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

[Audio Gap] Disclosures are available at the registration desk. For those joining us via the webcast, I'm Mike Rollins, covering the communication services and infrastructure categories for Citi Research. I'd like to welcome back to the conference, Founder and CEO of Cogent, Dave Schaeffer. Dave, thanks for joining us today.

Well, thank you, Mike, for hosting me. Thanks, Citi, for such a great venue. And I'd like to thank all the investors here in the room, to those that are joining us via webcast.

Great. Well, as you know, we have microphones set up around the table. I mentioned disclosures are at the registration desk.

And with that out of the way, Dave, always this time of year, we'd like to start with your operating and strategic priorities for 2020.

Cogent is a remarkably consistent business. Each year -- actually, on New Year's Eve, I send out 5 goals to the entire company and those are meant to be actionable things that each department can work on to help us achieve our objectives for the year. Those goals revolve around accelerating revenue growth, increasing profitability. So we are looking to increase the productivity of our corporate sales organization, having them sell additional new logos each month, and therefore, help us accelerate the growth in our customer base. And on the NetCentric side, we continue to look to add additional traffic and new AS numbers or unique networks to our network. We are the most interconnected network in the world. But it's like, you can't be too thin, you can't have too many networks connected to you. It's a good thing. And then on profitability, we constantly look at trying to optimize our cost structure and make sure that we're delivering the highest contribution margins possible. We have guided investors to expect about 200 basis points of EBITDA margin expansion. That's been our historical trend line, and we expect that to continue in 2020.

So maybe to unpack that a little bit. First, let's talk about on the revenue growth side. What are you seeing in NetCentric right now? You've obviously had a pickup in volume growth over the course of 2019. But what are you seeing both from a volume and yield perspective?

So our NetCentric business is approximately 1/3 of our total revenues. It is really a derivative of residential consumers, and our NetCentric customer base falls into 2 major categories: access networks, who distribute our bandwidth to their end-user customers, cable companies, telephone companies, mobile operators around the world. And I think today, we provide upstream connectivity for more than half of the networks around the world that connect end-user customers to the Internet. We continue to see people spending more minutes a day online and using more content per minute, i.e., consuming more bits per second of use. And then the second major group on the NetCentric side are the content players. And while we sell to a broad cross-section of content players, social networking sites, search sites, the dominant traffic generator on our network remains over-the-top video, and we've seen a number of new initiatives in the past year with new OTT launches. We've also seen some of our existing customers modify their product portfolios and I think create more original content. So we are seeing both an acceleration in traffic growth as well as a broadening of our customer base. That has had the effect of helping us moderate our revenue decline on the NetCentric business and start to see possibly positive growth in that business on a revenue basis. The NetCentric business is characterized by 2 key trends: unit volume growth in the industry, running about 25% a year. Generally, Cogent outperforms the industry by a significant margin, approximately 35% year-over-year in the last quarter. And I think similar trends should be expected for 2020. Secondly, average price per megabit falls at about 23% per year. That is due to improvements in technology to deliver Internet service. At the end of the day, when we sell Internet, what we are really doing is selling interface-routed bit miles connected to other networks. And each of those technologies continues to improve on a fairly steep price performance curve. And our ability to grow revenue is, we have market-based rates of price decline, but faster than market-based traffic growth. And I think you're seeing that play out, and we expect to kind of revert back to our historical long-term NetCentric revenue growth of about 8% to 9% per year.

So just to unpack that a little bit further. So you mentioned the 35% you were doing last quarter. This was pre-Disney+, pre-Apple TV. We've got other services, Peacock, HBO Max, and just a broader, as you said, just more consumption across a range of options that are out there for customers. So what are the opportunities for growth to actually go above where it's been over the last few quarters, just given these newer introductions into the ecosystem and people potentially spending more time on their smaller screens and streaming?

Yes. So we have relationships with both of the content producers that you mentioned as well as dozens of other newer entrants and thousands of existing entrants. We have about 4,500 content-generating customers, and we have 6,800 access networks. At the end of 2018, third parties concluded that approximately 16% of all video consumption was done over-the-top. Well, the corollary to that is that 84% was done through a more traditional distribution, usually linear broadcast or cable. We think these new products are going to accelerate the shift of viewership from linear to over-the-top. And I would suspect, within the next 5 years, those percentages to almost completely invert. I think there's a second benefit in that many of the customers will actually spend more time per day if they have more choices. It's hard to imagine how you can spend more than 305 minutes a day watching video. But as you pointed out, small screens, portability, make that much more easily done. And we provide upstream to the largest mobile operators in the world. China Telecom, GEO in India, are both Cogent upstream customers. So I think we benefit on both sides. And these new products have, I think, caught a lot of attention from consumers. I'm probably not the person to opine on the quality of that content or it's hit potential, but what we are seeing is significant traffic growth. The last point I'd like to make is, as these new services are launched, they're typically done with 2 technology platforms running in parallel, one being the provider's own CDN platform that then becomes a direct customer of Cogent buying bandwidth. And then oftentimes, they rely on third-party CDNs for overflow or for flex capacity. And virtually, all of those CDNs are and continue to be Cogent customers. And so we can win the business, either if it's going directly over the OTT's own platform or if it's going through a third-party CDN or any combination there above.

And just to round out the conversation on the volume side, what is the risk that these OTT providers create direct relationships with the ISPs? So, for example, HBO is a channel on cable systems, what if HBO Max just became another channel, so to speak, and in lieu of that, they get direct connections in these data centers and may not need the transit market. Is there a risk that transit somehow could get displaced from a share or volume perspective with these relationships?

So there's actually 2 separate questions in that question, Mike. So the first question is, is content that is originally designed for over-the-top distribution diverted from the Internet? And does it go inside of a walled garden? Does it become a linear channel? And there have been cases of that retrograde motion in content, where content that is designed for OTT becomes just a channel. Typically, though, customers like the flexibility of location, device and time, portability that they get with over-the-top and also the over-the-top products tend to be much less expensive. So the key trend in the pay television world continues to be OTT or core shaving, replacing linear or bundled services. So while there is some of that retro traffic going on to a closed network, most of the trend is in the opposite direction. Now to the second part of that question, which is, why does an OTT provider use transit versus some other method of connecting to access networks? And for every OTT player, they have effectively a 3-part waterfall that they test in getting their traffic to the end user. Their first choice is always to get free connectivity. If any access network will connect to them at no charge, and they're allowed to embed their server infrastructure at no cost in that network, they will take that free option. Typically, globally, that's a very small percentage of content, a few percent. The access networks don't want to give content to free ride. They don't want to necessarily favor one content player over another, and they lack any kind of way to monetize that content through those free relationships. The second method of connectivity is the most prevalent, and that is the purchase of transit. So while there are over 150,000 peering sessions globally, that's basically connections between networks that exchange traffic typically at no charge. Only a few percent of traffic actually goes between a Tier 2 and a Tier 2. Virtually, all traffic goes to one of the major global backbones such as ourselves. We are the second-largest carrier carrying about 22% of the traffic. But we have a dozen competitors in total and probably at least 3 other companies that can handle large volume customers. We're in a very competitive situation. We are the second largest of those players, and we sell at the lowest price point. And because the transit product is so competitive, it becomes the primary way that any content provider delivers a service. And the final and last way to deliver service is through a paid direct connect. And a paid direct connect is actually very problematic for 2 reasons: one, it's hard to administer; two, that paid direct connect is typically an acknowledgment of a monopoly relationship that the service provider has with its customer and therefore, the content player is effectively forced to pay a toll to reach the end user. That actually disrupted Internet traffic globally in the period 2013 through 2015. But post the open Internet order, even though it has been reversed by the current FCC, it's currently in appeal and all of the access networks are providing connectivity as if the order was still in place. I don't see paid direct connects as a material way in which traffic is exchanged. It's just harder to implement, and more importantly, it's more expensive than transit. I'm often asked a question, why does the Internet exist? And why does transit continue to effectively displace every other service? And it's really a simple answer. It's cheaper, it's easier to use and it's more ubiquitous. It's no more complicated than that.

And just to round up the conversation then on NetCentric. You mentioned the pricing and you talked about the average price declines of 23%. When we look at your numbers over the last year, but it's been better in the last couple of quarters, and you've talked about this publicly that the mix of where the volume growth was coming from drove the yield decline higher...

That's correct.

To that 23% average, and you thought at some point you would start to see that normalize. Are we in that direction now, where you're going to continue to see that normalize? Or are there some more mix issues to come?

So the 23% number that we report, and we've reported that consistently now for 15.5 years as a public company, is on a like-for-like basis. So that says, for 10 gigs, what does it cost this year versus next year? For a 100-gig connection, a 50-gig connection, whatever size, it's like-for-like. If, however, we're getting a shift within the customer base, where more of the growth is coming from the larger customers, the volume-weighted average price decline is greater than the nominal rate of price decline. You actually saw an improvement in that last quarter as NetCentric revenue declines moderated and sequentially, revenue improved. I think you'll expect to continue to see those trends in the fourth quarter and throughout 2020. However, it's not a perfect linear path and there could be periods where a very large player has particularly a hit series that drives a lot of growth, and it could push down volume average weighted price. But this general broadening of the content base is very healthy for the pricing environment.

So to then wrap it up. You mentioned that you were hoping that essentially we'd get back to the goal, that 7% to 8%. How far does Cogent get over the next 12 to 18 months?

It's difficult to make a firm prediction because it is a usage-based service, different than corporate, which is connection based. But I think the trends are in place that, first, we'll get to positive year-over-year growth. And I think that will happen very quickly. And I do think it's likely over the next year or 2 that we revert back to our normalized growth rate. And again, I want to remind investors, there have been periods in Cogent's history where our NetCentric growth has actually grown substantially above the average. And usually, that is when there is some new application or new group of content that is hitting its stride and growing substantially faster than the market. And I've been doing this long enough that I think I'm comfortable in saying that I don't buy the argument, everything that will be on the Internet has already been invented and is on the Internet. I am certain when we are here a few years from now, and I do think you and I, hopefully, will still both be here, Mike, and that we'll be talking about some new application we didn't even think of today.

And migrating over to the corporate side. In the past, you talked about the growth opportunities of expanding buildings, getting more penetration in the buildings. And you talked about being more linear that even though there's different buildings, different penetration rates, that you felt good about the mix that you were getting from existing buildings versus the growth in new buildings. Where does that sit today?

It actually sits pretty much at historic trend line. So we have approximately 1,800 corporate buildings that represent about 960 million square feet. So about 11% of the rentable office space in North America is on-net to Cogent. Our definition of on-net is the most literal definition, meaning, we have fiber into the building and on every floor with full distribution rights. So when a customer wants service, we do not have to construct any infrastructure that could turn up the average corporate customer in about 9 days. We guarantee 17, but we average 9. We have been very clear that we're very selective about the buildings we serve. We are still adding buildings, but the pace of corporate building additions has decelerated over the past 5 years, from about 7% a year to about 2.5% last year. And it should be in that same range in 2020. Most of our growth is coming from better penetration in those buildings. And we really have 3 things driving that demand: the migration to cloud-based computing, the migration to SaaS-based software and the migration to over-the-top VPNs using either SD-WAN or VPLS technology. We, today, are about 25% penetrated in those corporate buildings. Our average corporate customer has bought about 1.5 connections, or another way to say that is, half of the customers in those buildings take only Internet, half of the customers buy Internet and VPN from us. We think that we can easily double our penetration in those buildings. We can continue to add buildings selectively. Last year, in 2019, in my real estate operations review as we were going through targeted buildings, we identified 61 buildings in North America under construction that were designed to be multi-tenant of greater than 500,000 feet. We probably have connected about 30 of those so far, probably a couple we'll never get connected for a variety of reasons, usually geographic. And then the remainder will get connected later this year as those buildings generally take a couple of years to be constructed. I have not yet done that review for this year. I would suspect it will be a similar number, and we will target those buildings as well. Against the installed base of buildings that meet our criteria, there are actually only about 80 buildings in North America that we would like to get to that exist today that we have not gotten to out of our target list. And every one of those 80s usually have a story. Usually, there's an environmental issue, there could be an intransient owner, there could be an access issue in terms of some kind of public works in front of the building that just makes bringing fiber to the building uneconomical.

Maybe switching over on the revenue side of corporate. If I take the USF out, which is a pass-through, there's a little deceleration over the last couple of quarters. How much of that's driven by off-net pricing versus just maybe a change in sales mix versus volume?

Almost all are driven by off-net pricing. So our corporate product on-net has about a 3%, 3.5% annual price decline, almost exclusively attributable to customers signing longer contracts. So Cogent's rate card or list price hasn't changed in 15 years. Our average selling price has declined, that 3% to 4% rate, due to lengthening contracts. 60% of our corporate business is on-net, 40% of our corporate revenue is off-net, 20% of our corporate connections are off-net. So the way that relationship develops as a customer buys on-net, it then wants either Internet, VPN or both at a non-on-net location. We will review that location, very rarely will a customer order drive us to bring it on-net. So at that point, we look to see if that building is on-net for some other provider. We have 90 pre-approved providers, approximately 1.4 million buildings and about 4.5 billion square feet of on-net office space. Now on-net to those providers is typically only fiber into the building without riser distribution. So we typically do have an extended DMARC to build, but it's fairly limited because these tend to be much smaller buildings. And what we have seen over the past probably 5 years at an accelerating rate is significant competition from cable against telcos, driving lower off-net pricing. So our off-net ARPUs have been declining at about 8% a year, roughly 3% to 4% of that contract lengthening. And the other 3% to 4% price decline or half of the price decline coming from lower loop input costs. Our group costs represent 50% of the selling price of the product. There's always a 50% gross margin product by design. And we concluded that selling off-net as a stand-alone is not a viable business. And the reason for that is the subscriber acquisition cost actually exceed the gross margin over the customer's life cycle. It's why, I think, virtually all of the smart build providers fail. I think smart build was a unrealistic way of marketing and probably a wrong way to name that business model. We use a different smart technology. We take our on-net customers and try to sell them additional locations. It does result in a higher cost of revenue acquisition and a lower gross margin, but still acceptable for Cogent. So we generate 50% versus 100% gross contribution margin off-net, and our SAC off-net is about 3.5x that of on-net, but it's still substantially lower than what the typical industry subscriber acquisition cost is of selling just off-net of about $13 to acquire $1 of revenue, which turns out not to be NPV-positive.

And so when you -- you mentioned cable competition. Any concerns on cable getting more competitive for business revenues? The incumbents are also hungrier for revenue these days, and they've -- a number of them have reshuffled in different ways to reposition. Anything you're seeing on the competition front that would give you concerns on corporate growth?

You should always be concerned about competition, but let me take each of those separately. Cable tends to be more suburban, tends not to want to pay building owners, so they generally do not serve the largest buildings. And they are not willing to prewire the building, therefore, making installation costs high and installation windows very long. So we see cable extensively adjacent to their cable plant for residential. The final point on cable is just as they do on the residential side, they are very reluctant to sell dumb pipes, they want to sell full bundles. And anyone who's ever had the experience of calling their cable company and trying to turn off services knows that you can't get off the phone, and they'll switch you department to department and have promotion after promotion to make sure you buy a bundle. We, at Cogent, take a different approach. We go in and sell just a dumb pipe, and that really differentiates us. And then in terms of the incumbents, I think the incumbents are in a harvest mode and continue to try to retain legacy enterprise business but are not materially investing in that business. They are reducing their sales force. They're relying much more on channel and indirect, they are continuing to sell high value-added bundles, and again, are reluctant to sell dumb pipes. And whether it be AT&T, Verizon or CenturyLink, the 3 major incumbents, every one of them has a negative top line enterprise revenue growth rate. And I think there's 2 reasons for that: one, the customers no longer want the products they sell; and two, their Internet product is not generating enough revenue to replace the revenue that's going away as those OTT services are deployed.

Two other questions, first on margins and then on the return on capital. So on the margin side, the way I think about maybe the longer-term objectives you've had as a business, it was 10% to 20% revenue target and over 200 basis points of margin expansion. It's now over 10% the long-term revenue target, 200 basis points of margin expansion. Revenue performance -- partly because of NetCentric over the last few quarters, total revenue performance has fallen below that level...

Absolutely. It's about 7%.

But you're still guiding to get to 200 basis points of margin expansion. And as you described many times, you're an incremental margin business. So where are you driving the efficiencies to keep up with the opportunities for the same margin expansion, even though in the recent history, it's been a little bit slower on the revenue front?

So first of all, let's go back and look at historicals. We have averaged about 200 basis points a year for 15 years. Two, our contribution margins have averaged 45%. These are EBITDA contribution margins. And our base EBITDA margins are about 36%. In fact, last year, our EBITDA margin contribution drifted up to approximately 60%. We should continue to drive that type of efficiency. The second point is the 200 basis points probably has a little more flex involved in it than the revenue number. And that if we were hitting the higher revenue targets, we'd probably do better than 200 basis points. And then the final point is that roughly half of that margin expansion has come from gross margin uplift, about half of it has come from SG&A efficiency. That margin expansion will end at some point. But we have said that we should expect to deliver about 200 basis points a year from the roughly 36% margins we're at today until those margins asymptotically get to about 50%. So seeing another 7 or 8 years of margin expansion. And if you look at what we're doing, that's pretty close to what we are delivering. The gross margin could be a little tougher, particularly because of the USF pass-through that you commented on. Because while that's great for revenue, it carries no contribution margin. So if you're adding $1 of revenue with 0 contribution margin, it kind of is a drag. But yet, we're still delivering 100 basis points. On the SG&A side, we're pretty lean on general overhead. Most of our expansion in SG&A is in the S side, and that's very easily controllable. And again, by design, we are adding salespeople, but we're trying to do that in a measured way when we can absorb them. Throwing money at a problem doesn't fix the problem. Building a sales force and making sure that you're training them, bring them into the culture and making them productive is what it's all about.

And by the way, if there's a question, please, light up your microphone and we can get to you. And then finally, just on the return of capital front. You've done some outreach over the last year to investors. As time in the business has evolved, what are your priorities today in terms of choosing between dividend growth, share repurchase? Do you feel the need at all to delever at some point? How do you make the choices of where to deploy cash?

So I think it is the most important question the Board wrestles with. I mean, I think, the purpose of the Board is, one, to hold management accountable, and someday, they may decide I'm the biggest problem. But so far, they've focused on allocating capital. In fact, as we are speaking, there's actually an investor outreach call going on now with some of the Board members and a particular shareholder, and that is a topic. So it is a very important topic for the Board. We have delivered 30 consecutive quarters of sequential growth in dividends. Our dividend growth rate is just under 15%, about 14%. And we expect to be able to continue to grow dividends at a similar rate. We also have episodically bought back stock and have bought back about 20% of our float. We have not been very active in the buyback market over the past year or 2 since markets continue to melt up and are at all-time highs. The Board constantly weighs the tax efficiency of which method of returning capital. We also are slightly delevering. We're below the midpoint of our leverage target at 2.9, 3x net lever, the target range is 2.5 to 3.5x. And for a recurring revenue business with predictable growth, that's a very low leverage threshold in a low interest rate environment. To that end, we did a euro-denominated deal in June of 2019. We have our secured debt that's callable at par in March. We will probably take that out. We will move capital from the operating company up to the holdco when we do that. And then we will look to add additional leverage, both secured and unsecured. The unsecureds are callable at par in a year. So we are working on optimizing our capital structure for the current interest rate environment.

And then just for a couple of moments, if you do that, does that take away any of these builder basket issues that the company has had in the past?

They still continue, but we effectively deplete the basket by moving the money upstairs from the operating company to the holdco. And then as we layer new [ get on ], we start to incur new builder basket tests and have to go several quarters before we move that capital but we have sufficient capital at holdco to meet our dividend buyback -- dividend and buyback goals and have additional flexibility. We could totally eliminate those tests, but they would raise our incremental cost of capital. So what we're trying to do is optimize for lowest cost of capital, maximum amount of flexibility, minimum amount of negative carry. So it's kind of a triangulation on the best capital structure that fits our business. And listen, we're probably not perfect, but we've done a pretty good job on allocating capital.

Dave, thanks for joining us today.

Hey, thank you, Mike, and thank you all for coming.