Good morning. My name is Sebastiano Petti, and I cover the communications sector here at J.P. Morgan. It's my pleasure to introduce Dave Schaeffer, founder and CEO of Cogent Communications. Dave, thanks for joining us today.
Hey, thank you for inviting me, Sebastiano. I'd like to thank all the investors for taking time to hear a little bit about Cogent, and as always, thank J.P. Morgan for a great venue.
Awesome. Thank you. Dave, Cogent reported mixed one-key results last week, but.
You're kind.
Yeah. But you increased your long-term revenue growth target to 6-8%, margin expansion target to 150 basis points annually. We'll certainly dig into the details of the quarter and business trends momentarily, but can you unpack what underlies the confidence in the improving long-term growth trajectory?
Yeah. I think there are two components to that growth trajectory, the first being top-line revenue growth, the second being margin expansion. On the revenue growth side, we are near the end of the burn-off of revenues that we intentionally wanted to exit as part of the Sprint transaction. Just to remind investors, when we acquired the Sprint Global Markets business from T-Mobile, it was really two conjoined transactions. We acquired a money-losing services business that was doing $565 million in revenue and burning $1 million a day in cash. We were paid $700 million to take that business. When we took that business, it was experiencing a 7.4% compounded annual decline in revenues for the three years preceding the transaction. We actually did two things to that revenue trajectory. One helped and one hurt.
The first thing we did was tried to terminate services that were gross margin negative. Many of those services were governed by contracts. While we have exited the majority of those, there is still a tail. We also were able to go into the revenues that we wanted to keep and increase pricing on those services. Cogent's revenue growth coming into the transaction had decelerated due to the vacancies in offices as a result of COVID. While Cogent's organic business for 18 years had a compounded top-line revenue growth of 10.2%, all organic year-over-year, at the point of the transaction, the 12 months previous to the transaction, we were growing at 2.1%. You blend those two growth rates together, we would have a negative 1.8% growth rate, which is a little bit worse than what we have actually done since the transaction closed.
With the end of those undesirable revenues, we should return to positive top-line growth by the middle of the third quarter of this year, nearly a year earlier than we expected. The second thing on revenue is the demand for wavelengths, which is the second part of the transaction. In acquiring GMG from T-Mobile, we acquired an operating business, but we also acquired a network that was effectively fallow. It was a long-distance voice network that was built at a capital cost of $20,500,000,000 , and there was virtually no revenue on that network. In fact, 93% of the revenues that we acquired never touched the Sprint network. For those 7% that did, it was usually only at one location. We were able to repurpose that network to sell optical transport or wavelength services.
The demand set for those services turned out to be stronger than we anticipated, in large part because of AI, which is a key incremental driver of optical transport demand. As a result, we were comfortable in adjusting our revenue growth rates for the blended company from 5-7% to 6-8%, or about a 100 basis point improvement at the midpoint, still lower than Cogent's organic growth rate had historically been, but in line with what we anticipate the blended company to be able to deliver. In terms of margin expansion, we have actually exceeded 300 basis points a year in the two years that we have been operating the combined business, far better than the 100 basis points that we were projecting.
With the high contribution margins of wavelength services, we feel comfortable that the combined company should be able to deliver 150 basis points, still 70 basis points below where Cogent's contribution margins and growth rate had been prior to the acquisition. This was just an acknowledgment of the learning that we've had post-closing.
Got it. Okay. That's helpful context there. As we think about the first quarter from a Wave's perspective, revenue and installs came in a little bit light, with the installs concentrated towards the back half of the quarter. I think you attributed the shortfall to customers falling out of the backlog or just others perhaps not being able to accept the new services. Can you maybe double-click on that for us and help unpack some of the drivers there?
Yeah. Sprint had not been in the wavelength business. Its network was designed to support its long-distance voice business, which had been turned down in 2015. When we acquired Sprint, we acquired the network assets for $1, and we outlined a program to convert those assets into something that would generate demand in today's environment. That included two things: repurposing the physical fiber network for optical transport or wavelength services, the second being repurposing a subset of the central offices into data centers. For the wavelength opportunity, we needed to extend the Sprint network to touch carrier-neutral data centers across North America. The Sprint network terminated in 23 data centers. We have expanded that footprint to 883 data centers. Sprint did not have a metropolitan fiber network. We had to repurpose the physical long-haul fiber and then integrate it into our metro footprint.
When we announced the deal in September of 2022, we closed in May of 2023, and we outlined a program that would take till the end of 2024 to enable 800 data centers. We actually ended up with 802. We had deployed about 1,000 wavelengths on a custom-installed basis between closing and the enablement of the network. In that period, we garnered about 4,000 potential orders from customers starting in September of 2022 and extending through December of 2024. Most of those orders fell out due to the fact that customers could not wait that long, and we were unable to provision services. As we got clear line of sight to being able to deliver those wavelength services, we started to book new orders. We booked approximately 900 in the fourth quarter, about another 2,400 in the first quarter.
We typically install about 5% of orders in our backlog per month. We have developed the provisioning capabilities to be able to provision 500 orders a month. As we started to contact customers who had orders, many of them said, "We are still interested, but not ready now." We actually grew our wavelength business sequentially by 18.2% unit number of orders, but only 2.2% by revenue due to the fact that virtually all of that revenue installed the last couple of days of the quarter.
Got it. Okay. In terms of installations, you hope to scale Wave's installs to 500 per month early this year. However, now you expect to convert about 4-5% of your funnel each month based on the call last week, or about, call it 160 Waves or so based on today's backlog of 3,443 Waves. First, what's your confidence in being able to convert the 4-5% of the funnel each quarter? Why not go faster? You have this sizable backlog. What's the limiting factor there?
First of all, we have developed the install capabilities. This is field services and service delivery coordinators to be able to install 500 wavelengths per month. The network architecture that we have deployed allows us to do that install in a much shorter timeframe than industry norms. Today, the wavelength market is about 145,000 wavelengths. It is dominated by two other vendors with average install times of 90-120 days, with a nearly 50% fallout rate from order to installation. We were able to start with a clean site and architected a network very different than that of our competitors. We can install in those 883 sites in a 30-day window. We will ultimately get that installation window down to two weeks.
In the IP business, we can deliver IP services in 1,700 data centers in 57 countries, typically in nine days in an industry that was characterized with a 90-day delivery window. That is on our capability side. For the demand side, customers need to become comfortable that we can meet those delivery windows because they oftentimes have other capital expenditures associated with the use of these wavelengths. We are continuing to take in new orders at a faster rate. We have already installed wavelengths at 329 of those 883 facilities at quarter's end. As we have demonstrated our ability to provision quickly, we are beginning to see a larger percentage of the customers' Wave opportunities. Cogent is the largest provider of internet traffic in the world, carrying about one and a half exabytes a day of traffic.
We have 25% market share in that segment, and we anticipate getting to a similar market share within three years in the wavelength business. We have an existing sales force. We know who the customers are, and what we have to do is convince those customers that we can deliver the way we have outlined we can deliver after those customers have been accustomed to having delays and inability from their current providers. While our wavelength unit growth met our targets, our revenue growth did not due to the fact that most customers did not accept that service till the end of the quarter.
Just sticking with that last point, I mean, you have your ability to provision orders, right, is in excess of your customers' ability to accept them. Just thinking about the 30-day provisioning going down to 17 days, I mean, is that really an area of differentiation if customers are not necessarily accustomed to installing services that quickly, or is it just more about the last point of educating the, not necessarily educating, but getting customers more comfortable with your abilities on the Wave side and the faster provisioning, just trying to think about that area of differentiation and how it materializes from here?
It is absolutely an area of differentiation, and it is something that customers will value if it is available to them. The use cases for wavelengths fall into three primary categories: regional networks tying together islands of traffic, international carriers and content providers extending their networks transcontinentally, and then finally AI training tying databases to GPUs. In each of those use cases, because of the long lead times of our competitors and the inability to actually install 100% of the orders, the customers have become accustomed to living with long lead times, not because they want to, but because they had to. Secondly, oftentimes multi-sourcing, knowing that it was likely that one of their providers could not meet their requirements. If we are able to demonstrate ubiquity of coverage and quick installs with high reliability, that will be valued.
It is probably the way as a new entrant we gain market share. One of the other learnings that we have had in this short period that we have actually been provisioning is the fact that we did not have to, at least to date, be as aggressive on pricing as we thought we would have to. To just remind investors, we gained market share in our primary business, internet connectivity, by guaranteeing to underprice the market by 50%. We shortened the provisioning windows, broadened the number of locations customers could connect to us, and then finally drove down the cost per bit and gained the largest market share of any provider in the world. That process took us nearly 20 years to complete as an unknown company without a sales force and a much broader universe of competitors.
As we enter the wavelength market, we have an established sales force of about 625 quota-bearing reps. Roughly half of those focus on the wholesale customers in the wavelength market. Three quarters of those customers are already using Cogent as their internet service provider. What we need to do is port that credibility that we have on IP to wavelengths. If we are able to do that, we believe that we will continue to gain share without having to use the lever of price. Because we have a much lower capital basis in our network, we do have the ability to be more aggressive if needed.
Great. What underlies the confidence, I guess, in growing the backlog to 10,000 unique Waves by the end of the year and the install rate of about 500 per month? Yeah, help explain. Maybe also as part of that, take a step back and help explain what it means to actually be in the funnel and some of the different moving pieces around that.
Let me start there and then go to the cadence of building that funnel. The funnel is comprised of two types of orders: those that are actually signed, which means there's a specific pair of endpoints, a specific throughput, and a specific price with a specific date to install, firm order commitment date. Typically, customers are allowed to push those dates out up to three times. For orders that fit that category type in our IP business, about 91% of those orders ultimately install. About 9% of those orders fail for a variety of reasons. Usually, it's credit or abuse issues on the IP side. Those are probably less of an issue on wavelengths. It's a more concentrated market. Instead of 13,000 potential customers, there's probably 200 meaningful customers and 200 smaller customers. Less than 500 total customers. The customers tend to be more sophisticated.
The second thing that would be in the backlog is a verbally committed order. That means it has not actually gone to contract execution, but again, the two endpoints are agreed to, the delivery window is not set, and there is a general agreement on a price range, but there is typically a final negotiation to get to a final price point and terms and conditions. Obviously, a much higher fallout rate of those orders that are verbally committed than those that are committed in writing. Now, in terms of the confidence to build that funnel, in the first quarter, this is the first quarter that we could actually install and scale. As I said, we installed roughly 200 orders in the quarter, albeit back-end loaded. We built a funnel of nearly 2,400 incremental orders.
That pace of funnel additions appears to be accelerating due to the ability to actually deliver. You have to build credibility with your customer base. I think, again, we've done that on the IP side, and this is a new product for Cogent.
Great. Maybe shifting gears to just financials and 2025 EBITDA outlook. On the call, you said that you expect to "achieve the goals we've outlined" in regard to the 2025 EBITDA expectations. I guess in the past, you had described that as similar, right, or stable year on year. I guess what gives you confidence, I guess, in an improving EBITDA trajectory sequentially over the balance of the year? You talked about the revenue expectations for the third quarter with sustained growth thereafter. Just help unpack that or tell us more about that.
Yeah. First of all, we have spent the first 25 minutes talking about something that represents today about 2% of Cogent's revenues, and 89% of our revenues are in our IP business, which has continued to grow. We also have been able to demonstrate for the past year a reduction in top line as a result of that attentional grooming that I described, yet a growth in actual EBITDA, not just an expansion in margin, but a growth in EBITDA every quarter. We anticipate that growth to continue. Our payment stream from T-Mobile was bifurcated. In the first 12 months, the subsidies equaled $29.1 million a month, and then for the next 42 months, those payments stepped down to $8.3 million a month or $25 million a quarter.
In the first year we acquired Sprint, we acquired it in May of 2023, Cogent's EBITDA for 2023 versus 2022 went from roughly $260 million to $352 million. That payment stream stepped down mid-year 2023, from 2023 to 2024. As a result, our EBITDA for full year 2024 was $348.2 million. Looking at calendar year 2025, we have a $104 million headwind to overcome from that step down in subsidy payments. To help us offset that, we have been cutting costs, growing EBITDA, and will become total revenue growth positive mid-Q3. We also know that our wavelength revenue carries very high contribution margins. Our on-net services typically have a 90%+ contribution margin. With the acceleration of that business, we will see an acceleration in absolute EBITDA reported. We had said that our EBITDA will be relatively equivalent in 2025 to 2024.
Now, I somewhat soften my statement based on the $68.8 million or $69 million we did in Q1. We need that EBITDA to grow to get to that roughly $350 million range. If our wavelength business accelerates beyond kind of its current pace, we will easily achieve that. If it hits our current projections, we will fall slightly short of that. If we are able to pull in more orders to the funnel and improve the install cadence, we can easily achieve the guidance that we laid out. The underlying IP business, which is selling internet access and VPN services, continues to grow in mid-single digits, absent the non-core services that we are attriting from T-Mobile's acquisition.
Got it. Last week, Cogent slowed its quarterly dividend growth outlook to half a penny from a penny. I mean, why does it make sense to continue to grow the dividend given the leverage concerns and what we're kind of discussing? How are you thinking about, I guess, the return to a more comfortable leverage ratio?
Last quarter, we actually supplemented the dividend with additional purchases of stock due to the stock volatility. Cogent has returned approximately $1.6 billion to shareholders in a combination of dividends and buybacks, roughly $240 million in buybacks and about $1.35 billion in dividends. We've also been able to characterize the majority, and in fact 100% last year of our dividend as return of capital, making it extremely tax efficient for the recipient, reducing their basis, but not paying taxes on that distribution of capital. We have grown the dividend for 52 quarters sequentially. We have grown the dividend because of both the growth in EBITDA and the capital efficiency of our business. We have ratcheted leverage up. The step down in payments from T-Mobile has resulted in our leverage going above the range that we said we would be comfortable with.
We know that leverage will peak in Q3 of this year and begin to decline just because we lapped that step down in those payments. It is an LTM test. If you did it on an LQA basis, our leverage has already been improving for the past two quarters. In our use of capital, our primary objective will initially be to reduce leverage back into the comfort zone that we have outlined, two, to then accelerate the return of capital to shareholders. We have not been a company prone to doing a large number of acquisitions that could potentially destroy value. Rather, we give that money back to shareholders, and that intends to be our plan. Priority one is to reduce leverage. Priority two is to continue to accelerate the return of capital.
Because of the short-term increase in leverage, but the longer-term view of the growth prospects, it made sense to reduce the rate of growth in dividend from one penny a share sequentially to $0.005 a share. We will supplement that with buybacks. Then as we deliver over the next year and a half to two years, getting back down to the total leverage target that we have outlined, we will then be in a position to accelerate the return of capital, and we'll then judge whether it makes sense to either re-accelerate the pace of growth or to be more aggressive on buybacks.
Got it. I guess one area of opportunity as you think about the balance sheet and leverage is the data center monetization efforts. Following the acceleration in the data center conversion program, it sounds as though monetization efforts are progressing. I think you said you're moving towards initial contract negotiations with four letters of intent. However, I think we may not necessarily get a deal announcement mid-year. Can you help us, I guess, give us a little bit more color on why that might be the case and how the complexities of that and timing?
All of the forward-looking statements that I've made do not anticipate the monetization of surplus assets. We actually have three groups of assets that are non-revenue producing today. The first of those are our data center footprint. When we announced the acquisition of Sprint, we acquired 482 Sprint-owned pieces of real estate with 1.9 million sq ft and 230 megawatts of inbound power. Those facilities were primarily telephone central offices spread across the country. We had initially identified 45 of those facilities that we would convert into low-density data centers at minimal cost with no major capital expenditure. We were going to take 10,000 sq ft per facility, so about a quarter of the total footprint, and operate kind of a 1- megawatt, 10,000 sq ft data center.
Looks much like the 55 data centers that Cogent had prior to the acquisition, with the difference that of the 55, we only owned two of them, V Simple, and the other 53 were leaseholds. When we looked at the Sprint footprint and then realized the demand for space and power, the data center market change, we elected to spend $100 million, convert 24 of the initial 45 into much larger data centers with greater power density. We expanded the number of data centers that we would convert to about 125. That process is almost complete now. Of those data centers, we have a total of 109 megawatts in 24 facilities with 1 million sq ft that is completely fallow but ready for use. We do not believe we are equipped to fill those facilities up.
We're either looking to sell them or lease them on a long-term triple net basis. The second bucket of assets that we have that are potentially monetizable are our IP address spaces. We have approximately 38 million addresses. We have leased out approximately 13 million of those addresses. We have approximately 2 million addresses that were given to customers before those addresses had real value at no charge. So 15 million of the 38 million are in use, 23 million sit fallow. Those addresses have changed hands in open markets for between $40 and $60 an address. We are continuing to figure out what is the best way to monetize those. Of the 13 million addresses that have customers, we took 12 million of them and securitized them. We have raised $380 million against the stream of revenue from those addresses and an asset-backed securitization.
The final asset we have that could potentially be monetized is surplus fiber. We typically are using one pair of fibers for Cogent's IP network, and we have abandoned some IRU fiber that was originally used in our network. Secondly, we are typically using four fibers for our wavelength network. The cross-section of the Sprint network ranges from 24-144 fibers, and we would look to potentially monetize by selling off some of that excess fiber. There are three asset categories that potentially could raise capital. These are asset divestitures, not recurring businesses, and they are not baked into our financial projections.
Okay. Dave, that's a great way to end it. Thanks, everyone, for joining us. Dave, thank you for joining us today, and everyone have a great day.
Hey, thanks, Bashar.