I'm very happy to have Dave Schaeffer, Chief Executive Officer of Cogent Communications, here with us today. Dave, thank you for taking the time to be with us. So given the time of year, maybe you can just start off by discussing Cogent's key accomplishments over the last 12 months and the priorities as you look out into next year.
Yeah, so first of all, thanks, Chris, for inviting me. Thank you, UBS, for a great forum. And I'd like to thank the investors in the room for taking time to hear a little bit about Cogent. You know, I think as we look back at 2024, it has been a significant year of transition for Cogent. You know, our core business that we've had for 25 years has continued to grow and continued to improve coming out of the pandemic, but at a very slow rate. A lot of the effort that we have expended over the past year is related to the integration of the Sprint assets into the Cogent business model. And to that end, I think there are four key initiatives that we focused on. You know, first was the conversion of a TDM voice network into an optical transport or wavelength network.
That effort required the physical extension of the Sprint network to meet our metropolitan networks in 110 markets. That work was completed February of 2024. The second was the reconfiguration of nearly 17,000 route miles and 800 metropolitan rings in 110 markets across North America. That work is still ongoing, but is substantially complete and will be complete at the end of this month. Third is the deployment of transponders in 800 data centers in order to accept wavelength services. And then finally, the deployment of about 110 reconfigurable add/drop multiplexers to be able to direct those wavelengths between locations. All of that foundational work will be done this year, enabling us to provision wavelengths quickly between any two North American data centers.
While we have deployed approximately 1,100 wavelengths on a manual basis, our goal is to be able to automate that and cut those provisioning times to two weeks and do that across 800 data centers and 80. That has been our largest focus. Our second area of focus is migrating 12,900 miles of our North American IP network onto fiber that we acquired from Sprint. The Cogent IP network is the largest IP network in the world. It's the largest carrier of internet traffic. It spans 56 countries, 71,000 route miles of intercity fiber , and another roughly 19,000 route miles of metropolitan fiber. We needed to move 12,900 of those miles off of IRU and onto the Sprint fiber. That work is ongoing and will be completed in the first quarter of 2025, allowing us to return that IRU on schedule in mid-year 2025.
The third major initiative has been the consolidation of the operating businesses into a unified platform, whether it be customer care, billing, support, field services, network monitoring. You know, today Cogent operates with one platform globally. We don't have two different provisioning systems. We have one. We have one billing system, one customer care system, one design set of tools, one set of network management. And consolidating two large networks together in a relatively short period of time. We closed the transaction in May of 2023. And by, actually, October of 2023, we had virtually all of our systems consolidated. And we've completed that cleanup in the past year. And then the final piece is monetizing some of the assets that really don't show up on our balance sheet. You know, in 2024, we demonstrated the value of our IP address leasing business.
We did an asset-backed securitization and raised $206 million against that business in April of 2024. We also have identified 48 facilities to conversion to data centers, 21 of which have excess data center capacity. And we intend to sell that off or lease it on a wholesale basis. And we've begun that process talking to 133 counterparties about those either leases or purchases. We've also begun the work to start to monetize some of the excess dark fiber in our network that is beyond what we are operating today. And then overlaying all of these initiatives has been the cost reductions that we've been able to achieve in the operating businesses that we acquired with Sprint. That was a business that had - $300 million of EBITDA. Today we've got those acquired businesses down to below - $40 million.
Still a negative number, still a drag on our operating profits, but significant cost savings. As of the third quarter, we had achieved $165 million of cost savings against a target of $220 million. We had done that in 18 months against a target of 36 months. So we're heading both ahead of schedule and probably ahead of total target in terms of savings. So it has been a very eventful year.
Maybe we can start off with the Wave business since that gets a lot of attention and is a key part of the growth strategy. So you've talked about reaching 25% market share by 2028 or about $500 million. What gave you confidence you can scale to that level? And within that, we find that within the enterprise connectivity space, buying decisions take time. So how did you get comfortable with that timeframe?
So three parts to the answer. First of all, the customer base is not enterprise. It's other service providers, whether they be content generators or access networks. They are not end-user enterprises. They represent less than 10% of the market. Two, we looked at the market two very different ways. From a top-down, it is a market of about $2 billion North American intercity wavelength or transport services. It's comprised of approximately 140,000 waves. We have sold 1,100. So not even 1% of the market to date. But we know that we touch all of the endpoints. We have unique routes. We have the ability to provision more quickly. And we have the ability to price very aggressively. On the bottoms-up side, we looked at the usual buyers. 200 buyers account for roughly 80% of the market.
Of those 200 buyers, 150 of them are already buying transit services from Cogent. So we have a pre-existing relationship and credibility. Two, we have a dedicated sales force of 287 wholesale reps around the world who can help us sell to these wavelength customers. And we went to those salespeople, asked them to look within their account deck and make specific projections of how many waves they would sell, over what period of time, and at what price. We discounted that somewhat for the, I would say, unusual optimism that salespeople have. But we got comfortable that both from a top-down and a bottoms-up approach, we were triangulating on that $500 million run rate by mid-year 2028, which is five years post-closing. And I think with the preparatory work that we have done on the network, we're going to definitely hit that target.
You've talked about reaching, having two-week provisioning across 800 data centers by the end of the year. We're pretty close to the end of the year. Do you remain confident you'll be on track for that two-week target? As you think about your backlog of orders, it's pretty sizable. How quickly can you work through that backlog starting on January 1?
Two different questions. So in terms of our confidence, it is extremely high. We are substantially complete on that 800 target. We ended the third quarter with 657 data centers wave-enabled. We are very close to that 800 number today and will be there or slightly exceed it by year-end. And that includes all of the intermediate point preparatory work that we need to do as well. Second, we have the resources to provision about 500 waves a month. Today we provision approximately 3,000 IP ports a month. An IP port requires one field visit per port, and it's coordinated through our service delivery organization in conjunction with our network operations center and the field team. We can absorb a 33% increase in volume with the current resources. So we can go from 3,000 to 4,000. A wavelength requires two field dispatches, two points. It has two ends.
An A-end and a Z-end. We should be able to do that and add about 500 waves to our footprint every month. We have a funnel and backlog of approximately 3,400 waves. Some of that funnel will fall out, I am sure, because some of it's been sitting there a long time because we could not give guarantees around service delivery or locations. We are doing that now and should be able to eat into that funnel at about 500 a month. But we also anticipate, as we demonstrate the credibility of our assertion that we can add this number of facilities in this timeframe, the funnel will build at a more rapid rate. This is an industry where customers have become accustomed to a three to four-month provisioning window. This is very similar to what we saw 25 years ago when we entered the IP market.
So 25 years ago, if you wanted a high-capacity IP circuit in a data center, and yes, there were data centers 25 years ago, and there were about 100 of them globally that people bought high-capacity. And at that time, a high-capacity circuit was a DS3, a 45-megabit circuit, and it was going to take you three to four months to get it installed. Cogent came along. The smallest circuit we sold was 100 megs, so it was over twice as large as that high-capacity circuit. It was non-oversubscribed and non-blocked. And we provisioned that in an average of nine days. So many investors equate Cogent with low price. That is how we've become the largest IP carrier in the world. But we also have demonstrated our ability to out-provision our competitors. We think those same attributes will carry over into the wavelength market.
When you acquired the Sprint line, you also acquired this very large enterprise business. You've been in the process of grooming off lower margin or negative margin products, and it's created some noise in the financials. Where are we in that process? And any help sort of sizing how much is left as we look into 2025?
Yeah. So there were three things that we wanted to exit. The most obvious were products that were gross margin negative. These were all resale products, typically of value-added services: firewalls, managed router services, DDoS mitigation, security products. We are roughly 80% of the way through exiting those non-core products. We did have to take contracts. It is part of the reason why we were paid $700 million by T-Mobile to take this business, knowing that we had to honor these contracts as they roll off and that non-core revenue tail will extend through 2026. That is the longest duration contract. The second thing that we want to migrate is services that were delivered in countries where we and Sprint were not licensed. So we are licensed and operate in 56 countries around the world. We actually sell services to customers in 180 countries.
But the differential is we require the customer in a non-licensed country to buy the access loop and then connect to Cogent. That is different than what Sprint was doing, where they were purchasing the access through a resale agreement. We were uncomfortable with the regulatory risk associated with that, and we're unwinding those agreements. We're probably about halfway through that unwind. And then the third area were services that were delivered at very low speed over non-fiber facilities. So these could be satellite, wireless, twisted pair, or coax. And in those services, we are looking to migrate all those customers to fiber-based services. We are probably about 60%-70% of the way through that migration. So with these three migrations, two things are happening. Our revenues are shrinking and our margins are expanding.
When we announced the transaction, Sprint was doing $565 million of revenue with - $300 million of EBITDA. Between announcement and closing, T-Mobile affected a number of changes in conjunction with us that reduced revenues to $490 million and reduced the burn rate to -$190 million. As I've stated earlier, we have further reduced that revenue burn rate to about $400 million, but we have also reduced that burn rate to $40 million. The ultimate goal is to get the remaining business to be profitable and generate about $80 million of our aggregate EBITDA from that legacy base without anticipating growth.
Maybe if we can pivot and talk about the legacy Cogent business prior to the Sprint transaction. Your NetCentric business was one that benefited during the pandemic from rising data traffic. So first part of the question would be, how sustainable is the growth in data traffic across the internet? And along those lines, we saw recently a high-profile event with the buffering of the Netflix's Mike Tyson-Jake Paul fight. When something like this happens, are your customers knocking on your door? Does it create an incremental revenue opportunity? Anything you can share about that?
Yeah. So I think three different points, so Chris. The first one is the internet has grown in unit volume at about 23%-25% a year for the past 30 years. I know some investors recently have become concerned because some of the CDNs have seen their unit volume growth slow or even go negative. One public CDN went bankrupt. There's been some consolidation. And I think that's not a perfect window into the internet. The internet continues to grow at that kind of mid-20s rate and probably will continue to grow for the foreseeable future at that rate. The maturity of the internet has been predicted for the past 25 years with a number of technologies eventually making the internet obsolete. And the internet wins because it is the cheapest, most ubiquitous, and easiest-to-use way to communicate, and it is application agnostic. Those factors are going to continue.
Internet traffic will continue to grow. Cogent will continue to gain market share in that growth based on the number of networks we're connected to, over 8,200 customers by their upstream from Cogent, by the number of data centers, by the number of content players. Now, the primary driver of growth recently has been professionally produced video, mostly store-and-forward video, now becoming real-time streaming video, sports. The Tyson-Paul fight is a good example of that. You know, I think there has been some deficiencies in network architecture that were exposed during that event. I think many customers are working to resolve those issues. The fact that our bit volume has continued to grow in the mid-20s has allowed us to gain market share even in light of a 23% per year price reduction. The internet is very deflationary.
We will continue to capture share based on being ubiquitous and lower cost. The pandemic was a huge boom. We went from 18% of video consumption in the OECD countries in 2019 to now being 55% of video consumption and 2024 being streamed. That was a huge inflection, and it drove outsized growth. Our NetCentric revenue growth spiked. It went from growing at 3% up to a peak of 26% and is back down to being a more normalized 9% growing business. So I think, you know, the internet for content distribution remains the network of choice.
And then maybe just quickly touching on the legacy corporate business. Obviously, that was on the opposite end of the pandemic in terms of the headwinds that that faced with more employees working remotely, businesses right-sizing their footprints. The growth has rebounded. It's not quite at the 10% that it used to be historically. What is the pathway to get back there, and how quickly can it happen?
Yeah. So Cogent's end-user business, we're selling directly to corporate and now enterprise customers, had historically grown organically at 11% year over year for 15 years, from 2005 to going when we went public through 2020. Our compounded corporate growth rate was 11% a year. When the pandemic hit, that growth rate declined to a - 9% at the trough of the pandemic. It has recovered to about a 4% positive growth rate. It is being impacted by the number of tenants in the buildings we're in, their willingness to make architectural changes, and their aggregate need for bandwidth. It is slowly improving, but not nearly at the pace we expected. So when we blend this together, we see the classic Cogent selling IP services, which really account for two different products: internet access and VPN services over the internet.
91% of our revenues, and they're growing at kind of mid-single digits, at kind of 5%-6% growth rate, so in understanding a dollar of revenue at Cogent, we try to characterize each dollar four different ways: by network type, on-net or off-net, customer type, corporate, enterprise, or NetCentric versus service providers, by geography, U.S., rest of the world, and then by product, whether it be an IP-based service, and that can be two subcategories: internet access or VPNs. It could be colocation. It could be IPv4 leasing, and now adding on wavelength services.
You mentioned earlier the undermonetized assets that you now have within the portfolio. You've talked about the data centers, the IPv4 addresses. I guess taking a step back, if we're going to look out 12 months from now, is it fair to assume that we should expect some sort of material transaction to come forward on both of these fronts? You did announce the incremental CapEx that you're putting behind the data centers. So as investors, should we see that there would be some sort of payout by this time next year?
The answer is yes to directly answer the question, and there are really three buckets of assets that don't show up on Cogent's balance sheet. I'm going to start with our IPv4 addresses, which investors did not focus on. We were leasing them out, and in April of 2024, we demonstrated that roughly a quarter of our address space was generating enough revenue to validate a $206 million asset-backed securitization, so we effectively monetized that stream of cash flow through an ABS, and we remain flexible on whether we lease out more addresses and monetize them through an ABS or do an outright address sale. We operate approximately 38 million addresses. We are currently leasing out a little less than 13 million or about a third of our inventory, and the value of these addresses have typically traded for between $50 and $60 per address.
Therefore, over $1 billion of latent value on our balance sheet. The second area would be excess fiber. When we acquired the Sprint network, it had three types of fiber: intercity-owned fiber, 19,000 route miles. There are between 24 and 144 strands. In those routes, we need between four and eight to run our business. So we have substantial excess dark fiber, and we've actually sold one or two strands in a few instances, not as a concerted effort, but rather as favors to some companies we had worked with historically. But next year, we should be more aggressive in selling that excess inventory. We have 1,200 route miles of metro-owned fiber. We'll also look to use and sell some of that.
And then finally, we inherited 4,500 miles of uneconomic leased fiber, and we have exited most of that, and the remainder of that will be exited over the next 18 months, helping us improve our cash flow and profitability profile. And then finally, to directly answer the question you posed, is the real estate footprint that we inherited. Prior to the acquisition, Cogent had 54 data centers, 634,000 ft of raised floor, and 69 MW of power. So about 12,000 ft per facility. And these facilities were primarily leaseholds. So 52 of the 54 were leases, two were owned fee simple. Our initial thought was to take the 482 buildings that we got with Sprint that were owned fee simple, that had 1.9 million sq ft and 230 MW of inbound power, and just put a retail Cogent data center in 45 of them.
That thinking changed as it became obvious to us that there was an acute shortage of power in the market, and we could begin to extract value out of the power we had delivered to the sites. Earlier this year, we started to reposition actually 48 of the facilities. We increased it by three, and then identified 21 of the 48 that have in aggregate 1 million sq ft and 100 MW of excess power. We started to clear out the old phone equipment, and in April, we started talking to counterparties to buy or lease these facilities. We got enough market feedback that we were comfortable by June committing $100 million over the next 12 months to convert these facilities.
Our ask on these facilities is either a sale at $10 million a MW or a lease at $1 million a MW triple net per year, representing about $1 billion in value. We got comfortable with the risk that we're taking based on the number of counterparties and the interest level in these facilities. So by the end of this month, we will have three facilities fully converted. By the end of June, we will have all 21, the remaining 18, converted and will have that full footprint of 1 million sq ft and 100 MW available. And I would suspect some of those will either be leased or sold sometime next year.
You've talked about a meaningful ramp in the EBITDA profile of the business over the next few years. I think you've suggested EBITDA would be flattish here in 2025 and then scaled over $500 million by you get to 2028. Maybe just help us break apart the moving pieces there and what are the biggest drivers of that inflection as you look out?
So let's start with what Cogent looked like before Sprint. We were $600 million in revenue and $260 million of EBITDA. That's reported numbers for 2022. We layered on top of that $490 million of revenues declining and negative $190 million of EBITDA. But we received $350 million from T-Mobile over the first 12 months in the form of $29.2 million a month subsidy payments for that acquired business. As a result, our reported EBITDA for 2023 was $352 million. I think there was a lot of skepticism among investors that we could maintain that pace as the T-Mobile subsidy payments stepped down in May of 2024 from $29.2 million a month to $8.3 million or $100 million annually. That's a significant headwind. And even with that headwind, we will report EBITDA approximately equivalent to where we were last year.
Might be a little over, a little bit under, but in that general area. And then as we look at calendar year 2025, we have a further $100 million headwind from that step down in subsidy payments, but we'll make that up with the operating leverage of the wavelength business and the remaining cost synergies. By 2026, we will have lapped that headwind, and we should start to see meaningful acceleration in EBITDA. By 2028, the payments from T-Mobile go away, and the business will be generating about $500 million of EBITDA out of organic business as opposed to subsidy payments. And then from that point, we expect to grow the business at roughly a 5%-7% rate with an annual margin expansion of roughly 100 basis points a year.
Maybe just one final question. It's well flagged that your leverage is going to increase next year as the T-Mobile reimbursements step down. So I guess three parts to this. One, what is the peak leverage we should be mindful of as you go into next year? Two, does this impact your urgency to sell off some of those underutilized assets? And three, are you still comfortable raising the dividend at the current pacing?
We have grown our dividend for 49 sequential quarters. We expect to be able to continue to do so. Our leverage had actually ticked up prior to the acquisition of Sprint due to the impact of the pandemic. So between 2010 and 2020, we had hovered between two and a half and three and a half times net leverage. At the beginning of the pandemic, pre-Sprint, so May of 2023, our leverage had ticked up to 4.2 times net leverage, above the range that we were comfortable with. With the acquisition of Sprint, we actually delevered back down to 3.2 times last quarter. On an LQA basis, that is our worst quarter. We will improve from here. On an LTM basis, leverage will get worse and will peak over five times.
We are comfortable with that based on the growth in the business and the operating leverage that we have. And while we would consider and would like to sell assets, we're also mindful of trying to get the maximum return out of the assets. And it may make sense to lease as opposed to sell some assets and then potentially gain cash through the cash flow generated through those leasing activities.
I think we'll leave it there. Thank you very much, Dave.
Thank you very much, Chris. Thank you all very much.