% is international ISPs procuring capacity in the U.S. So and then the rest is just sort of cats and dogs, and we like it that way. That you know the wireless carriers have been spending less the past couple years, as we all know, in the digital infrastructure industry, and it's had virtually no discernible effect on Uniti's business, because we're diversified with all the other use cases that are making up the difference.
So we're always looking to grow the funnel, and that's important, but also very important that we keep it as a diversified funnel.
Okay. Does that include—I think, you had mentioned that there's a broadening set of customers that's increasing your TAM. Can you talk a little bit about that?
Yeah, yeah, I think it's definitely the hyperscalers. You know, the hyperscalers have been customers of ours and other fiber businesses for years. I mean, we can if you go back three, four years, five years, we've been selling capacity for for some time. But really, towards the end of last year, and certainly coming into this year, the opportunity set has expanded materially. And I've said publicly that going into last year, we had roughly 10 opportunities in our sales funnel related to hyperscalers. Today, we have close to 100, and these are not small opportunities, right? These are tens of millions, hundreds of millions of dollars of opportunity. And so that alone just gives you a sense for the amount of increased opportunity that we've seen from that, from that group.
Like I said, it's partly driven by AI, but not entirely. I think it. There's there's many other use cases that support that digital infrastructure spend.
So when you're actually competing for some sort of a large hyperscale type requirement, like, why do you win? Is it time to market, existing conduit, something else?
Yeah, good question, and certainly one we're getting a lot. And and I'd say the short answer is the reasons we win are not that different than the reasons we win in other cases, in other cases where people are procuring large network solutions. But let me come back to that. I think one thing that's important to point out, the hyperscalers are are a different type of customer than the wireless carriers, for example. The hyperscalers are more inclined to pay larger NRCs-
Mm-hmm.
And take smaller MRCs as a result. And so whereas the wireless carriers tend to prefer smaller NRCs and larger MRCs over time. And so as a result, the economics on transactions that we and other digital infrastructure providers do with hyperscalers bring a lot of capital in upfront to help fund network builds, whether it be data centers or or fiber builds. And then the recurring revenue, as a, as a result, over time, is lower. But. And I think there's been some confusion about, well, are those deals worth doing or not-
Mm-hmm.
As a result? The answer is: they absolutely are if if that build is strategic to your network. So for us, for example, we announced the deal a few weeks back, where we're building a long-haul route between Mobile, Alabama, and Montgomery, Alabama, two very strategic markets for us. Very high NRC, very low recurring revenue associated with it, but it's a very strategic route for us that we have tremendous lease-up potential. And so the recurring revenue benefit will come over time as we lease up that asset. So I wanted to point that out. And the ultimately, I think that type of demand is going to, is going to continue, and I think over time, you'll start to see the the the lease up potential associated with that.
We've talked about bookings, and which doesn't fully capture your entire sales funnel necessarily because there are different types of services. So a peer of yours has talked about $5 billion in contract value. Is there a similar number that you care to share, or...?
We do have a number. We're not prepared to share it. We haven't shared that publicly, and and I don't think we will anytime soon. Partly because it's, there's different ways to talk about the funnel. You can talk about contract value, you can talk about MRR. We think about it as MRR because it's the leading indicator that that results in bookings, and we we we do report bookings. But we try to be transparent about the fact that just because bookings may be elevated one quarter and may be lower another quarter, that doesn't necessarily, it's not the perfect leading indicator of future business.
And so for example, in the first quarter of this year, bookings were a little bit lower than normal, but we had a very large hyperscaler deal in the first quarter that investors won't see the effect of that deal until maybe late this year, but probably next year. And you'll see it in revenue, and we'll call it out as a little bit of a lumpy impact to revenue, but it had no effect on bookings. And conversely, we had a sizable quarter of bookings this past quarter that was the result of MRR hyperscaler deals as well, but they were largely MRR-driven deals. And so they did show up in bookings, and they will show up in recurring revenue later this year and early next year.
And we try to call those out so that the people can see it, because I do think over the next probably year, year and a half, we're gonna see continue to see that lumpy opportunity set coming from the hyperscalers.
Great. So on CapEx, how do you see that playing out over the next few years, and what are you assuming for BEAD within that? Do you intend to fund BEAD, driven CapEx acceleration? Just talk a little bit about that.
Yeah. So with our two fiber businesses, our fiber infrastructure business, which is our current Uniti business, plus the Windstream Wholesale business, I think capital intensity there will continue to come down. That trend has been one that you've seen over the past few years with us. We've we've really evolved from being a greenfield builder now to more of a lease-up-focused fiber business by design, and so that capital intensity is gonna continue to come down. Really, the the the bulk of the capital over the next few years is going to be spent in the Kinetic vertical, where we're building fiber to the home.
And when we announced the transaction with Windstream, we were very deliberate about saying that our current build plan, inclusive of the incremental one million homes, was fully funded with liquidity on the balance sheet, plus cash flow from the business in the outer years. But we also said that if we wanted to bring that build in, if we wanted to accelerate that build, that we might raise incremental capital to fund that build. And that's exactly what we're doing. We're actually in the market right now, raising capital to refinance some debt, plus some incremental capital to bring that build in. And we're gonna give more clarity on the size of that acceleration and the expanded build plan in the coming months.
But I would say that we're, I would say we're emboldened by industry themes and trends and the performance of the business, and that's leading to us being more bullish on the build plan.
So to touch on the balance sheet a little bit, Windstream announced a consent solicitation for some of its 2028 notes, which included some amendments to modify some covenants. So does this allow for the collapsing of the debt silos? Can you touch on or is there something else that needs to be done? It sounds like there's more to be done. If you could just kind of give us a general sense of what's left.
A little bit, but not much.
Okay.
So, when we announced the transaction, we talked about having the dual silos as a way to help facilitate getting the transaction done. But we were also very candid that we wanted to collapse those silos as soon as we could and completely eliminate the MLA relationship between those two two silos. And I would say subsequent to announcing the transaction, we really heard from the market, both equity holders and creditors in our, in our existing silos, that they wanted that to happen sooner rather than later. That would be good for the story, reduced complexity. And and so we heard that, and we're acting upon it. And so to Bora's question, last week, in partnership with Windstream, there was a consent effort to among Windstream's bondholders to allow that silo to be collapsed.
And and we're currently in the market today, refinancing Windstream's term loan, and as part of that, that debt will also be portable into our silos. So as a result, with with one remaining consent required, we will be able to collapse those silos at closing, and we're. We couldn't be more happy about that. We're very excited about that. The one remaining consent will be from Uniti's revolving credit facilities, lenders, and we're confident that we can get that done. So, I think that this is a result of us having a plan, but also the market encouraging us to be more proactive, and us listening to the market and working with the market to actually make it happen.
So I think it's good for both us and for our various securities.
For new Uniti, you're targeting about $100 million in OpEx savings. How do you see that getting layered in over the course of the years, and can you talk about the sources of it? Is it more gross margin or SG&A?
Yeah. So we, we talked about cash flow synergies in our combination with Windstream of roughly $100 million of OpEx synergies and somewhere between $20-$25 million per year of CapEx synergies. On the OpEx synergies, I think they'll all largely hit the SG&A, I mean, the EBITDA margin line. So these are, there's some headcount redundancies in there, but most of those savings are going to be network related. So when you look at Uniti's business today, we're very much an on-net infrastructure-focused business. Windstream's wholesale business is mostly off-net, so using fiber from other providers. By combining the two, we're gonna be moving most of that business on-net. And so as a result, there's gonna be operating savings as we move off of other other network providers' networks.
And that'll take several years to materialize. I think we've talked about those savings coming in over a three-year period ratably.
Mm-hmm.
That's still our expectation. In addition to the cash flow savings from those synergies, there's really a lot of benefit to being a true owner having true owners' economics and selling and operating a business on an own network. I think we'll also see benefits in the key KPIs of the business as a result. Churn should be lower as a as a result, and we'll see the benefits of that over time as well.
Okay. And then, lastly, on the managed services business, you've mentioned that you think of that as being non-core. You might explore potential disposition of it. How should we think about what the market demand is for that? What sort of valuation you might be able to get, as well as are there other non-core assets in the combined company that you could potentially monetize?
Yeah. So Windstream has a business they call their enterprise business. We've started calling it the managed services business to help distinguish it from our enterprise business. And it's a,it's a good business. It's roughly $1 billion of revenue. It's mostly an an an enterprise cloud-based business, SD-WAN, UCaaS, SASE, security offering, and it's mostly off-net. So it's it's not core to our fiber infrastructure strategy, to be clear. We have, we're focused on the infrastructure, we're focused on simple products, but that doesn't mean it's not a good business, because it is. I mean, the Windstream team has done a terrific job with that business, and it's very strategic to a number of the large enterprise providers around the country, like the cable providers, for example.
So we've talked about that business as being non-core. We've talked about a valuation of that business being roughly a 3-5 times EBITDA multiple. And so, you know, there's about $300 million of EBITDA in the business. So think about just or call it $200-$300 million of EBITDA in that business. So you can think directionally about the value that we would place on it. We haven't talked about selling it. But we've signaled that we think it's non-core, and and that's how we're currently viewing it. I think with respect to your question about other non-core assets, I would say no, I don't think there are, but I also would say that if if you want some interesting reading, go to...
Go look at the proxy that we filed and look at the background reading. There's tons of interesting background in there about some of the strategic conversations that both we and Windstream have had over the years, and you can see that we've been very active in those conversations. There's some pretty interesting value metrics, actually validating value metrics, in there. So the point of that is, even if something is core, that doesn't mean that we're not, in the interest of our shareholders, buyers and sellers every day, so.
Great. I think we'll leave it there.
All right.
Thanks very much for your time.
Great. Thank you, Bora.