Good morning, everyone. Thanks for attending our fireside with Cogent Communications. My name is Jiten Joshi, and I'm the Deutsche Bank High Yield Telco Analyst. It's my pleasure to be hosting Dave Schaeffer. Welcome, Dave. Thank you. He's the CEO of the company. I think we all in the room know that. He's very, very present and in front of investors at all times, and we appreciate him coming to our conference here as well. So Dave, welcome.
Well, thank you, Jiten, and thank Deutsche Bank for a great venue, and I'd like to thank all the investors for taking some time out of our day to hear a little bit about Cogent.
Fantastic. So, you know, obviously, we're gonna talk about the Sprint acquisition. That's the most topical thing for the company right now. But before we get there, you know, let's maybe talk about the corporate side a little bit and the core business. You had said in the Q2 press release that you'd started during Q2 to see a little bit of improvement in vacancy, you know, lower vacancies and higher occupancy. Midway through quarter three, that. You know, at this point, have you seen that sustained? Do you? Are you saying that there could be some runway here, or is there still uncertainty out there?
Well, Q3 is over, and we're in Q4.
Sorry, in the middle of Q4.
Fair enough,
Beginning of Q4.
So just to remind investors, within Cogent, there are kind of three pieces. There's the classic Cogent corporate business, where we sell internet access and VPN services in approximately 1 billion sq ft of multi-tenant office space in the central business districts of major cities. That represented roughly 60% of classic Cogent's revenues. We also sell wholesale bulk internet connectivity in 51 countries and 1,600 carrier-neutral data centers around the world. Jiten is talking about our corporate core business. That business had consistently grown prior to the pandemic at about 11% a year for 15 years. Since the pandemic, it had been experiencing negative year-over-year growth, and it is now finally positive. So at its trough, that business went to -9% year-over-year growth. It is back to +1%. We are continuing to see improvements, and those improvements come in three dimensions.
Maybe the most important is customers' willingness to make a decision. During the pandemic, many companies put their IT reconfigurations on hold. Now that companies are back to whatever the new normal is, they're much more willing to make those architectural reconfigurations and buy more bandwidth. The second thing that's been a positive for our business is, throughout the pandemic, the need for higher bandwidth applications has continued to increase. Companies continue to put more applications in a remote or cloud environment. They support hybrid workers who need an ad hoc VPN, and virtually all of the software that a modern company uses is sold as a service as opposed to a perpetual license on-prem. All of this is driving the need for more bandwidth per capita.
Then the final trend has been, and it's no surprise to people who walk up and down streets, central business districts of cities saw an exodus and are now seeing employees return, but that pace of return is not consistent across all geographies. We are seeing vacancy rates decline, occupancy rates increase, customers making decisions. For all those reasons, we're back at about + 1% growth in Q2, and I expect that trend to continue going forward.
Okay. That's, that's certainly encouraging. Can you kind of... You're talking about the customer mix on the corporate side. Can you just run through that, again, for the audience? Is it, are you more... You're in large buildings, but the average profile of your corporate customer is more in the SMB?
Yeah. So, Cogent's typical corporate customer will generally have three locations. They will have about 30 employees at each location and be occupying about 8,000-10,000 sq ft at that location. Now, we typically start that relationship by selling their primary location an on-net service, and then we sell off-net services in addition to on-net at additional locations. So within our corporate segment, 20% of the connections are off-net. That means we buy a local loop from a third party. It's a building that does not have enough demand to justify the capital for us to build into.... 40% of our corporate revenues come from off-net services. Those corporate customers typically buy about three-quarters of their purchases are Internet access, about a quarter of their purchases are VPN services.
Those businesses are located in 1,850 skyscrapers that average 41 stories in height, 550,000 sq ft, and would typically house about 51 tenants pre-pandemic. What we have seen is the vacancy rate in that footprint roughly triple, the number of tenants decline to about 40. We are seeing that number increase now as companies come back, but the new tenants tend to take smaller floor points. In the long run, that should be a positive for Cogent, because the same 550,000-sq-ft building, with each tenant taking 20% less space, will give us an addressable market of nearly 60, as opposed to the 51 businesses we had pre-pandemic. Now, Cogent also sells to large enterprise customers. This is a result of the Sprint acquisition. Those large enterprise customers typically have hundreds of locations.
The majority of those locations are off-net. 97% of the revenues that Sprint was or excuse me, 93% of the revenues Sprint was, were deriving were off-net, and only 7% on-net, as opposed to Cogent, which is 75% on-net, 25% off-net. So one of the ways we're improving the profitability of those enterprise customers is migrating them from off-net to on-net.
What percentage do you think of the current Sprint corporate base can move on-net with reasonable CapEx?
We think we'll eventually be able to bring about half of that revenue on-net, and that's going to come in two locations. That's going to come from those customers that have offices and buildings that we're in for end use and for those customers buying services and data centers. Sprint was only connected to 24 carrier-neutral data centers. Cogent is connected to over 1,600. So it's very easy for us in those data centers to put those customers on-net.
Have you discussed the margin difference, both in the legacy Cogent for the on-net versus the off-net, and what the Sprint profile will look like? Looks like currently with the 97% off-net, and what kind of an improvement you'll get?
Yeah. The on-net business is inherently much more profitable. It carries a 100% gross margin contribution and 95% EBITDA. In an off-net service, roughly 50% of the cost of that service goes to buy that local loop. So by definition, it's a 50% gross margin and about a 45% EBITDA margin contribution. In the acquired Sprint business, those roughly 1,400 large enterprise customers and other customers are improving their margins with Cogent three ways. One, we're eliminating non-core products that were actually gross margin negative. Two, we are requiring those customers to only get connectivity to us via fiber, which would mean higher reliability and higher throughput and lower costs.
And then third, for those customers that operate in very exotic markets, we are assisting the customers in purchasing the loops themselves, as opposed to having Cogent purchase those off-net circuits in places like China, which are regulatorily very burdensome. So in the acquired Sprint business, it was negative EBITDA of about 70% at acquisition. We believe over a 4-year period, we will be able to improve those margins and stabilize that business at approximately a 20% positive EBITDA margin. Cogent's EBITDA margins on a blended basis prior to the acquisition were approximately 39%. So much better because of a higher mix of on-net services. In our net-centric segment, which is the third type of customer we sell to, that business is almost 90% on-net and only about 10% off.
Right. So, you know, we've talked obviously a lot about Sprint. As I said, it's very topical at the moment. You've closed the deal on May first of this year. And what I wanted to sort of kick off with is, I think there's probably naturally a bit of skepticism, right? On the value of what you bought. You effectively bought it for $1, and on top of that, you have an IP transit services agreement where you're going to be getting $700 million in cash payments over time. So, maybe you could just give investors a little bit of, you know, the history of the asset so we can understand it better.
What you saw in the asset that T-Mobile or frankly, any other buyer did not see, and why this deal, you know, makes sense?
Yes, so Jiten, there, there are actually two very different assets that we acquired, but they needed to be acquired together. The first of those is the physical network. It was a network built along railroad right of way, with fiber buried, direct buried six feet under the track to deliver long-distance telephone services. 19,000 route miles of inter-city fiber, 1,200 miles of metropolitan fiber, and 485 fee simple-owned buildings, of which 45 are large tandem switch sites. This network was built in the mid and late 1980s at a capital cost of $20.5 billion. It was the country's first nationwide fiber optic network. At that time, all long distance was delivered via fixed microwave by AT&T and MCI. MCI actually stood for Microwave Communications, Inc. Sprint was a number three player in that market.
It realized it could not out-brand or out-market AT&T or MCI, so in 2002, it had agreed to sell itself to MCI for $129 billion. That deal was ultimately blocked by the Justice Department and the European regulators. At that point, Sprint elected to under-invest in its wireline business and pivot to become the country's first nationwide mobile phone operator. Again, remember, you know, for investors, at that time, there were two licenses, an A and B license. The A license was always given to one of the seven RBOCs, and the B license was typically an entrepreneurial endeavor, and Craig McCaw was rolling those up, but it was very far from having a national footprint. McCaw eventually sold to AT&T Long Lines. The Bell companies eventually consolidated.
Sprint did build a nationwide wireless footprint, but again, found itself with a higher cost of capital and out-marketed by Verizon, AT&T, and T-Mobile. Sprint sold itself to T-Mobile. At that point, the wireline business had atrophied to $1.1 billion in revenue and zero in EBITDA. T-Mobile initially hired a consultant to figure out what to do with the business. They came back with the same answer that Cogent came up with: turn it into a wavelength network and turn the switch sites into data centers. They attempted to do that and realized it was going to be too hard. They didn't have a management team focused on it, and it would not be material enough relative to their wireless business. At that point, in late 2019, they took a different tack.
They hired a different consultant and did a wind down analysis, said, "Let's just shut this business down." They concluded it would cost about $1.5 billion and could potentially damage the T-Mobile brand. At that point, Cogent stepped in, acquired the asset, and when we acquired it, we acquired two very different things. We acquired a physical network for $1 that we had appraised by KPMG in its current state at $1 billion. That network is fallow. It's much like an empty building with no tenants. It needs to have a purpose. Our purpose is to connect that network to our metropolitan footprint, offload some of our IP traffic onto it, and then use that excess capacity to sell wavelengths and dark fiber. The second thing that we acquired was a remaining enterprise customer base buying 28 products.
Those customers were 1,396 of them. They were spending $560 million in revenue, and it was cash flow negative $330 million. $300 million of negative EBITDA, $30 million of CapEx. And again, you can go look at this all in T-Mobile's public disclosures at the point of acquisition. We requested that T-Mobile shut down 24 of those 28 products. That process is underway. The deal did close sooner than we expected. It was about $480 million of revenue going to about $440. We are end-of-lifing those gross margin negative products as quickly as possible. We are going to retain those enterprise customers.... we will consolidate network assets, we will rightsize headcount, and we will be able to turn that business into a roughly 20% positive margin business with no growth.
You know, the enterprise customer base will be buying two services: internet access and VPN services. Those VPNs are typically delivered over MPLS. That's an antiquated technology. We will support it, but we don't see a lot of growth there.
Got it. So we'll get to the numbers again, as to what the trajectory is going to be to get it from what the current run rate is, which I think you said is a negative $180 million-
That's correct.
To 20% margin. But before, you know, we get to that, let's talk about wavelengths a little bit. You know, before we get to what it is, I want to ask you about what you need to do to get the network ready to sell the wavelength product. But from your perspective, can you describe the wavelength market the way you see it, the size, the players and the competitors you're going to be going up against, the demand drivers, and, you know, essentially, how do you plan on inserting yourself into that market, which is currently, I believe, dominated by Lumen?
Yes. Let's first of all, start with what we're going to be selling.
Right.
What is the product? So the Internet is the most ubiquitous, easiest to use, and cheapest way to move data around. However, there are three attributes of a wavelength network that will justify a premium from customers. One, it's deterministic. You know exactly how long it takes to get from point A to point B. Two, it's well suited for very large file transfers, and three, it's totally secure.
Before you, could you just explain the deterministic part of it?
It just means you know where point A and point Z is, and you know the exact path. The Internet was designed to survive a nuclear attack. There is no concept on the Internet of end-to-end connectivity. When you type something in on your computer, the only thing that computer does is it knows what the next router is, where it's sending the packet. The Internet dynamically, real time, optimizes how that packet gets from its origin to its destination and back by a series of those next hop determinations. If it's inside of your network, it's known as IGP, Interior Gateway Protocol. If it's between networks, it's BGP or Border Gateway Protocol, and that's how packets move.
The average packet goes through about 8.5 routers between origin and destination, and two endpoints may end up with dozens of different physical paths between them based on traffic at that instant in time. A deterministic route says, "I have a fixed circuit from location A to location Z. It never varies, and I know exactly how many milliseconds, and there is no restriction on packet size going down that route." So the customers for those services, those deterministic services, are content producers, regional access, and large corporates. Cogent already has relationships with virtually all of those potential customers.
When you say content producers, you're effectively talking about hyperscalers?
Hyperscalers, some CDNs, some hosting companies-
Okay.
But really think of companies like Amazon, Microsoft, Google, anyone doing AI, anyone doing large data replication would be the customer base. They tend to want to move very large files on a fixed schedule, which is not conducive to the Internet. So for those customers, we will compete with Lumen, Zayo, and others. What Cogent will have is an advantage. 90% of our routes are unique to us, meaning the physical path the fiber is on is not shared. That's not true of our competitors, who oftentimes are on railroads or pipelines that are shared. Secondly, we will have more endpoints on-net because we sell transit today and more endpoints than anyone else by physically interconnecting the networks. Third, because we are designing this network solely to deliver wavelengths, we can optimize it to be able to provision those much more quickly.
We anticipate being able to deliver a wavelength in 800 endpoints. Any permutation of those 800 can be delivered in a 2-week window. That is not where the market is today. Today, it's more like 350 locations and 3-6 months to provision. And then finally, because we have an asset that we paid $1 for, we will be able to have very aggressive pricing. Remember, we were paid $700 million in the form of a transit agreement to take over a money-losing enterprise business.... we will probably spend $400 of that $700 stabilizing that business. To your question about the trajectory to positive EBITDA, it was - $300 at signing. It was down to - $190 at closing within a year of closing. So the run rate in May of 2024 will be - $80.
By May of 2025, it'll be breakeven, and by mid-year 2026, should be about a 20% margin business through all of those restructuring efforts that I described.
What type of revenue potential are you looking at retaining and growing to at that point?
We think the acquired business will be flat at about $440 million-$450 million in revenue. We think the core Cogent business, which was a little over $600 million run rate at acquisition, had grown at an average rate of 10%, had slowed to a 5% average growth rate organically during the post-depression or post-COVID era, I think will return to being about a 10% growing business. So $600 million growing at about 10%, $450 million flat, and then finally, an embryonic wavelength business that will grow at an astronomical rate for the next few years. Over the next seven years, it should linearly grow from the $8 million run rate we inherited to $700 million, relatively linearly. And that business, being an on-net service, will carry 95% contribution margins.
We've guided to five years post-acquisition, so that's kind of a May 2028 run rate of a $1.5 billion revenue, $500 million EBITDA business, requiring about a $100 million of CapEx.
In cumulatively or every year?
Annually.
Okay.
Yeah, but that is for the entire business.
Right.
The capital is really in three major spending categories. The core Cogent network requires about $35 million of maintenance CapEx. That's been pretty static. The acquired Sprint network requires about $30 million of maintenance CapEx. And then finally, the expansion footprint augmentation is about a $30 million a year expense.
Right. Just to go back a second to the wavelength business, you'll be a new entrant. You talked about the flexibility you have because you effectively have a zero basis in the business, flexibility you have with respect to pricing, underpricing the product. But this is more differentiated wavelengths than standard transit.
That's correct.
So you mentioned that you have the ability to do that, but doesn't sound like you will take that path. So what is... What are the bells and whistles, if you will, that you'll be able to offer that Lumen can't, or that you can offer better to be able to pull away, for example, a hyperscaler or a large bank, to your network over time?
So first of all, the physical path, it was built by one company. It was not a roll-up, so we have GIS information with one-meter accuracy for the entire path. We've built a tool to be able to generate maps with that level of accuracy with each quote. None of our competitors can do that. The path will be unique, which provides redundancy and flexibility.
Can you, can you explain that a little bit again? Because you said, "I believe the initial Sprint network was along railway rights-of-way.
That's correct.
But then you also said, "I believe that the competing networks-
Are wrong.
have that element." They're
They're on highway and pipeline right-of-way-
Okay.
for the most part.
All right.
The city pairs are common. If you need to go from Chicago to Houston or Minneapolis to Seattle, any of us can do that. But the physical path of the fiber that we offer is different than what Level 3, Zayo, or others offer.
What if... Maybe this is a basic question, but why, why does that matter?
Because this is an unprotected service. The Internet is designed, if a cut occurs, it constantly reroutes around it. On a wavelength, you buy it between point A and point B, and if there is a physical cut, there is no connectivity between point A and point B. For that reason, virtually all customers buy multiple wavelengths and build a level of physical redundancy. So if your goal was to go, say, from Dallas to Phoenix, you will buy two different paths, maybe one path going up through Denver and back down to Phoenix, another path going through El Paso and Tucson and up to Phoenix. So if a cut occurs on either of those paths, you still have redundancy, whereas in an Internet network, the router makes that decision and reroutes the traffic. In a wavelength network, you need some kind of optical switch to make that rerouting decision.
Understood. While we're still on the topic of what it is that will attract customers to your, to your network, what about the actual fiber? ... That's in place. It's sort of decades old, if you will. Are there concerns about sort of degradation of the signal? How are you going to address that?
So first of all, the number one reason for degradation is splice loss due to cuts. We are a customer on Lumen, for example, and they have experienced twice as many cuts per meter as the Sprint network, which is 10 years older. The Lumen network was built in the late 1990s, the Sprint network in the late 1980s. It was deeply buried in armored cable. The physical fiber itself was the most basic fiber, single-mode fiber or SMF-28. That fiber was thought to be obsolete in the late 1990s. As companies were increasing the throughput per wavelength from 10 - 40 gigs, companies felt that SMF would not scale. That was true. All of the competing vendors deployed a non-zero dispersion-shifted fiber, so MCI, AT&T, Lumen.
In the late 2010s, kind of like 2012, 2013 timeframe, the industry switched from noncoherent to coherent transmission to go from 100 gig- 400 gig- 800 gig to 1.6 terabits per wavelength. In that transition, the Non-Zero Dispersion-Shifted Fiber fails, and the SMF-28 actually works better. So again, I wish I could tell you Sprint was so forward-looking and thought about that. It had no idea. It was just by dumb luck, the technology turned out to work better on the older fiber than the newer fiber.
That's, that's very interesting. Appreciate that. Well, we've really got maybe one or two minutes left. I just want to give the audience the opportunity. If there are any questions out there? Okay, I'll just wrap it up with just one. You know, if you could just kind of give us the bullet points on, you know, all the cost savings that you're going to generate through, from the integration, I think that would be helpful.
Yeah. So we will save $25 million by turning down the least international network of Sprint and putting that on the owned Cogent network. We will take $180 million of costs out of the North American network by eliminating people, redundant sites, optimizing the architecture of the equipment, and then finally, we'll achieve $15 million of savings by migrating 12,900 miles of Cogent-used fiber onto the Sprint network and eliminate that third-party O&M expense. So in total, we will save about $220 million, which is what allows us to get to be a 20% positive margin business.
Right, that'll take roughly three years-
About four years.
On the same trajectory.
That is correct, yeah.
Okay. With that, you know, Dave, thank you. Thank you very much for being with us and appreciate the insights into the company. Thanks again.
Hey, thank you very much for hosting us.
Great. Thank you.
Thank you all for your time.