Good evening, everyone. Welcome back. I'm John Hodulik, Media and Telecom Analyst here at UBS. I'm very pleased to announce our next guest is Dave Schaeffer, the CEO of Cogent Communications. Dave, thanks for being here.
Hey, John. Thanks for hosting me. Thank UBS for a great forum. Most importantly, let me thank all the investors for taking time out of their busy day to hear a little bit about our business.
We've got about 40 minutes to discuss current trends and strategy and of Cogent Communications. I've got a bunch of prepared questions and anybody else please get on the app and shoot them over to me, and I'll weave them into the conversation. Dave, as I always do, as I always ask at this conference, you know, here at the end of 2021, give us a sense of how you think the company performed during the year, and more importantly, what are your priorities as we look out into 2022.
I think, you know, COVID has been a tale of two cities in terms of impact on Cogent. You know, as we look back at 2021, hopefully our growth rate has decelerated, but our outperformance relative to other wireline telecom companies actually accelerated. In our corporate business, we have had significant headwinds as many companies shifted to remote work and remain either completely or partially remote. It's for that reason our corporate business has decelerated from its historic average growth rate of about 2% sequentially to about -1% sequentially. Now, the corporate sector of our business has been improving over the past three quarters, and we expect that trend to continue. We do think it's still a quarter or two before the corporate business turns positive and probably another quarter or two before it regains its historic growth rate.
If we look at the past 16.5 years since Cogent has been public, the corporate business has averaged 11% growth and has only previously had two negative quarters at the end of 2008 and beginning of 2009 during the financial crisis. The pandemic has been longer and more severe, with six quarters of negative growth and probably one or two quarters continuing negative growth before turning positive. The good news, however, is that our NetCentric business has actually accelerated, and we had the best two quarters in the company's history in the last two quarters, as we've seen an acceleration in streaming, a diversification of the streaming content base, the continued internationalization of the streaming market, and the increase in traffic where Cogent is paid both by the sender and the receiver.
Putting these factors together, the NetCentric business, which has averaged 9% top-line growth, actually grew 30% last quarter and grew 25% the quarter before. Now, some of that is foreign exchange, as roughly 55% of the business is outside of the U.S. Even on a constant currency basis, the business is growing at about 25%. Putting this together, roughly 60% of Cogent's total revenues are declining at about 1% sequentially, and 40% of our revenues are actually growing at approximately 7% sequential. That gives Cogent an aggregate positive growth rate and the ability to expand margins. As we look to our priorities for 2022, it's to continue to see the corporate business improve as businesses decide on what their post-pandemic network architecture will require.
We think there will be a return to office, a continuation of an increase in larger symmetrical connections at the primary locations, a acceleration of office to office VPNs, and the migration of ad hoc VPNs from the corporate premise to a carrier neutral data center. All of these trends should accelerate Cogent's corporate growth. On the NetCentric side, we think that while streaming will continue to grow and will revert back to a more normalized growth pattern, and we would expect our NetCentric business to decelerate to kind of historic norms. All in all, we see Cogent's total revenue growth re-accelerating to our historical average of about 10%. Now, our guidance is not meant to be specific to any one quarter or even year, but rather multi-year in nature.
We anticipate being able to deliver about 200 basis points a year of EBITDA margin expansion. That should allow Cogent to continue to grow its dividend. We have 38 consecutive sequential quarters of dividend growth, and we see that continuing into 2022.
I think that's a great summary, and I think we're done. No, I'm kidding. I've got loads of questions to follow up. That was corporate. Let's dive into the corporate business first, and then we'll do the NetCentric business. First, you know, obviously, the pandemic has had a big impact. I mean, can you tell us what you're seeing in terms of the corporate sort of spending environment here in the fourth quarter? And any issues with the Omicron variant or, I mean, are you seeing maybe a slowdown in the return to work or maybe even a reversal in some of the gains that you've seen recently?
The U.S. is not monolithic. Our corporate business is only in the U.S. and Canada. In the South and the West, we've seen good increases in corporate activity and about 60% of employees on a given day returns to their offices. In the Northeast and Northwest, we're actually seeing office occupancies at around 30%. Clearly a hugely divergent customer base. Many companies had initially planned to return to office after Labor Day. Delta variant pushed that out. For many of those companies, they then announced a January return to office. While many companies are sticking to that plan, we have seen some companies push that date out till February or even March. Now, we are seeing a clearer picture of what the new corporate work environment is gonna look like. Most of our customers will have a portion of their workforce hybrid indefinitely.
To remind investors, our corporate product is sold on a flat rate, not a metered service. So it's binary. The customer either has the service or doesn't. We're also seeing companies groom their locations, consolidate within any given MSA to one location, but maintain locations in multiple MSAs. We're also seeing companies begin to revisit MPLS replacement. Many companies prior to the pandemic had expressed interest in replacing those MPLS networks with either SD-WAN or VPLS. Those programs typically were placed on hold during the pandemic. Companies now I think have enough confidence in what their post-pandemic architecture is going to be, that they're requesting proposals for actually ordering some of these services. We are seeing a increase in quotes being issued, orders being processed, and actually customers being installed.
We've also, I think, been through the worst of whatever churn would have existed from customers under financial duress. The good news for Cogent is that by serving only the largest multi-tenant office buildings in central business districts, the landlords did a pretty good job of vetting our customers. We've actually seen our DSOs decline throughout the pandemic, and our bad debt as percentage of revenue also improved. I think this is really a testament to the necessity of the services Cogent sells its corporate customers. Finally, we're seeing a large number of corporate customers start to inquire about a new service, and that is to be able to put their VPN concentration for these remote employees into a data center. Cogent sells internet access, but over that internet connection, the customer basically has three different functions that they can buy from the service.
One is just internet access for general corporate purposes. The other is an office-to-office VPN that's either provided using IPsec encryption at the customer premise, known as SD-WAN, or using a VPLS encapsulation provided by the service provider, therefore being delivered as-a-service VPLS, giving office-to-office isolated connectivity. Then finally, virtually every company has had to be able to support remote workers with a dynamic ad hoc VPN service. For the very largest of customers, companies like UBS, that had already been occurring in a data center. For those mid-sized to smaller customers, they perform that function at their corporate firewall in their primary location. Because companies now realize this is probably a permanent change to their network architecture, many of them are looking to add a redundant ad hoc VPN concentration port in a data center.
This is an incredible long-term opportunity for Cogent, and we think that, coupled with the fact that we're only about 25% penetrated in the 980 million sq ft of office space that we're in, gives Cogent the ability to grow its corporate business at double-digit rates for the next decade.
Got you. You hit on a lot of different topics there. I'm gonna maybe work through a few of them. First, you know, you said the consolidation of satellite offices and branch offices. I mean, have you seen that process? Are we just starting to see that process? I'm just thinking from UBS. I mean, you know, as you said, there's more and more workers probably gonna end up in more of a hybrid environment. Companies are gonna start to look to rationalize some of their space. Is that something that you think will be a bit of a headwind for a period of time or, you know, sort of how should we assess the impact of that?
I think the companies that are trimming those satellite offices within the same metropolitan statistical area have already done so. In many cases, they've exited those leases. In other cases, they've been able to continue to lease space, but turn off whatever telecom services at those offices that they're not intending to reoccupy. You know, the companies that have multiple offices, either across the country or around the world, I think they've made the decision around which of those offices will continue to be staffed. For that reason, we've actually seen this acceleration in request for service, new orders being written and new orders being installed for our corporate customers off of a trough in corporate business, which was probably Q4 of last year.
I think that was the worst environment where companies had no visibility, there was no vaccines available, and there was no plan to return to office. Companies now, I think for the most part, have come up with a plan. They are, though, in different stages of implementing that plan, and I do think incremental variants like Omicron could end up slowing those decisions down, but at least the decisions now have kind of a clear roadmap to go forward.
Got you. Next, can you remind us of sort of, you know, your network reach, how many buildings you've got connected, how you expect that to trend over the next few years? You've mentioned the building penetration levels, and then, you know, as it relates to that 16% in the South and average on the Northeast and Northwest, then you even mentioned the West. You know, where are these buildings? I mean, do you have more leverage right now to the southern part of the U.S., the Northeast or how would you characterize the sort of geographic dispersion of these buildings?
Yeah. Cogent's corporate business model is to focus on the largest multi-tenant office buildings in North America. We have about 1,825 buildings directly on our network, which means we have fiber in the building, up the riser, and can serve customers on every floor and provision services in an average of about 9 Gb. We have about 980 million sq ft. That represents about 11% of the multi-tenant office space market in North America. In addition to that footprint, Cogent has approximately 1,350 carrier neutral colocation data centers on its network in 50 countries around the world. That's significantly more data centers than any other provider. Finally, we have 54 of our own Cogent data centers on-net.
The network that we operate is about 60,000 route miles of inter-city terrestrial fiber, coupled to about 17,000 route miles, about 40,000 fiber miles of metropolitan fiber and wholesale market in about 215 markets around the world. That network is comprised of about 1,030 physical rings in those 215 markets. Attached to those rings are the 1,825 skyscrapers, the 1,350 data centers, and the 54 Cogent data centers. One of those locations is served entirely by our facility, so we need to buy no services from a third party.
We do have an off-net corporate business, which represents about 20% of our corporate connections, about 40% of our corporate revenue in buildings that we have concluded are not economic to serve, meaning we cannot justify building fiber, but we'll buy WISP services from other providers in those buildings. We have about 7,700 of those types of buildings. They tend to be much smaller with far fewer potential customers. We buy those services from 90 different suppliers, and we serve just under 12,000 off-net connections in those buildings. Our primary business, in fact, 81% of revenues come from selling internet access, and the vast majority of that is on-net. We have about 75% of all revenues and about 90% of all connections are directly on-net.
How should we think of ARPU or the trajectory of ARPU in the coming years? I mean, at this point, are you seeing anything? You know, I can imagine that you have these large resellers on the corporate side. A lot of people have you know working from home or in a hybrid environment. At the same time, maybe some of the smaller offices, if they're being consolidated, maybe some of your lower ARPU lines going away. There seem to be some conflicting trends there. I mean, what's your view on and I think at the same time, you're in the process of migrating other legacy connections to 1 Gb connections. There's always this need for faster and faster speed. How does all that fit together and drive ARPU over the next few years?
I'm going to actually answer the question in three parts. First of all, for corporate on-net customers, we have two trends going on. One is the migration from 100 Mb to gigabit, which increases ARPU. The second is the average contract term lengthening, which decreases ARPUs. If you look at our corporate on-net ARPUs, they have declined at about 3% per year. We think that trend will continue at approximately the same pace. In our corporate off-net connections, the rate of decline is between 7% and 8%. It actually has the same two factors that I described for on-net, plus a third factor, which is lower loop costs from those WISP providers. We have approximately 90 different providers who have built fiber into 4 million endpoints. We buy WISP services for those off-net locations that we cannot justify building into.
Our margins are lower, and the quality differentiation is not as great. It is the reason why the vast majority of our corporate connections are in fact on-net, and we sell off-net as an ancillary service to those companies that we have an on-net relationship with. The off-net ARPU declines will continue as there are more fiber overbuilds providing true competition between overbuilders, telcos, and cable companies. Today, we can serve approximately 55% of the office space in North America. About 45% of it through these off-net connections, about 11% through our on-net connections. The third place I'm going to talk about ARPU is in our NetCentric business, which is almost exclusively on-net. Those ARPUs are impacted by four factors. One, whether or not the customer is taking a larger port. So we typically sell 10 Gb and 100 Gb ports.
If the customer migrates to 100 Gb NetCentral ports, that pulls ARPUs up. Secondly, our NetCentric business is a metered service. Therefore, the customer pays for what they actually utilize. We are seeing an average price per megabit decline of 23% per year, in line with the same rate of price declines that we've seen over the past 20 years. Third, we're seeing customers increase their average utilization on those ports. So bigger pipes, and they're using more of those pipes, increasing ARPU. And then finally, about 55% of that NetCentric footprint is outside of the U.S., and we are impacted by currency as the dollar strengthens, that can weaken the reported ARPU in those non-U.S. connections. Putting all of this together, our price per megabit will fall at about 23%.
Our volumes will typically increase faster than that, allowing us to grow revenue, and our ARPUs will be volatile, but generally flat to slightly up on the NetCentric side of the business. We also have benefited from a higher effective yield per megabit as a greater percentage of our sales have come from customers who also terminate on a Cogent connection. Over the past several years, we've seen the percentage of our NetCentric traffic that is two-sided, meaning we're getting paid by the sender and the receiver, increase from 50% to the low 70s%. The other traffic is going from a Cogent customer to a peer, so we're only being paid on one side. The final impact on NetCentric revenues is the size of the customer. The bigger the customer, the lower the price is.
What we have seen is a growing diversification in our customer base, which has helped us increase the effective price per megabit. While the nominal price is declining at 23%, our effective price is declining at a much lower rate, allowing us to have these record growths in NetCentric revenues.
Got you. Last follow-up on, from your earlier comments is SD-WAN uptake. You summarized it. Just remind us, maybe where we are in that process. Again, getting back to UBS, you know, we had a kind of a tech expert meeting where even UBS, you know, company very, very focused on security and privacy and massive compliance issues on a global basis, is a major RFP for SD-WAN. So where are we in that process and how does that help you again, and now that we're kind of recovering, coming out of the pandemic, what does that mean for your customers?
Maybe a slight history on private networking. Private networking began after World War II, and for 30 years was entirely based on private lines rented from phone companies. In the mid-1970s, private networking migrated to Frame Relay, which dominated the market for 20 years to be replaced by MPLS. During the rise of MPLS, the public internet really became ubiquitous and more reliable. As a result, and the fact that the public internet is so much less expensive per bit than these private networks, there have been two major technologies developed to allow companies to build private networks that are fully secure over the public internet.
Much like VoIP overtook plain old telephone services for voice conversations, where streaming companies such as Netflix have displaced linear video, we're seeing the roughly 1.2 million ports of MPLS in North America at peak and a $45 billion spend on those services migrating to these over-the-top strategies. There are really three strategies possible. For the smallest companies, they will just use the internet and not be concerned with security or encryption. For mid-size companies, they will typically use a SD-WAN customer premise device that will take every bit, encrypt that bit to 2 to the 128th power, and then send it over the internet in an encrypted form and decrypt it on the other end, making a highly secure connection utilizing low cost public internet with a piece of customer premise equipment.
The challenge is that equipment does not deliver full line rate throughput for larger companies. Most of the mid-size and larger companies have decided to use a different technology of VPLS encapsulation. What it's basically doing is taking every bit, putting it inside a wrapper that makes it invisible on the public Internet. It's sent down the Internet and then unwrapped at the other end. Both of these technologies, VPLS and SD-WAN, represent about a 15x or 15-fold reduction in the cost per bit, and both technologies are easier to deploy and easier to manage and are more ubiquitous in their availability. Many companies had already begun the process of turning down those MPLS networks before the pandemic started.
Between 2015 and 2020, you saw that market shrink from about $45 billion to about $35 billion in revenues for legacy providers. With the pandemic and the uncertainty around it, many companies who had planned to migrate to SD-WAN or VPLS put those plans on hold. We're now seeing those companies reaccelerate those migration plans, and our expectation is that those legacy MPLS networks will fall by about another 90% over the next five years as that traffic uses one of these two superior technologies at a 15-fold cost advantage.
Just one last question, then we'll move to next one, which is we had an announcement from AWS early last week, and then we talked to Verizon and AT&T about it. Just the notion of private 5G for enterprise. Is this potentially a disruptive technology? You know, we are from our side struggling with how broadly it's gonna be adopted by corporations. You know, what services can potentially be cannibalized. What's your view on that? Has it changed the need for broadband connectivity within enterprises, either by making it larger or potentially migrating applications to these wireless carriers?
I do not believe 5G wireless will be a game changer. If you think about the Internet, the Internet began as an overlay network. It could sit on top of a telephone network, a coax network, or a mobile phone network. What we are seeing is as Internet volumes have grown, the fixed line networks are transitioning from those legacy transmission platforms to fiber. Now, fiber is not completely ubiquitous, but it is becoming much more available both for residential and for businesses. 55% of all businesses today have fiber connectivity to their locations. The deployment of 5G is a much more robust wireless service than, you know, 4G or even going back to original AMPS, you know, mobile telephony, which was at about 0.5 bit per hertz.
5G is an attempt to increase that throughput, and it is successful in doing so, but the cost per bit is about three orders of magnitude more expensive than the cost per bit over a fixed line network. That differential is actually increasing, not shrinking. You know, the oversubscribed shared nature of a 5G network will probably not be either cost-effective or reliable enough for most corporate users. Maybe for the very smallest businesses, those that have two or three employees that rely only on the Internet, yes, you know, a 5G, you know, solution may be sufficient. For an office with even three or four employees and maybe one or two of those employees being remote part of the time, a fixed line fiber-based solution will be much more viable.
You know, I think the reason why you're hearing the wireless carriers propose this is they're struggling to be able to convince investors that there's an additional addressable market for the tens of billions of dollars of capital that they will need to spend to go from LTE to 5G. My belief is that that capital is necessary to maintain market, but not grow market size.
That's probably the concern, you know, in these days. Dave, certainly a pretty significant portion of your business and I'd like to touch on this. You know, obviously the big driver right now being streaming, whether it's new content spend, new subs, new subscribers, new services, sort of the movement of these services from sort of traditionally U.S. to more international. I mean, so we have a media team as well here. I mean, we think it's just beginning, right? If you look at the time spent, and Nielsen actually has some good data on this, the majority of time, at least in the U.S., is still spent on broadcast and cable TV. You know, services like Netflix and YouTube are still surprisingly small as a percentage of total viewing.
It would seem to suggest that's a multi-year opportunity. I guess as part of your opening remarks, you talked about the Netflix business really slowing. Why, I mean, I get that there's law of large numbers here, but why can't that continue as viewership migrates to these platforms both in the U.S. and globally? I would imagine we're even earlier from a lifecycle standpoint outside of the U.S.
The rest of the world is behind the U.S. in terms of streaming adoption, and that's been a tailwind that will continue for Cogent as we derive 55% of our revenues outside of the U.S. Secondly, the number of minutes a day that people consume video has remained relatively constant at slightly over 300 minutes in the developed world. That number may creep up slightly as there is more content available and more choice. Also, the total cost of video consumption declines with streaming. Choice goes up and costs come down. Now, going into the pandemic, we were at about 18% of all video content streamed. That meant 82% of content was delivered linearly through broadcast, cable and satellite or DVDs. The pandemic accelerated that conversion, and today we're at slightly over 40% of all content being streamed.
Now, over the next several years, whether it be three, five or seven years, that 40% is probably gonna rise to somewhere between 80% and 90% of all video is streamed. The other thing that's happening is the quality of that streaming is increasing as we've moved from standard def to high def to 4K, even 8K. Really, you know, what Facebook is proposing with Metaverse is the idea of a much more robust fifth screen that can be much more immersive. Apple, Netflix, Google, Facebook have all been experimenting with augmented reality and virtual reality, and they aren't really different services. I think we'll definitely see augmented reality supplementing very high definition, whether it be 4K or 8K. There's still a shortage of 8K capable monitors.
Even at 4K, there's a lot of advantages for the consumer by having superimposed on the image an additional stream of information that augments reality. While it's not a completely immersive experience, it's something that cannot be replicated on television or in exhibition theaters. That will be a key driver of differentiating streaming content from originally linear content that has ported over to the Internet to be a whole genre of content that is Internet designed for streaming. Ultimately, I think that augmented reality will be supplemented and replaced eventually by true virtual reality, which will be a completely immersive experience that is more interactive than any linear content today. I think we have three big drivers of traffic growth. Number of people connected, number of minutes a day they consume content, and then the bit content that they consume per minute.
All of those are leading towards faster growth. Offsetting that is the fact that technology is continuing to improve. While investors may be familiar with Moore's Law and the roughly 55% price performance improvement in computing, i.e. millions of instructions per second, we're seeing a similar rate of decline, albeit at 23% in the price per megabit. Local access monopolies have retarded the rate of price declines in their monopoly footprints. The wave of overbuilding that is occurring is creating duopoly and triopoly competitive situations, and with that, we will see a much more rapid rate of price decline in the access market. The core of the Internet has declined at about 23% per year for 20 years. The edge has typically only seen price declines of 2%-4% on a per megabit basis.
This was part of the basis for some of the antitrust and net neutrality work that was done several years ago, both by the Department of Justice and FTC, PUC and FCC. Now we're seeing the availability of capital for competitive fiber overbuilding, driving a competitive war where ARPUs are meant to be flat, but service packages are increasing, reducing the price per megabit. All of this is good for Cogent because we make money by hauling those bits across the globe. We're the second largest global carrier. We're the most connected. We have more networks connected to us and more data centers. All of these trends bode well for continued growth in our NetCentric business and the ability to kind of push the law of large numbers issue out for decades.
Okay. You're not changing that really good picture of my cable or wireless coverage with that. I have two more questions. A couple more minutes. I got two more questions for you. First, maybe on the regulatory side, it looks like we're gonna have a new FCC in place here shortly, and they'll probably move forward with net neutrality. I know it was an issue in the past. I mean, how does that potentially affect Cogent at all?
Very different than other network operators, we have historically been a very strong supporter of net neutrality along with the tech community. The reason for that is anything that increases Internet traffic helps Cogent. Legacy providers are trying to retard the rate of Internet adoption and Internet substitution, whether you're a cable company, a mobile phone company, or a fixed line phone company. At the end of the day, the Internet is bad news. It drives up your capital requirements, it accelerates the obsolescence of your legacy network, and it reduces the revenue per bit substantially. The world has moved on, though. You know, the government spoke in 2015 and implemented net neutrality rules. The administration in 2017 attempted to roll those back, but states stepped into the breach and continue to maintain net neutrality principles.
Every major access provider is today adhering to those principles. They should be codified by the FCC with new regulation. At the end of the day, the consumers are now so entrenched in getting services that net neutrality provides, any provider attempting to throttle traffic and basically create a walled garden would be skewered by its customer base and by regulators. I think we have a de facto net neutrality world, and I think we will see a de jure world as well as the FCC finally codifies what states have implemented and what the FCC initiated in 2015. We feel pretty comfortable that the internet is safe. Access networks do have a challenge. It's part of the reason why we have been so hyper selective in choosing only corporate locations in the central business districts and skyscrapers to be an access provider.
We think that in rural markets and residential markets and smaller buildings, the economics of those fiber builds are very challenged.
Last, capital allocation. So a number of ways you guys use excess capital, I would say. I think you're at a higher end of your leverage target. So how should we think of the sort of continued pace of the dividend increases we've seen versus buybacks versus potentially things like that?
We have returned over $1 billion to shareholders, mostly through dividends, roughly three-quarters dividends, about one-quarter through buybacks. We continue to have a buyback program, and we use that in periods of market volatility, but are committed to a growing dividend. We have 38 sequential consecutive quarters of growing our dividend. We actually have excess cash on our balance sheet. We are looking to slowly repatriate that cash to shareholders by pivoting out more than 100% of cash flow and using our balance sheet as a strategic asset to continue to borrow in a low-interest-rate environment. Different than other telecom companies that built their businesses on debt, Cogent built its business debt-free by buying distressed assets. That allowed us a unique position in the market. As interest rates fell, we elected voluntarily to take on debt to accelerate capital returns.
We think our ability to consistently grow our dividend, you know, is pretty safe, and we remain underlevered in a low interest rate environment and will be prudent in using our balance sheet but are committed to returning capital. I think investors should expect a similar pacing of return of capital. You know, our free cash flow growth is in the low teens, 13%, 14%, and our dividend growth kind of mirrors that at 13% or 14%. It feels like, you know, all of the market forces are aligning to allow us to grow the business at an equal or faster pace.
Fantastic. Dave, I think that's all we have time for. Again, you're always very generous with your time. We appreciate all your insights into Cogent, but the rest of the telecom industry. Thanks again.
Thanks, John, and thanks to all the investors for their time. Everyone stay safe. Take care.
Sure.