Clients only. Welcome back to Citi's 2025 Global TMT Conference. For those of you I haven't met, I'm Mike Rollins. I cover communication services and infrastructure. Disclosures are available at the back of the room, and if you don't have access or would like another copy, please email me at michael.rollins@citi.com. We're pleased to welcome back Dave Schaeffer, Chief Executive Officer of Cogent. Dave, thank you for joining us today.
Hey, Mike. Thanks for hosting me. Wanna thank investors for taking time out of their day to listen, and always thanks Citi for a great venue.
You know, thanks so much. You know, maybe we'll start with a high-level question for you, which is just an update on the strategy to grow and create value for shareholders.
So within Cogent, we had an organic business that grew at 10.2% a year for 15 years, with no acquisitions between 2005 June, when we went public, and the beginning of the pandemic. And the growth rate in that business slowed because of COVID to about 5% per year, and the rate of margin expansion also decreased from an average of 220 basis points a year, down to about 100 basis points a year.
As the pandemic waned, that business began to improve, but opportunistically, we had a chance to acquire the original Sprint Global Markets business from T-Mobile. And in that acquisition, there were things that both positively and negatively impacted our growth trajectory. On the negative side, it was a business that, for the previous three years, had declined at 10.6% per year, and it was a business that was burning $1 million a day in cash.
To help mitigate that burn, T-Mobile agreed to pay us $700 million over a 54-month period, and we still have payments that we expect to receive between now and the end of Q1 2028, with a total net present value of about $244 million. But when we acquired that operating business, we actually accelerated the rate of revenue decline intentionally to purge that business of unprofitable products and certain unprofitable customers. As a result, over the 8 quarters since that acquisition has been completed, our growth rate on a combined basis has declined at about 2% annually.
So we went from being a 5% grower to a -2% grower, and we have been able to improve underlying EBITDA margins and actual EBITDA in light of that top-line decline at an average sequential rate over those eight quarters of about $5.2 million a quarter. We anticipate the growth in that business that we acquired has plateaued, and the combined company will now start having positive top-line growth sometime in this quarter.
We also know that there are additional synergies that we can take out of the combined business, and then finally, the monies that we are spending on integration efforts will taper off over the next six quarters. And then finally, the primary reason for doing that acquisition was the opportunity to repurpose the original long-distance telephone network into a wavelength transport network.
We spent nearly two years completing that integration and optimization of those assets. We have begun selling those services. What is a brand-new business inside of Cogent is actually growing very rapidly. In fact, last quarter, the sequential growth rate in revenues in that wavelength business was 40%, or excuse me, 27% sequentially, and 149% year-over-year.
We anticipate that business to keep growing off of a larger base, and as a result of these three different drivers of growth, one being, you know, roughly 70% of our current revenues growing at about 5%, the second being about a third of our revenues declining at, you know, kind of 1%-2%, and this very small but rapidly growing business growing, the aggregate growth rate of Cogent should return to between 6% and 8%, below our pre-pandemic levels, but above the level that we were achieving, prior to the acquisition.
And that's really driven by the inclusion of very high-margin wavelength revenue, and with this, we anticipate growing margins again at about 200 basis points year-over-year. So the combination of moderating capital intensity and accelerating EBITDA growth should allow Cogent to materially grow its free cash flow.
Really helpful, and it gives us a lot to drill into. You mentioned a few things that's gonna affect financial performance going forward, you know, the returning to revenue growth, the synergies, and then getting through the integration expense. Maybe working backwards on that, on the integration, how much is in the, you know, quarterly EBITDA that's a drag on that from integration that goes away?
So there's actually two components. There are costs that we have identified that are directly related to that business that we can continue to take out. There is approximately $20 million of those costs that will come out over the next year or so. In addition to that, we are spending about $4 million a month, or about $12 million a quarter, on direct integration work.
When we initially acquired the business, the primary focus was to consolidate all of the customers onto common IS platforms, whether it be billing or customer care or accounting or network monitoring. Everything was collapsed into a common platform within the first roughly six months of the acquisition.
The second area of grooming took much longer and was not completed till earlier this year, and that is the complete elimination of the Sprint network and the migration of 100% of the customers and services onto Cogent's infrastructure. And we've now been focused on the third leg of the integration, which many companies choose to ignore, and that is the standardization of all of the customer information, systems, network elements into a common nomenclature, a common, set of reports that allow us to more effectively manage those customers. That work is ongoing. It encompasses over several hundred discrete projects, and those projects will taper off over the next year and a half, and we anticipate that cost disappearing.
By the end of the eighteen months, that four million goes to zero?
That's correct.
You don't need anything in its place to get the results that-
That is correct.
The $20 million of savings you mentioned, does that go in the synergy bucket that's remaining?
That is going in the synergy bucket. We had originally projected $220 million of synergy. We achieved that goal earlier this year, and then revised that goal upward by about 10% or another $20 million.
That's income over the next year?
That is correct.
And then, you know, going back to the revenue, and there's a lot to get into on the revenue side. You know, you mentioned that the underlying business growing mid-single digits. You know, when we've looked through some of the detailed disclosures, and we look at on-net corporate, and we look at on-net NetCentric, we're seeing just based on percentages, and I know there's some rounding errors, so we have to be careful not to be too precise with this, but it looks like the revenue from those core pieces could be flat to down in the first half of the year. Is that what you're seeing, and is there some, you know, perspective to-
That is not what we are saying. We are seeing that number up, but up modestly.
Okay.
So I think there are three reasons why you may come to that conclusion. The first is, approximately 20% of the revenues that we acquired from Sprint were classified as either corporate or NetCentric, not as enterprise, and those revenues have continued to decline. The second reason is, some of those customers were purchasing non-core products, and if the non-core product was associated with an on-net service, it got classified in on-net.
If it was associated with off-net, it was classified in off-net. We have, by design, tried to eliminate all of those non-core products. We have taken the run rate in those non-core products down from just under $60 million annually at acquisition closing, to today in the order of about $15 million. There is still a tail of those services that we have to support due to contractual obligations.
The third point is that there were actually customers moving from off-net to on-net, which actually helped on-net to some extent beyond just organic sales, and we have groomed a significant amount of off-net traffic. Virtually all of the customers that we acquired with Sprint, whether they be enterprise or corporate, were off-net. Because of the locations of some of those corporate customers, we chose to terminate services to them because there was not a fiber solution to deliver those services.
For the enterprise customers, we have concluded that the percentage of their services that will be off-net will continue to remain much higher than our installed base. Prior to the acquisition, in our installed base of end users, corporate customers, we were doing roughly 60% of revenues on-net and 40% off-net.
In that acquired enterprise business, even with the grooming that we've been able to achieve, we're still running 88% off-net and only 12% on-net. That has two consequences: one, higher churn due to lower quality, and two, much lower margins associated with those off-net services. But the kind of core on-net services, whether it's a NetCentric or corporate customers, are growing in the mid-single digits. Probably lower single digits for corporate, you know, kind of in the 3%-4% range. It's bounced around over the last couple of quarters, and then on the NetCentric, there has been some deceleration down to about 8% revenue growth.
For the 8% growth, does that include the IPv4?
It does include the IPv4, which is roughly 85% a NetCentric product, it is 14% corporate, and 1% enterprise. And again, like any of these ancillary services, if that is associated with a on-net customer, it gets counted in on-net. If that service is associated with an off-net customer, it gets counted in off-net. The growth rate in that portion of the revenue stream has been much stronger. It grew nearly 40% on a year-over-year basis.
So when we zoom out a bit, and maybe we'll stick with NetCentric, you know, for a moment, you've got traffic growth that's slowed down to, like, low double, upper single, right?
Yeah, 8%-9%.
We're accustomed to thinking about price declines on average being about 20% per year. So is it a situation where that on-net, you know, IP transit piece is shrinking, but being made up with, like, off-net and IPv4, and that's kind of the recipe, you know, for now, and until, you know, maybe traffic growth re-accelerates in the future?
There is one error in your calculation, and you're assuming that all NetCentric customers are homogeneous. They are not. If we are selling to smaller customers, the effective price per megabit may actually go up, while the average price is going down, and the reason for that is smaller customers pay higher cost per bit than larger customers.
The second distortion in that analysis is geographic. We have seen, over the past couple of years, a much faster rate of traffic growth in the less developed world than the developed world. As a result, the percentage of traffic that is coming from rest of the world has gone from around 45% three years ago to 55%, and that increase in international traffic ends up increasing the effective price per megabit, because our pricing tends to be much higher in less developed markets.
While we are not actively selling today in India, but intend to be in that market by the end of this month, after an eight-year saga to get a license, the pricing in that market is roughly about fifteen x the North American pricing per megabit, with a very limited number of competitors. While we will not go into the market at current market prices, we'll look to disrupt. We will not be at pricing nearly as low as North America or Western Europe, and probably not even as low as some of Latin America and Africa, where there is more competition. As we get growth from a market like that, the effective price per megabit tends to be much higher.
One thing I also didn't mention in that kind of calculation is waves. Is waves in that number?
Waves are, again, classified by the type of customer. To date, 100% of waves have been on-net. We have not sold an off-net wave. That doesn't mean we won't. There could be a situation where a customer needs a wave to a single-tenant building that we choose not to build into, and we will combine a local wave from a third party with our long-haul wave.
We also look at waves by customer type. The majority of waves have today come from NetCentric customers, with over 90% of waves being NetCentric. We have had a handful of waves to either corporates or enterprise customers, but combined they are less than 10% of the wave base.
But again, to remind investors, every dollar of revenue at Cogent gets four different disclosures, and they're not meant to be mutually exclusive. They're designed to give you four different views of the business: customer type, corporate, NetCentric, and enterprise, network type, on-net, off-net, third, by product type, and then finally, by geography.
And maybe just, you know, connecting a little bit more on the waves business, you know, over the summer at some other conferences, you laid out some of the backlog that you're seeing. Do you have an update on how the backlog is progressing and how the sales of that backlog is progressing?
I have no specific update, other than we feel comfortable with our ability to exit the year at a $20 million-$25 million run rate. Two, that we have a significant backlog of waves that have been installed, but not yet customer accepted. Third, we continue to build the funnel and the backlog, and what we are hoping to do is compress the time from our installation to customer acceptance and our ability to recognize revenue.
We feel very encouraged about the wave market in three distinct ways. One, there is incremental use cases that we were not expecting. Two, the level of loyalty to existing providers is lower than we expected, due to many customers perceiving quality of service issues. And then, three, the uniqueness of our routes have been viewed as a huge positive for path diversity among customers. A wavelength product is an unprotected product by design, and by having common city pairs, common data centers, but having a completely physically diverse route, there is a value to that, that the other providers cannot deliver.
Sorry to ask this, but just given how fast waves are growing, when you say a $20 million-25 million run rate in 4Q, what does that translate into in terms of like what we would see on the income statement for 4Q revenue?
That means we will be on an exit run rate at the end of the quarter for $20 million-25 million.
A year.
A year.
Yeah.
But that is measured on a monthly basis, so the variable in that will be how quickly the customers accept, and we will have sufficient number of waves installed with adequate ARPU to hit that number. What we cannot yet predict accurately is the pace from install to acceptance, and we also know that with our IP services, after a several-year period of allowing customers to consistently push out orders, we eventually implemented a forced billing regime, and it is our intention to do the same with wavelengths, but I think we need to have more market share before we can take that more heavy-handed approach with customers.
And so then maybe, coming back up to the corporate side of the business, can you give us an update on how you're doing with individual relationships, that penetration in the buildings, and then what's happening with the sell-through of services? Because, you know, you provide the VPNs, you have, of course, the higher connection speeds that you can deliver. So maybe a little bit about that volume number, and then that, like, mix in ARPU relative to that.
So three different questions. First of all, the number of tenants in our footprint has decreased because of COVID and the increase in occupancy rate. Partially offsetting that has been the fact that new lease signatures in our one billion sq ft footprint have typically been for smaller suites. So therefore, there is a potential for more aggregate tenants if the building returns to pre-pandemic occupancy levels. While we have seen a reduction in the rate of vacancy increases, there have only been a few markets where we've seen net absorption be positive and vacancy rates coming down.
Probably the biggest exception to that is DC with DOGE, where we are now at all-time high vacancy rates, and they are continuing to increase. Even markets that had been challenging after the pandemic, like San Francisco and Seattle, have peaked in vacancy and are now seeing some positive net absorption.
The second point is, there is a consistent transition from lower speeds to higher speeds. We went through a significant rotation, probably right before the pandemic, and then continuing for the first year or so, where we were replacing fifty or a hundred megabit connections with 1 Gb connections.
And today, probably 93% or 94% of the entire installed base is on one gig or greater speeds, and we have seen an increase in demand for 10 Gb connections, which had historically only been a NetCentric product. But as we have wave-enabled the CNDC footprint, we also, in that process of reconfiguration, our metro networks completed the process of creating a 100-gig transport out of each of the buildings, allowing us to sell 10-gig connections.
So we are actually seeing corporate ARPUs go up, even though the price per megabit is coming down because of that speed shift that is going on. And then the third point that you raised, which is by product type, and our business is predominantly a DIA business or internet connectivity business, but we do sell a significant number of VPNs to end users.
Today, the percentage of VPNs as a percentage of total revenue is actually at an all-time high, but it's distorted by the fact that most of the Sprint customers were receiving VPNs, and only a very small percentage of them were DIA customers. Now, the technology that Cogent had been deploying was virtual private LAN service, or VPLS. We continued to deploy that. Sprint had been delivering a much more antiquated service on MPLS, or multi-protocol label switching. Initially, our thought was we would be able to convert all of the Sprint customers to VPLS, and they would welcome that because it was cheaper, more flexible, and it was more scalable.
What we quickly realized in conversations with those customers is that they had no interest in migrating to a different VPN architecture, and that actually sent us back to the drawing board, and we actually revived some older equipment that had been pulled out of the Cogent network and deployed 273 global MPLS aggregation points, and we provide that MPLS service over our public internet backbone, but with MPLS aggregation devices at the edge to allow customers to connect.
That allows us to continue to provide MPLS for customers without the threat of turning off the service, and in fact, we've given customers a ten-year written guarantee that we will support that technology. That's actually been very welcome, and for that reason, our aggregate VPN business has grown. Even though for the legacy corporate customers who were buying VPLS, there has been some replacement of just public internet. In aggregate, our VPN business has never been higher as a percentage of our revenues.
That's really helpful. You know, maybe migrating to capital allocation in the few minutes that we have left, how are you thinking about a possible optimization or shift in going from, you know, paying out the dividends that you're paying to maybe migrating to buybacks or focusing on de-leveraging, you know, with the cash flow that you generate?
So Cogent has returned approximately $2 billion to shareholders. We have bought back 10.9 million shares out of a 49 million share float still remaining, and we have a dividend that we have grown for 52 quarters consecutively, sequentially. We historically hovered in a leverage range of about three times levered, while we were doing both buybacks and dividends. With the onset of the pandemic, our leverage went up to 4.2x .
When we initially acquired Sprint, we immediately de-levered down to 2.7x due to the front-end loading of the subsidy payments from T-Mobile, meaning we got more upfront than we needed to take out and cover the burn, and that helped fund the wave enablement of the network, as well as accelerate the purging of undesirable revenue. Now, going forward, we have peaked in leverage in the most recent quarter.
That was clearly telegraphed in advance. If you look at on an LQA basis, we are already declining in leverage. Secondly, with the growth in EBITDA and the additional growth in EBITDA that will come from selling wavelengths, we generate incrementally more EBITDA, and our capital intensity is declining as our reconversion and power optimization in the data center footprint is now complete, so that leaves us with increasing amounts of cash that we could deploy in one of the three different venues.
The market has clearly decoupled our dividend from our stock price, so we have to listen to what the market is saying to us, and that they just don't believe that. We believe that we can both delever and continue to grow the dividend concurrently. The market does not believe that. It is possible that we pivot more to buybacks.
It is possible that we decide to leave more net cash on the balance sheet, and therefore net delever. What is true is we are certain that we will return increasing amounts of cash to shareholders, and one, if not all three of those different factors, and at minimum, we will be able to delever from the 6.6x leverage we're at today, down to about five times by the end of 2026, and then we need to go beyond that.
So in our final minute, one other way you could delever is sell some of the non-core assets. You talked about data centers in the past, you talked about IPv4, you talked about maybe there could be some fiber. What's the opportunity? How do you see that in terms of if... you know, very quickly, in terms of magnitude, timing, likelihood?
We are highly motivated to divest of those non-core assets. We have concluded that we cannot generate revenue fast enough to justify us holding those assets. In the data centers, we had a lot of preparatory work based on feedback from potential customers. That work was done in June of this year. We have six LOIs as of the end of the quarter, last quarter in hand.
We are continuing to negotiate. We also have looked at the IP market, IPv4 market there. We have been less willing to try to sell at this point, as the two largest buyers have not been in the market. While the market is still robust and is broad, I don't think it's deep enough to absorb our inventory. And then on dark fiber, we've done a handful of one-off deals, and we will continue to evaluate those on a case-by-case basis.
We think that non-core asset divestitures are helpful. They will clearly help us delever in the short term, but they do not build long-term enterprise value in the way that we can by growing our recurring revenue business. So our primary focus is on growing wavelengths and growing our core on- net services.
Dave, thank you so much for your time today.
Hey, thanks, Mike, for having me, and thank you all for listening. Take care.