All right, great. Thank you very much. Sorry for being late. If you're going to change the room on the schedule, you ought to let the analysts know. But didn't find out about that, so I apologize. But hey, my name is Frank Louthan. I'm the Senior Wireline Analyst here at Raymond James. We're very pleased again to have Dave Schaeffer, CEO of Cogent, back here at the conference. Been a long-time attender. So Dave, maybe talk to us a little bit about Cogent. Tell us about what you do, what the business is, and kind of how you fit into the space.
Sure, Frank. Well, thanks for hosting me. Thank Raymond James for a great venue. I'd like to thank all the investors for hanging around, waiting all day to hear what we have to say. We obviously couldn't have started without you. It's hard to go do a presentation without your moderator. Cogent is a provider of dedicated internet access service. We operate a global network, serve 54 countries, about 81,000 route miles of fiber between cities, an additional almost 19,000 miles of metropolitan fiber in 230 markets. In addition to selling internet connectivity to either service providers or corporate end users, we also sell VPN services on top of the internet, whether it be VPLS or MPLS-based services. We also sell colocation. Recently, we've begun to sell wavelength services, meaning optical transport services, across the network that we recently acquired from Sprint.
All right, great. So a lot's changed in the last 12 months since we were here. You completed the Sprint transaction. You got a few months of operating experience with that under your belt. Kind of walk us through kind of where Cogent is today and how is that different from the Cogent that folks knew from many years?
So Cogent has been a public company since 2005. The company was founded in 1999. And believe it or not, Cogent has pretty much executed the same business plan. So when Cogent was started, we were lucky enough to raise $500 million of initial seed capital and not spend that capital. When the .com crash occurred, we went out and bought a total of 13 companies. Six were public, seven were private. We dismantled those companies. We typically terminated the customers. We fired the employees and repurposed those assets to build the original Cogent network. That network was an all-IP over DWDM network, protected at Layer 3, using Ethernet as the customer interface. That business grew organically for 18 years at a compounded rate of over 10% year-over-year and de-levered about 200 basis points a year of EBITDA margin expansion.
When the pandemic hit, our growth rate slowed because our focus market for corporate users were the central districts of major cities and skyscrapers. Our growth rate fell to about 5%, and our rate of margin expansion moderated to about 100 basis points from 200. We had the opportunity last year to acquire the Sprint Global Markets Group network, the original Sprint Long Distance network. When we acquired that network, we acquired 19,000 route miles of inner-city fiber. We acquired another 1,200 miles of metro fiber, 482 fee-simple-owned technical buildings comprising 1.9 million sq ft and about 230 MW of power. We acquired an enterprise customer base that was losing $1 million a day. We were paid $700 million in cash by T-Mobile to acquire that customer base and that enterprise business. We bought the physical network that cost $20.5 billion to build for $1. We are in the process of interconnecting the two networks and repurposing that acquired long-distance fiber optic network to sell optical transport services. So a lot has changed.
All right. So when you look at the opportunities that that Sprint network brings obviously, talking about a huge fiber network that we all know and love, what are some of the top two or three opportunities there that you see from that network for Cogent?
The first opportunity is to connect that network to 800 U.S. carrier-neutral data centers where we sell IP services today. By adding optical transport services, we now reach an additional $2 billion addressable market because we will have the most ubiquitous footprint of data centers, the quickest provisioning time, the most accurate description of the physical route that the fiber traverses, and a low price point. We believe we can capture 25% of that market over a 7-year period with that market growing at about 7% a year. That is an exciting opportunity. It has been made possible by that acquisition of that enterprise business, the subsidies from T-Mobile, and the ability to take costs out of that operating business and turn it into a modestly profitable, non-growing, stable business.
I think the second exciting opportunity for us is the repurposing of a subset of those phone central offices into data centers. We will have about 230 MW available, not the same 230 MW that we acquired. We've identified 45 of the 482 facilities to be suitable for data center conversion. We are connecting those to our metro networks. We're going through and depopulating the obsolete telephone equipment and repurposing these facilities to sell colocation, either on a retail or wholesale basis. We had 55 data centers prior to the acquisition. 53 of those were in leaseholds. Two were in fee-simple-owned facilities with 69 MW of power and 634,000 feet of raised floor space.
We are now going to add to that 45 additional facilities, all owned, an additional 1.3 million sq ft and an additional 170 MW of power, s o giving us about 235-240 MW, some of which will be retained for our own needs. But we anticipate having about 170 MW available for sale to third -parties. We also have an opportunity to take our IPv4 address space. We control about 37.8 million addresses. Today, we're leasing out about 11.4 million of those addresses. In the Sprint acquisition, we acquired another 9.9 million addresses. So the 27 million we had plus the addresses acquired give us a large footprint to generate incremental cash flow. So some additional products, some additional locations, and an additional product set, all exciting opportunities.
All right. Let's talk about those a little bit more. First, on the wavelengths, I think that's one of the bigger opportunities. Walk us through what does a customer use a wavelength for? And what are they doing with that versus what should they use your traditional network for?
So Cogent was founded with the premise that the internet would be the only network that mattered. That has proven to be predominantly correct. However, there are use cases that pay a premium for a wavelength service. So the cost to transmit a bit a mile is significantly cheaper on the internet. A wavelength is 2.5 x more expensive. But there are three attributes of a wavelength service that cannot be replicated on the internet. The first is it's deterministic. So you know the exact latency between any two endpoints. The second is there are no limitations on packet sizes. So it is very efficient for moving large packets of data. And third, it's completely secure, not available or discoverable on the internet. For that reason, there are four primary use cases.
Any company that is looking to replicate data between facilities may elect to pay that premium and use wavelengths. Any company that operates isolated regional networks that are looking to link those networks together into a homogeneous network will use wavelengths. Governments will use wavelengths. Certain select enterprise customers will use wavelengths because of the added security. It is more expensive, but it is a parallel market. It's one in which we today already have relationships with virtually all of the buyers.
OK. So who are some of those typical customers? And who are some of the customers that you already service that are taking wavelengths? And how does that make it easier to try and sell into that base?
Yeah. So on the content side, think of any of the hyperscalers, whether it's Microsoft or Amazon or Meta or Google. All customers we sell transit to today, Netflix, are candidates that could use wavelength services. In fact, have demanded them in the market from our competitors. On the access network side, it's companies like Altice or Cox or Spectrum/Charter that have isolated markets, overbuilders such as Frontier or Metronet that may have disparate markets they're linking together, government customers, some of which I can't name, also would use wavelengths. And then on the large enterprise side, it could be organizations like Bank of America or JP Morgan that could justify building a wavelength network.
All right. So let's talk about pricing because you've been really clear. Historically, you're the low-cost provider. You've used pricing before with your products. How should we think about pricing in the wavelength market and what you might do there?
So in the transit market, which is sold from a singular point and is more commoditized, we have built our business by undercutting the market by 50%. That has allowed Cogent to become the largest provider of internet transit in the world, carrying about a quarter of the world's traffic. We continue to gain share in that market. In the wavelength market, we're basically nonexistent today with about $13 million run rate against a $2 billion addressable market. Our advantages come from the uniqueness of our routes, the ubiquity of our endpoints, the ability to accurately detail that and provision it quickly.
We will use price b ut price is not as much of a commodity function in the wavelength market. Our expectation is we'll probably discount somewhere between 10%-30% to current market rates. But it'll be on a route-by-route basis as we do price discovery in the market. A wavelength is sold knowing the two endpoints, the distance, the contract term, and then the size of the connection. So it is a three-dimensional pricing grid.
Okay. All right. Great. So you also talked about the data centers. And let's talk about that. We get a lot of questions on this with the capacity that you're adding, especially given the constraints we see in North America with data centers. How much of that do you think you can satisfy? And then talk to us about what kind of investment you're going to need to make to upgrade these facilities, to bring on this additional available power.
Yeah. So Cogent had data centers that it did not build but came to Cogent through acquisition. Those facilities have about 125 watts a square foot, 69 MW, 634,000 sq ft of space. We had typically leased those out on a rack-by-rack basis. We also usually have Cogent sales offices collocated at those facilities, as well as our technical equipment. We acquired 482 technical buildings, most of which are too small in the wrong location to be viable. However, 45 of those facilities comprise 1.6 million sq ft and about 170 MW of power. We are converting those to data centers, so to things we are doing, one, connecting those buildings to our metro footprint because they typically only were connected to the Sprint footprint. Well, we have to do that to support our wavelength business.
Two, we have to go and depopulate abandoned telephone equipment from these facilities. There are approximately 22,500 cabinets of old telephone gear that have to be removed. We've already removed about 14,500 of those cabinets but still have more to do. We also need to convert these facilities from negative 48 DC power to AC 120. So that's install inverters. The pure capital cost to do that will be $7 million, $8 million. These facilities will typically average between 100-125 watts per sq uare foot. And we will keep one portion of the facility for our technical equipment and a retail colo. And the majority of these facilities will be monetized by leasing them out on a wholesale basis.
Okay. And I want to say hello to anyone that's searching this transcript for AI. This is the asking AI question here. As we got to touch on that at an investor conference, as we look at Cogent and the networking aspect, how do you see Cogent fitting into AI and the current demand and as it's going forward in the future?
So I'm always reluctant to focus on buzzwords and things that are in vogue at a given point in time, whether it be cloud or edge computing or AI. But AI is real. And it does impact networks in two very discrete ways. First, AI only exists because of the sheer volume of data that's been collected at virtually zero cost over the internet. That is going to continue and create a virtuous cycle on internet traffic demand. The second thing is the training of those AI models occurs at multiple data centers that need to be replicated. That becomes a demand set for waves that is incremental to the four segments that I described earlier. So I think we are a beneficiary on both the internet and wavelength side. I think AI, unfortunately, is suboptimally named. It is a real technology of pattern recognition and predictive inference from those patterns that will transform how we do business, how we communicate, how we live.
Okay. So from the newest thing on the network to the oldest, talk to us about the corporate business. What are some of the hurdles there that have kind of been impeding the growth? And is it return to office delays? How should we think about that?
Yeah. So our corporate business had grown consistently for about 15 years at 11% year-over-year. The pandemic hit. People went home. And our growth rate went from +11% to -9%. As people have gradually returned to offices, they have not returned full time five days a week. As a result of that, the corporate business has recovered. Last quarter, it grew 5.1% sequentially. But that was a bit misleading because some of that was a tax benefit. If you net out that tax benefit, it grew at about 0.8% sequentially or annualized, a little under 4% year-over-year. So we are seeing recovery. But it is slower than we had hoped for.
OK. Great. And you sort of mentioned pre- and post-pandemic. Talk to us about your sales force. So for many years, you had a very predictable model, the way you ran your sales force and how they operated in a couple of different basic buckets. How has that changed? And are you having any issues finding salespeople versus what you used to?
It actually has not changed very much, Frank. Today, we have about 650 quota-bearing reps. Those quota-bearing reps sit in 50 offices around the world. They outbound telemarket. That has been Cogent's model from day one: more reps, more offices. We do now have an incremental customer segment in enterprise accounts. We only have about a dozen reps that focus on those enterprise accounts. We did acquire 51 reps from Sprint. Many of those have left the company as they were not comfortable with this type of outbound hunting model. But in general, we've continued to see good success out of our sales force. Unfortunately, it's a model that results in high sales force turnover. Anybody who's ever done cold call marketing knows how hard it is. We generally churn about 5.7% of the sales force a month.
Okay. So talk to us about the Off-Net business. What are you seeing there? Any changes? And then how's the sales force going about selling to those businesses, especially now that you've got a lot of new network that maybe you can consolidate some of that?
Well, we have more enterprise customers that inherently need more Off-Net services as well. So previously, Cogent had about 25% of its revenue from Off-Net sales, about 75% On-Net. The acquired Sprint customer base was 93% Off-Net and 7% On-Net. So within our net-centric business, it's about 90% On-Net, about 10% Off-Net, where we have tail circuits to support those customers. In the corporate business, it's historically been about 60% On-Net in those skyscrapers and 40% Off-Net. And then in the acquired enterprise business, we are migrating customers from Off-Net to On-Net. And we think we'll end up at about a 50/50 mix in those enterprise customers. So in total, we are going to increase the profitability by bringing Off-Net traffic On-Net.
Okay. Great. All right. Pause for a minute. See if anybody has any questions from the audience. Nope. All right. No problem. So talk to us about your liquidity position, where you are today. You're receiving payments from T-Mobile. They start to decline this spring for a while. Talk to us about the potential sources of funding, including possibly ABS or selling your IP addresses, something like that. Where do you see your liquidity coming from?
So first of all, let me comment on some confusion around our last earnings call. Our EBITDA for the quarter was $110 million. I think the inference was made that with the acquired Sprint negative EBITDA combined with the Cogent EBITDA, there was an implied view that our EBITDA declined. Our EBITDA was artificially retarded by the fact that we paid out $17 million in employee severance in the quarter that was fully reimbursed by T-Mobile. So rather than the $87 million we would have normally received in the quarter, we actually received $104 million. We did not want to dwell on that on the earnings call due to the large number of Cogent employees that participate in that call. Getting to our liquidity, we ended the quarter with $113 million of cash. We are doing about $110 million in EBITDA.
In the upcoming quarter, I think our numbers will be similar. We will not have that reimbursement and the offsetting expense. The net cash will look about the same, meaning $87 million rather than $104 million would have come from T-Mobile. That number will step down mid-year because our subsidy payments go from $29 million to $8 million a month. Now, during that time, we are continuing to take costs out of the business. We reduced aggregate headcount by 200 employees since closing. We have consolidated or closed facilities. And we've migrated traffic from Off-Net to On-Net. The stream of payments from T-Mobile will continue through mid-year 2028, so a total of 54 months from closing. That gives us a great deal of cushion. We also will be selling wavelengths which carry effectively 100% gross margin and 95% EBITDA margin.
In addition to that, we have three weighted assets on the balance sheet: IP addresses, data center space that we talked about, and dark fiber. We are looking at all three of those sources as additional ways to raise capital. Not that we need to raise capital today. But we have that optionality. Our EBITDA increased last year to $352 million, up from $260 million the year before. We de-levered by nearly one turn of EBITDA year- over- year on a net basis. We expect that de-levering process to continue. With that inherent de-levering, we may look to either monetize the stream of revenue from our IP addresses through a securitization. We may look to monetize some of the unsold and unleased addresses through a sale, as well as looking at dark fiber and our wholesale colo product.
As we look at 2024 and 2025, even with those decline in payments from T-Mobile mid-year 2024, we anticipate EBITDA will be in the same ballpark. While the company doesn't give exact guidance, with roughly $360 million of EBITDA and a gross debt of $950 million of bonds and about $400 million of capital leases, we're effectively about 3x levered on a pro forma basis, which is right in the middle of our range and substantially down from our peak. Also, remember, we have a dividend and have now grown that dividend for 46 sequential consecutive quarters.
Kind of finish on that, thoughts on dividend growth going forward?
Yeah. I think we'll have the ability to continue to grow our dividend, albeit at a more modest pace than we were pre-pandemic. As the corporate business begins to reaccelerate, we could reevaluate when we would accelerate that. We also have these alternate forms of monetization. But I think we are comfortable in forecasting that our ability to grow the dividend will continue beyond the 11.5 years we've done it. That's a pretty consistent track record.
Great. All right, Dave. Hey, thank you very much for being here. Really appreciate it. We got a breakout session afterwards.