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

Good morning, everybody. Tim Horan here, hosting Cogent. The CEO, Dave Schaeffer, who's attended most of my conferences over the years and always a real pleasure. And thanks again, Dave.

Thank you, Tim and Oppenheimer for hosting us, and I am here to answer questions.

And we have endless questions. If anyone has a question, there's a chat, let me just type my name in, sorry, feel free to fire away on there. So obviously, Dave, there's a huge amount of interest in the name right now, in your stock right now, and it's been fairly volatile. And I guess the key question really is how is the T-Mobile integration going? And what do you think about maybe ways to monetize some of those assets here over time. But why don't you just start out with where are you with the -- integrating the networks and upgrading the data centers, yes.

Yes. So there was actually 3 questions there, Tim. So let's start with the integration. We are 62% of the way through achieving the cost savings that we outlined when the deal was announced, that is $220 million. We're 62% of the way through achieving the cost savings that we had outlined, that was to take 3 years. We're 15 months into the process. So we believe we will achieve those cost savings early. I think we're both ahead of schedule as well as probably likely to exceed the original $220 million target that we initially laid out. Also, ahead of schedule in terms of integrating the networks into a single network. We have outlined that by year-end, we will have wave-enabled 800 carrier-neutral data centers in North America. We have completed that effort in 554 of those facilities. We've actually sold wavelengths in 156 of those 554 facilities. And we are rapidly completing the 4 steps necessary to be able to provision wavelengths across the entire footprint of 800 facilities within a 2-week provisioning window. We anticipate having that work completed by year end. And then on the third point, which is systems, we have completely integrated the systems where there is one billing system, one customer care system, one set of provisionings, one CRM system. So I feel that we are doing very well on the integration of the Sprint assets into the Cogent business.

So obviously, the big kahuna here is once you get the networks upgraded and the provisioning cycle times down is the wavelength market. Could you talk about the demand that you're seeing out there and the backlog that you think is out there? And just maybe any updates on what you think the size of the market might be and what it's growing at?

Yes. So the total addressable markets for wavelength services globally is about $7 billion, $3.5 billion of that is in North America, where we are focused. Of that $3.5 billion, $2 billion is intercity wavelengths, $1.5 billion is intracity or metro wavelengths. We are primarily focused on the intercity portion of the market. We have installed about 800 waves to date. We have about 2,700 in our backlog. We have been reluctant to install wavelengths out of that backlog because they need to be done on a manual, somewhat bespoke, process as opposed to a fully automated process. What we are doing and enabling our network is designing it in a way that we can go from any data center to any data center with a wave in 2 weeks. We can auto group those waves through our CRM system, produce a KMZ or route map with 1 meter accuracy [ that's ] part of the growth and because of the ubiquity of the footprint and the uniqueness of the routes, we have some key differentiators from the current players in the market. We've also been somewhat surprised at how much customers appreciate a new entrant into the market. It's -- what was surprising to us to have customers say, "We're unhappy with our current supplier." Normally, from a negotiating perspective, they would prefer to say, "Oh, well, you got to win our business." And we've seen just a very strong set of demand from a broad set of potential customers. We have been reluctant to provision that as quickly through the manual process because if we divert resources, we may put in jeopardy our ability to do this in an automated way across the entire footprint. So we feel pretty good about where we are in the wave process and the demand. In the IP transit market, it took Cogent 17 years to achieve 25% market share. Here, we're confident we'll get to that same level of market penetration, 25%, in a 5-year period. And I think there are 3 reasons for that: One, we have credibility with customers. Two, we have an existing sales force that calls on these customers. And three, most of these customers are already today buying transit services from Cogent and just expanding our product book by one additional product is much easier than winning a customer for the first time.

That's great color. So the 2,700 lengths in backlog, you're not even really out there marketing all that much at this point, I'm assuming you're just letting customers know you have it but you're not really taking like many orders. Is that fair?

That's right, Tim. We've been reluctant to sell and scale. We also -- we do realize that without being able to give customers a specific install date, we probably will have some fallout out of that funnel. But we also know that once we demonstrate that we can provision the way we are telling people we can, just as we have done in IP, it will become a self-fulfilling prophecy and that customers will have confidence in us and be able to then buy more services. And we think we'll see an acceleration in the rate of new signing early in '25 as well as the -- working through of the existing funnel.

Well, I mean, our channel checks always suggest you guys have some of the highest quality services globally of any communications carrier and that reputation, everybody kind of is well, well, well aware of that. Could you talk about what the average speed of a wavelength is and what's the average ARPU?

Yes. So a wavelength comes in 3 speeds: 10 gig, 100 gig and 400 gig. A wavelength price is determined in 3 dimensions: The speed, the length of the wavelength and the length of the contract or the services. Our ARPUs actually ticked up slightly this quarter to $1,670 per wavelength. I think somewhere between $1,600 and $2,000 is a reasonable ARPU to model as we get to scale. The dominant wavelength speed today is 100 gig. There are still 10 gigs being sold and only a handful of 400-gig waves are today being installed, but we've enabled the network to support 400 gigs at all locations. We also know that over time, there will be a migration just as there was from 100 to 400 to 800 and 1.6 terabits. So we've designed the architecture to be able to scale to those higher speeds as they become commercially available.

And what's the industry standard now for provisioning time? And how different -- or how important and how different is your -- having alternative rights of way or alternative routes essentially?

Yes. So if we elect to provision a wavelength and we only do that if the customer really pressures us, we prefer to do it in this more automated way. We would have a 90 to 120-day provisioning window. That provisioning would take at least 6 field dispatches to complete and would require custom engineering. What we are doing is deploying our perspective base's transponder shelves in all of the data centers. We are configuring the line systems so they are completely transparent. We are deploying ROADMs or reconfigurable add-drop multiplexers at key intersection points, and with that employing the work effort that is necessary to deploy a wavelength on a point-to-point basis for any data center to any data center is plugging in optics at those 2 end points that would require 2 field dispatches and then the wave path being constructed by the network operations center, and we can do that just the same way we do it for IP services in a less than 2-week window.

And what is like someone like Lumen provision now time wise?

Both Lumen and Zayo would install a wavelength the same way Cogent does today, which requires these multiple field deployments and a custom engineering package. We had the advantage of taking over a fully built network with no traffic on it that was built for voice. There is both good and bad in that [ statement. ] The bad news is we had to clean out all of the old voice equipment and repurpose the network. The good is we can architect it from the ground up in a way that would be much more efficient for this product. Companies like Lumen and Zayo sell hundreds of products across their network, I think they've recently announced they are hoping to reduce their product portfolio to just 300 products. Cogent, prior to selling waves, literally had 3 products now it has 4. So product standardization, system standardization, we have one provisioning system, one CRM system, one network management system, one billing package globally. That makes us very different in terms of how quickly we can provision and then how accurately we can support that customer once they are provisioned.

And now important is having the alternative route?

What were you saying, Tim?

How important is having an alternative route?

Extremely important. So every player has all of the major U.S. city payers. Today, on a nationwide basis, you could buy from Lumen or Zayo and get any of the 100 major markets in America connected. There are at least 3 or 4 regional players that can do that within their region. Crown, typically in the East. Uniti, Southeast. And Windstream, Southeast, Midwest. Cogent will have that ubiquitous set of city payers, but 90% of the routes that the Sprint fiber occupies are unique to Sprint because a wavelength is an unprotected product, having that diversity of right of way is extremely important. So our advantages are not only speed to provision, not only ubiquity of footprint, not only the uniqueness of routes, but ultimately, the biggest lever we're going to have to pull to gain market share is our ability to price at whatever the market requires. We have a network that we have a 0 cost basis in, which is a competitive advantage that others that are saddled with debt in building their networks can't compete with.

That's for sure. Just on maybe that point, Lumen announced last week or the week before, sorry, that they maybe have up to like $12 billion of dark fiber sales. Do you think that could maybe improve their competitive advantage over you or maybe cannibalize the wavelength market somewhat?

So we fully intend to also sell dark fiber next year once we have completed the wave enablement of the network, the same resources would be necessary for dark fiber provisioning as well as doing this wave enablement. So Cogent has built its IP network out of buying dark fiber from a total of 356 vendors around the world, including Lumen. Lumen actually regretted selling fiber to us. I had conversations with 2 previous CEOs who both said they enabled us to compete with them, and they wish they had never sold those fiber. In fact, we could have gotten that fiber elsewhere. For the past 20 years, Lumen has been very reluctant to sell fiber. I think now based on their liquidity profile, they are selling them. The deal will -- or at least the deal that was announced was a deal that we looked at as well and chose not to participate in. That deal required a significant amount of capital to build into brand new data centers with only 1 tenant. These proprietary data centers don't make economic sense. There's too much [ monopsony ] power on the part of the customer. The assets would be fully stranded if you've lost that customer, and then those proprietary builds were joined to their inventory. It's also important to remember that these deals are typically 20 to 30 years and like our deal with Lumen or the predecessor, Level 3, was, in fact, a 30-year deal. So while some of the proceeds may come in upfront, the significant portion of that will be paid out over time. And we know from our negotiations with a couple of the hyperscalers that we interacted with, they were looking for us to do those builds at low single-digit rates of return. And quite honestly, that's below our cost of capital for that reason, we elected not to participate in those processes.

So do you think that will have any impact on the market at all positively or negatively?

So it clearly takes some demand out. So if you were looking for the cheapest and easiest way to move bits, you would do that over the public Internet. The next most expensive or inexpensive way to do it, but about 2.5x more expensive is over a wavelength. A third way, which is even more expensive is to go out and buy dark fiber and then build your own optronics on that. Typically, you're not utilizing the fiber as effectively as a carrier does and therefore, your cost per bit is higher. The fourth and most expensive way to move a bit is to go out and build the fiber yourself. So selling dark fiber will, in fact, siphon off some wavelength demand, but it is demand that is paying a substantial premium above a wavelength for an incremental amount of control. I think with the hyperscalers as well as others, we have seen that they use a combination of all 4 technologies. We're selling transit to Microsoft and Amazon and others, but we also sell wavelengths, we will sell dark fiber. We are not a construction company. So I think there's just a continuum of products that customers want, and we feel there's more-than-adequate addressable market for Cogent to hit its revenue targets in the wavelength market.

Out of curiosity, if you were to sell a payer a fiber as dark fiber nationwide, how much would you get for that about?

Well, I can tell you what we pay for a payer nationwide, which was $54 million, that's public record. It's in our filings. Now that was some time ago, but dark fiber prices have not really moved. I think we would typically sell dark fiber on a route-by-route basis rather than on a nationwide footprint. That means you can still assemble a footprint, different routes have different value and to be candid, our inventory varies route by route. The Sprint fiber is direct buried. That actually gave it an advantage of not having many cuts and therefore, being high quality, but it also means that it is not very easy to add more fiber to that given route.

Got it. Got it. And I think you do have probably 30 payers of fiber -- dark fiber you could sell on average.

It really varies, Tim. Our routes go anywhere from 24 payer up to 144 and it just varies on span by span-by-span basis.

Got it. Got it. Helpful. So if you were to guess how much do you think you can get in dark fiber sales in the next 5 years?

Again, I want to be clear, there are 3 assets, you asked this in your first question, that we have not built into our revenue models. Those are the sale of excess data center capacity, the sale of dark fiber and the sale of IP addresses. I think if you looked at current pricing at our inventory, it would be realistic to assume that there's at least $1 billion of value that could be had from that dark fiber sale...

Just on dark fiber. How about for colo and the IP addresses, how much do you think you could sell them for over the next few years?

Yes. So on the IP address space, we have 37.9 million addresses. We are currently leasing out about 12.5 million of those. We have about 2 million that have been allocated in no charge. So we have roughly 23 million excess addresses. Market price is around $50 an address. So well over $1 billion if we chose to sell those. We are evaluating an opportunity while at the same time continuing to lease out addresses and we have raised places for those reasons. We also have a data center footprint that has 1 million square feet of excess space, 108 megawatts of inbound power provision, 88 megawatts of protected power in place today, and we are working to either sell or lease that footprint. If we sell it, it would be at $10 million a megawatt. If we leased it, it would be at about $1 million per megawatt per year on a triple net basis with CPI increases, we would lease in terms of anywhere from 3 to 20 years. We are in discussions with multiple counterparties for multiple facilities. None of these facilities were built as data centers. They were telephone central offices for tandem switches that are being converted. And by year-end, we will have at least 10 of these facilities converted and ready to be fully marketable and someone could close on them and put in servers. We're continuing that work, and that will bleed over in the first quarter, but we've begun the marketing process. On the dark fiber, we haven't done anything yet. We've shared our inventory with people, but we've been very clear, we have to do the wave enablement. And on the IP addresses, we are having discussions with other counterparties, but at the same time, we've raised prices to our installed base by almost 50% going from $0.29 an address to $0.51 an address.

I mean so you have potential to sell well over $2 billion of assets that's not in the model. I don't think anyone's really modeled out. Can you just talk about the overall interest level and demand level just at a high level?

Yes, sure. So I'll take each of them. On the data centers, we went out to 133 counterparties. We are in active discussions with 58 of them. There are a total of 20 facilities that we are discussing. Some parties are interested in multiple facilities. Some are interested only in single facilities, and we're having those discussions. Those parties include private equity, regional data center operators, private equity-backed management teams looking for a platform and actually some hyperscalers looking for edge locations. So we have 4 different types of potential buyers and we're in discussions with all of them. For the dark fiber, we have had dozens of inbound requests. We have not initiated active conversations based on the fact that we know that we don't have the resources to provision until next year. And then on the IP addresses, we have talked to probably a half a dozen significant buyers. I think our conclusion was with Amazon and Microsoft not actively buying right now, this is not the optimal market to be a seller in. And therefore, we have decided to raise pricing on the leasing side and try to accelerate that leasing activity as I think was demonstrated in our last results. And then after the market kind of firms up and maybe some of these key buyers return, we'll revisit the sale market.

Got it. And just to clarify, I think you said you raised prices for the addresses for existing customers. I think it was for new customers. Or have you done it for existing customers already? And what is the new price?

Yes. So our leasing today for new address, this is $0.51 per address per month. Microsoft and Amazon are at $3.60, our historic pricing had been at $0.29 per address per month.

And you've raised prices for all the legacy customers now as well as new customers?

No. So we've only raised, starting in April, prices on the new customers. We ran a full quarter and measured the impact of that, it was fairly de minimis in terms of aggregate demand. So now we are pivoting and September 1, we are going to be raising prices on a about a quarter of the installed base. We will measure the impact of that price increase on churn. Churn has historically been very low at 0.8/10 of 1% annually. That is much lower than our bandwidth churn of 1.2% per month. If we do not see an acceleration in churn, we will then go and raise prices on the remainder of the installed base, and then after measuring that impact for several months, we actually may go back to the new sales and do a subsequent price increase. This is a scarce resource that is in demand. We have a significant pricing umbrella with Microsoft and Amazon establishing the prices that they have in the market. Their goal is to maximize value out of this asset.

Very good, very good. It seems like to me, you should probably be able to charge half of what Amazon is charging, which is $40 a year as you mentioned, and you're at $6. So it seems like you do have a lot of room to raise prices there over time.

I think we do. But I want to temper expectations in 3 respects: One, we don't have the brand and marketing prowess of Amazon or Microsoft. Two, we have a different customer base, 85% of our customers are service providers and 100% of theirs are corporate end users. So there's a difference in the distribution model, the difference in the customer type. And then I think in terms of the ultimate use, there is a difference. I think most of our customers are facilitating either their content delivery or their access customers, whereas, I think, Amazon and Microsoft are facilitating customers locating virtual servers in their cloud environment.

So Dave, we're almost out of time. We'll go a couple of minutes over because we started a few minutes late. But I am getting a question here. Could you talk about your legacy business, the corporate business and the transit business. If you break it up from Sprint, roughly, was there any real change in the revenue or EBITDA growth rates in those businesses since you acquired Sprint's asset?

Yes. So historically, Cogent had sold IP services to corporate users and service providers. Our corporate business had declined because of the pandemic, and it returned to positive growth. And in fact, it was growing as last quarter and this quarter at about 4% year-over-year. Our NetCentric business where we sell to service providers had average 9% growth, had peaked during the pandemic at 26% and has slowed to about 4.5% this most recent quarter. The quarter before was at about 11%. There is some seasonality slowdown that normally occurs. But in general, our corporate business has stabilized at a much better level than it was pre pandemic. It is growing. We also have acquired corporate business where we are managing out those negative margin products and low-margin corporate customers. I spoke about that on the earnings call. And then on the NetCentric side, we are continuing to see healthy volume increases and relatively consistent price declines. We are continuing to gain market share. So kind of the legacy Cogent business is growing at mid-single digits. The acquired Sprint businesses, enterprise, corporate are actually declining due to these product terminations, and we are growing our high-margin wavelength business organically. So when we blend those 3 components together, we anticipate the entire business growing at between 5% and 7% annually.

And so I do have one question here. Can you just update us on the vacancy rates in your office buildings of the legacy corporate business? And how is that starting -- or trending and how you expect it to trend the next year?

Yes. So across our national footprint, the vacancy rate in those buildings is at around 15% and that is down from the -- just on 17% and change that it peaked at, at the worst of the pandemic. There is a modest improvement, but definitely not back down to the 6% vacancy rate in our 1,868 MTOBs that we were at pre-pandemic.

How long did it take to go from 17% to 15%?

It's taken 2 years, it's stopping even. You operate sometime of the year in South Florida. That market is very tight. And it's hard to get off a space running at high, mid-90s occupancy, whereas D.C. has now taken the baton as the market with the highest vacancy rate at 21% across the D.C. market. San Francisco is second at about 18.5%.

Well, great. We've run over a little bit. But Dave, I really, really appreciate the time. A very interesting time for the company and the whole industry and congratulations on what I think is a phenomenal asset, and good luck monetizing it.

Well, thank you for your interest. Thank you, investors, for your time. And Tim, we'll be talking soon. Take care guys. Thanks.

Absolutely. Stay safe, bye.