Cogent Communications Holdings, Inc. (CCOI) Earnings Call Transcript & Summary

Welcome to Citi's 2024 Global TMT Conference. For those of you I haven't met, I'm Mike Rollins with Citi Research, and we're pleased to welcome Cogent Communications Founder, Chairman and CEO, Dave Schaeffer. This session is for Citi clients only, and disclosures are available at the back of the room right next to the AV desk. And for everyone in the room, you can see there's QR codes up here on the screen and in some of the rows. You can scan that, and we're going to welcome you to participate in our upcoming live audience surveys. And importantly, your participation, your submissions, they're anonymous. And so with that, Dave, thank you so much for joining us today.

Okay. Thanks for inviting me, Mike. Thanks for Citi for a great venue. Thanks for investors taking your time. And again, I apologize for being a couple of minutes late, as I was upstairs in the other tower, but I'm here now. Little challenging to get an elevator down. I asked, can you walk? And of course, you can only walk to the fifth floor, and then you still have to wait for an elevator. So anyhow, enough. Let's get to questions about Cogent.

Great. So maybe get us started. It's been a couple of years of change for Cogent, both in terms of the acquisition that you completed, and you've been integrating, and also the operating environment. So just curious, if you can provide an update on the strategy for Cogent to grow and create value for shareholders.

So Cogent has undergone more change in the past 18 months than it probably had for the past 15 years. Just to remind investors, as an independent public company between 2005 and 2020, we organically grew at a rate of about 10.2% a year with no acquisitions. And during that period, our margins typically expanded about 220 basis points per year compounded over that period, resulting in about a 15% growth rate in EBITDA. And in that classic Cogent business, there were 2 major segments of customers. Corporate customers that reside in multi-tenant office buildings that bought Internet access and VPN services from us and then would have branch offices in other locations, that represented roughly 60% of our revenues and actually grew exceedingly consistently at about 11% a year. That business was severely impacted by the pandemic, and that growth rate decelerated to actually a negative 9% at the trough of the pandemic, and then it has slowly returned to being about a 4% growth rate. The second part of Cogent's classic business was the sale of Internet access in data centers around the world. And prior to the acquisition of Sprint, we were in 48 countries and about 1,500 data centers at the time. And that business had grown at an average of 9%. It was about 40% of our revenues, but tended to be a lot more episodic in its growth, and there was a lot of volatility. We actually went into the pandemic with that business growing at 3%, and we saw the growth rate in that 40% of our revenues accelerate to 26% at peak. It has since decelerated and growing at about 10% on a year-over-year basis, actually slightly lower than that in the most recent quarter at about 4.5%. We acquired the Sprint Global Markets business in May of '23. That business had 2 components to it. We acquired an operating business, in which roughly 80% of the revenues were coming from large enterprise customers. And 93% of those revenues were off-net, meaning they did not touch the Sprint network. That business was burning nearly $1 million a day when we signed the acquisition through some efforts of T-Mobile. Prior to closing, that burn rate was reduced to $190 million of negative EBITDA and $30 million of CapEx at closing in May of '23. We have subsequently shrunk that business, reduced costs. We targeted taking out $220 million in costs over a 3-year period. In fact, we've taken out 62% of that target in the first 15 months. And that business will eventually stabilize as a 20% margin business, with a significant part of that customer base being off-net and no real ability to have meaningful growth in that segment. The growth opportunity in the Sprint acquisition was the second part of the transaction. So for the first transaction, we were paid $700 million in cash by T-Mobile over a 54-month period to acquire that money losing business. The second part of the transaction was the acquisition of the Sprint network that was originally built to carry long distance voice. That network was constructed between 1984 and 1990 at a capital cost of $20.5 billion. We paid $1 for that asset. It was effectively dormant. That network had not carried voice traffic for a decade. In fact, the 482 facilities that we acquired had old telephone equipment in them, comprising 22,500 cabinets of obsolete equipment. We are still removing that gear. We're down to about 6,000 remaining cabinets to be removed. The physical network was 19,000 route miles of intercity fiber and 1,200 miles of metropolitan fiber, with effectively no revenue on that network. We then have a number of initiatives underway to turn that network into a viable business and convert it and repurpose it to sell optical transport or wavelengths. The first thing we had to do was interconnect the Sprint network to our metropolitan networks in 110 markets. That effort was completed in February of '24. The second thing we had to do was remove all of that dead telephone equipment. That is still an ongoing effort. The third thing we had to do is reconfigure our metropolitan networks to be able to sell wavelengths in 800 targeted data centers across North America. We're about 75% of the way through that reconfiguration project. We have to deploy transponder shelves in all of those 800 locations to accept those wavelengths. We're about 70% of the way through that exercise, and we have to deploy reconfigurable add-drop multiplexers at the intersection of the metro and long-haul networks, where we're about 70% of the way through that effort. By year-end, we will have completed that, and we will be in a position to sell wavelengths across 800 facilities with a 2-week provisioning window. So we are expanding Cogent's core business in 2 dimensions. one, we're picking up a new product, wavelength services; and two, servicing a new type of customer, large enterprises. So you are right, Mike, it has been a busy 18 months.

It has, and that gives us a lot to drill down on. The first maybe survey question we'll get to, let me pull this up on the screen, and you can also, as a reminder, hit the QR codes with your camera to be able to participate in the survey, completely anonymous, and we're not tracking. So let's turn this on. The first question. So does Cogent have the right products and strategy to return to significantly positive annual organic revenue growth within the next 2 years? And just a simple up or down yes or no. While those results kind of come in, we'll keep my eye on the system here. Maybe briefly, remind us, now that you have additional products within the portfolio, what your addressable market for revenue is and what your share position currently is.

Yes. So let's start with our IP Transit business, where we are the largest carrier of Internet traffic in the world. That is a $1.5 billion addressable market. We have 25% market share in that business. In our corporate footprint, in North America, it is about a $9 billion total addressable market. Roughly, 11% of that market is on-net for Cogent, and we have about a 20% market share in the on-net portion and a sub 2% market share in the remaining off-net portion. The wavelength market is a $7 billion global market, $3.5 billion of that is in North America, $2 billion of that is intercity, $1.5 billion intracity or metro. We are primarily focused on the intercity portion of that. We today have a run rate in that business of only about $15 million out of the $2 billion market, so less than 1/10 of 1%. We are targeting a 25% market share within 5 years of deal closing. We also have an IP leasing business. Our 2 primary competitors are Amazon and Microsoft. There's not a well-defined market share or even total addressable market, but we're probably the second largest leaser of addresses and generate about 4% of our revenues. About 2% of our revenues come from our data center footprint. We're a de minimis sub 1% player in that market. But we do have nearly 2 million square feet and over 200 megawatts of power, which would actually rank us as probably the third or fourth largest player in the market. So a lot of different market segments. I'll just close on the enterprise segment. There, we're doing about $160 million in revenue, serving large enterprise customers. That is a very large market, but shrinking very rapidly.

Really helpful. And let's go to our first survey results here. So according to the survey, 89%, yes, you have the right products and strategy to return a significantly positive annual organic revenue growth within the next 2 years. 11%, no. So as you look at your conviction to get back to revenue growth, and I'm also seeing that the Visible Alpha consensus also has you growing revenue in 2025, does this seem like a fair expectation?

It definitely is a fair expectation. I obviously agree with the majority of the respondents. We do have revenues that we are intentionally attritting, and we've been very clear with investors that there are noncore products that we are end of lifing. There are access customers that we are turning off service due to the remote locations and the inability to serve them via fiber. And then there are some international customers that we are bifurcating their access into a port-only versus local access loop in countries where we are not sufficiently licensed. But in total, I think getting back to positive revenue growth is extremely likely in less than 2 years.

Great. I'm going to introduce the next survey question, and then we're going to talk a little bit more maybe about the revenue and the EBITDA environment. But we'll throw this one out there. Will Cogent monetize a significant portion of the noncore assets previously identified as the unused IPv4 addresses, the data center capacity and the dark fiber over the next 12 months? So we'll get our responses, and we'll come back to this in a moment. Just on the revenue side, you mentioned, I think, on the earnings call, or one of the investor conferences recently, that the core corporate performance is back to growth.

About 4%.

About 4%.

On-net corporate.

And so as you look at what's happening with the sales force, the intake, the churn, do you feel like this is now back to a sustainable growth position for that segment?

So the answer is sustainable, yes, suboptimal though. Remember, this is a business that for 15 years had grown organically at 11%. 4% is clearly better than negative 9%, but it's also clearly much worse than positive 11%. And I think we continue to see the impact of the pandemic recede. It's not completely equal among markets, but that business will continue to gradually improve.

And on the NetCentric side, we saw some deceleration there and some deceleration in traffic. What's happening there? And is this something that you view as temporal, the decel? Or is there something more significant that we should appreciate from a trending perspective?

So if we look at Cogent's traffic growth year-over-year over the past 20 years, we've reported that number consistently, it's been in the low 20s, 22% or so traffic growth. It did decelerate to 17% year-over-year in the most recent quarter. There is just some volatility in that. But in general, Internet traffic growth is driven by the number of users, number of minutes a day of use and the bit intensity of those minutes of use. All 3 of those are still trending positive. I think the Internet will cease to grow at some point, but I think we're still far from that. The drivers of traffic growth have been episodic. Most recently, the key driver has been video. We have hit the point where now about 50% of video consumption in the developed world is stream, but that also means 50% of video consumption still can migrate to the Internet. And I think probably one of the key inflections over the past year has been the movement of live sports to streaming. So a year ago, all of the streaming content was more static or prepackaged in nature. And now we're seeing a migration of real-time sporting events to streaming. I think they'll continue to drive growth. And therefore, I think traffic growth will continue for Cogent in the low 20s.

Another question on different products on the wave side of the equation. You've talked about the importance of getting the 800 online, getting the install windows to the 2 weeks, and that would be the catalyst to accelerate the bookings relative to the pipeline you have. For that pipeline, is it a little bit of a cycle where you have stuff coming in, customers say it's going to take maybe longer than I'd like, and they're going elsewhere for their waves, and then you're replenishing that with new interest? Or is there a stickiness to this pipeline that you're seeing for the interest to use Cogent for wavelength service?

I think it's actually both. We will definitely lose some customers that are in the pipeline. We also will acquire new customers. The wavelength market is more expensive on a per bit mile basis than using the Internet. If your goal is to move bits at the absolute lowest cost, you use the public Internet. However, wavelengths have 3 attributes that the Internet cannot match. It's secure, it's latency defined and it allows for very large packet transfers. All 3 of those things don't work on the public Internet. And there are businesses that value those attributes and generally pay about 2.5x as much per bit mile to use wavelengths. Now for Cogent, we gained market share in the transit business by doing 2 things: dramatically undercutting our competitors by at least 50% on price; and by provisioning more quickly. In the 25 years Cogent had sold transit, we saw our price differential shrink from 97% discount to a 50% average discount, and we transformed the provisioning window from 90 days to 9 days. When we look at the wavelength market, we think we'll get to similar market share much more quickly. And we're going to do that because of the uniqueness of the routes, the ubiquity of the endpoints, lower pricing, but also faster provisioning. So something we are doing different than the companies we're competing with in that market is we have built an architecture that is solely optimized for wavelengths, just as we have a parallel architecture optimized for IP Transit. Our competitors typically run heterogeneous products over a network that is much less scalable and has longer provisioning windows. Today, the average time to provision a wavelength from our competitors is between 90 and 120 days. I think the fact that we can cut that time down, coupled with the other attributes and lower price, should allow us credibility. We also have a developed sales force that calls on this market. We have 280 quota-bearing reps focused on the wholesale market. They have relationships with all 200 significant buyers of wavelengths in North America. And in fact, 3/4 of those customers are already buying transit services from Cogent. That credibility should transfer over into wavelengths and allow us to get to that 25% market share much more quickly.

And then just thinking about the use cases, how can the emergence of GenAI capabilities help revenue as well as your cost? And which is the greater opportunity for value creation at Cogent?

So I think AI will be built into every application in every business process, independent of what industry you're in, and will reduce your cost. But I think that will be a gradual program over a very long period and not meaningfully change Cogent's cost structure over the next 3 to 5 years. On the revenue side, I see 2 discrete opportunities. One, GenAI is practical today because of the massive amounts of data that have been collected on the public Internet at little or no cost. That data now has a purpose. It now is used to train large language models. And I think because of that, the value of that data has gone up, and more data is being collected. That is a positive for Cogent's transit businesses, historic business, where we have 25% market share. The second thing that's happening is the actual training itself is today constrained by power availability. So in the ideal world, you would do the training where the data sits. You would have a shorter path, and you would have as much compute power as possible, where these large databases reside. That is not realistic. Because of power constraints, that training is being distributed where power is available. As a result, wavelengths are a primary beneficiary of that distribution as the way in which to move data from the repository to the trainer and back again. So I see that as an incremental use case, so the use cases that existed previously for wavelengths, and probably responsible for some aggregate growth in that $2 billion TAM.

Let's see what the survey results are for the second question we asked, which was will Cogent monetize a significant portion of the noncore assets over the next 12 months? 40%, yes. 60%, no. So Dave, from your perspective, if you can review for us the opportunities for monetization and how you see these opportunities playing out over the next 12 months.

Yes. So there are 3 asset bases that don't show up on our balance sheet and are not built into our forward guidance for revenue and EBITDA. And again, to remind investors, these are multiyear targets. They were laid out 5 years from acquisition. They said we would be at $1.5 billion of revenue scale. We're at about $1.1 billion today. We would have $500 million of EBITDA. We're at about $350 million of EBITDA today. And about $500 million of that $1.5 million would come from wavelength sales versus the roughly $20 million today of wavelength revenue on an annualized basis. The 3 assets that we have that could have value are IP address spaces. We have 37.9 million addresses. About 13 million of those are leased at this point. About 2 million are in use for free, and we have about 22 million. We are monetizing those that are leased. We did an asset-backed securitization earlier this year, where we raised $206 million against a portion of that lease income. I think the question is, will we sell off that excess inventory? And the answer is maybe. We are raising prices on the leased addresses. We are also exploring the sale of the addresses. There are exchanges that are quoted on a daily basis. We also know that addresses have generally appreciated over time and are continuing to still appreciate. And finally, the 2 largest buyers of addresses are temporarily not buying, that being Microsoft and Amazon. I'm going to skip over to dark fiber, and then I'll come back to data centers. We have 19,000 route miles of intercity fiber, 1,200 route miles of metropolitan fiber. The fiber counts range from 24 strands to 144 strands to spending on the segment. We only need 6 of those strands to run our business. We will sell off that excess fiber. It will be on a route-by-route basis. We don't know who the buyers are and what we can get for it, but that is something we're willing to do once we have completed the wave enablement, so we could focus on that revenue opportunity. The same field resources are necessary to provision the dark fiber, but we do anticipate some dark fiber sales. And then data center capacity, that's maybe been the most surprising to us is how strong the demand for data centers with existing power. We had initially not expected to monetize that well at all. We were going to have a small percentage of our data center space for retail colocation and leave the rest follow. As demand picked up because of GenAI demand, we started to rapidly convert those centers and market the excess capacity. We have 1 million square feet of data center space and 100 megawatts of excess capacity. We began talking to counterparties in April. We have a number of ongoing discussions, some are from multiple centers, some are for single centers. And we're offering that excess capacity in 2 models, either on a lease for $1 million per megawatt per year on a triple net basis or an outright sale of a facility. You have to take the whole facility at $10 million a megawatt. And we have counterparties interested in both. If I had to guess from where we sit today, that is going to be the most meaningful of those 3 assets to generate incremental revenue for Cogent.

And one of the features of the data center, so there's a number of them, and I think based on the math that -- the numbers that you gave us, the math is something like 2 megawatts a data center on average.

Yes, there's a great deal of variability. The largest one is about 13 megawatts, and then we have some small as sub-1-megawatt.

And is there a difference in value per megawatt between the larger facilities and the smaller facilities?

It's not clear. Today, we look at public comps and trades in the market, and there's a fair amount of dispersion there, but we have priced our sales and our leasing at about a 40% discount to where the market prices are. And we've seen both lease and purchase interest from both large and small facilities. And I've actually been a little bit surprised at the geographic locations that have generated the most demand. They were not the ones I would have thought of it first.

We've got a few minutes left, so I want to hit our last survey question and then hit a question on your EBITDA opportunities and as well as on capital allocation. So the question is, do you view Cogent's current dividend payout and recent policy to increase the dividend quarterly by $0.01 as sustainable? And the choices are, for our audience who's listening in: yes, the payout and the recent policy to grow the quarterly dividend are sustainable; yes, the payout is sustainable, but Cogent is unlikely to grow the dividend per share in the future; no, expected dividend payout will eventually need to be cut. So we'll see what our audience thinks of that. Before we get there, can we talk a little bit about the EBITDA opportunities? So if I'm doing the math right, the EBITDA from the core business without the payments from T-Mobile was about $40 million.

$39 million.

$39 million in the second quarter.

That's correct.

So $160 million annualized. And I think there's -- you mentioned the percentage of synergies done. So there's about $85 million left for synergy.

That was on point, yes.

And that takes you to $245 million versus the $230 million pre-deal.

That's correct.

So can you walk us through -- you mentioned the $500 million EBITDA target. How do you think about scaling the EBITDA through these cost synergies and then towards that $500 million level?

Right. So on the cost synergies, 2 things: one, we are ahead of schedule; two, we believe we will have the opportunity to increase that number as we've uncovered some additional synergy areas. Two, the incremental margins from IPV for leasing and wavelength leasings are substantially higher than the margins in our core business. Three, the operating leverage of the core classic business had slowed due to the slower top line growth, but we're still continuing at about 100 basis points a year versus the 220 that we had averaged over the previous 15 years. So when we kind of package that together with this -- the classic business slowly improving and margins expanding, the bleed down of the negative margins from the acquired Sprint business and then, finally, the layering on of the higher-margin businesses, I think our 5-year targets are very realistic, probably somewhat modest.

Great. And that gets us to the capital allocation question. So if we look at the results of the survey: 25%, the payout and recent policy to grow the quarterly dividend are sustainable: 50%, the payout is sustainable, but Cogent is unlikely to grow the quarterly dividend per share in the future; 25%, expect the dividend payout will eventually need to be cut. Can you walk us through the capital allocation, your outlook for the dividend and what investors should expect?

So Cogent is very lucky that it built its business without debt. In the early acquisitions we did in the 2001 to 2004 time frame, we were actually producing free cash from operations by the end of 2006, which was about 18 months after going public. We began buying back our shares. And between the end of '06 and mid-2010, that was our only mechanism of capital return. We had no debt. We elected to add debt to our balance sheet and accelerate our return of capital. We have returned more than 100% of free cash flow from operations for 14 consecutive years, and we are 3.17x levered. This is the combination of the pace of dividend growth, the supplementing of that with buybacks, we did some buybacks last quarter, and the expansion in EBITDA. I believe we have the ability to return increasing amounts of capital and should do that through growth in the regular dividend and opportunistic buybacks.

Great. And then in terms of -- as you look at over the next couple of quarters, are there any step function opportunities that we should be expecting in terms of cost or cash flow performance, especially as some of those payments from T-Mobile start to wind down?

So our payment subsidy from T-Mobile was front-loaded, $350 million in the first 12 months, those stepped down in June, and then it goes to about $8 million a month, $8.3 million a month going forward. Our EBITDA in 2023 was $352 million. It will be similar in '24 and in '25, and that is with this step-down in subsidy payments. Partial year, half year in '24 and full year in '25. So the improvement in the underlying EBITDA offsets that step down. We'll start to see meaningful expansion in EBITDA in '26 as the payment stream from T-Mobile continues through midyear '28, at which point, it has ended. We do have a working capital true-up that occurs at the end. But at that point, we're organically producing over $500 million of EBITDA and expect EBITDA to continue to grow from there. And with that growth in EBITDA, we will have the financial flexibility to continue to pay out, hopefully, more than 100% of free cash flow.

Dave, thank you for joining us.

Thank you very much for taking the time. Mike, thanks for hosting.

Thank you.