Recurve Capital 的 Aaron Chan 谈 Cogent $CCOI

You're about to listen to the yet another value podcast with your host me, Andrew Walker. Uh today's podcast is a deep deep dive into Cogent. The ticker there is CCOI. It's with Aaron Chan from Recurve Capital. He has done some incredible work on the name. I'll include a link to some of his writeups in the show notes. You should definitely check them out. We dive really deep into the name. It's a really interesting stock. A lot of my different friends who who run the gamut of concentrated value investors, hedge fund people looking for uh a an inflection point, everything are really interested in the name because it's just it's done some incredible deals. It's got a really interesting CEO. I think you are really going to enjoy this podcast and learn a lot. So, uh first a word from our sponsors and then Aaron Chen. Today's podcast is sponsored by DUPA. Are you still manually updating your financial models after earnings? Ask yourself why. Every quarter, analysts lose hours copying numbers from filings, adjusting templates, and double-checking for errors. It's tedious. It's timeconuming, and it's a terrible use of your time. Dupa changes that. They automate your model updates with near real-time precision using AI that's been trained on thousands of companies filings across every sector. The result, you get a fully updated model in your format with your logic faster than ever before. Every KPI, every footnote, every guidance figure exactly where you need it with source links built in. So stop wasting times on data entry and start focusing on what really matters. Analysis, insights, and alpha generation. DUPA doesn't just save your time, it gives you time back where it matters most. Book a demo with the DUPA team today at dupa.com/demo. That's dupa d a l o pa.com/demo.

All right. Hello and welcome to the yet another value podcast. I'm your host Andrew Walker. With me today, I'm happy to have on for the second time from Recurve Capital, Aaron Chan. Aaron, how's it going?

I'm doing great. How are you?

I'm doing great. I'm super excited for this podcast. So, before we get there, a quick disclaimer: Nothing on this podcast is investing advice. Full disclaimers are at the end of the podcast. As disclosed, I have a small position in the stock we're going to talk about today, so people can keep that incentive in mind.

With that out of the way, Aaron, the company we want to talk about is Cogent. The ticker is CCOI, probably the most interesting stock in telecom outside of maybe the looming disaster of the corporate restructuring that is EchoStar/Dish. I'll just turn it over to you. What is Cogent, and why are all of my hedge fund friends so interested in it?

I didn't know all of your hedge fund friends were so interested in it. First of all, thank you for having me on again. I had a great time last time, and this one is even more complicated than Carvana, so it'll be fun.

It's easy to get drawn into the weeds on Cogent, so I'll try to keep it very high level. I'll probably fail, but I'll try. It's just an internet provider. The legacy business is an internet provider. They basically built a global internet network that reaches almost all the IP addresses in the world. This is legacy Cogent. We can talk about the new stuff later.

They connect into about 3,500 buildings, and they sell internet to corporate customers in skyscrapers and to wholesale customers in data centers. They've taken all these legacy assets that they bought out of distressed bankruptcies and similar situations in the early and mid-2000s and reassembled them into this narrow footprint of a single-purpose internet network. It's very scalable, very efficient, and very profitable at scale.

They were the first telecom company I encountered that admitted and leaned into the idea that they're a dumb pipe. Everyone else is running specialized services. They think their voice is differentiated, their video is differentiated, and these other products are differentiated. Dave Schaeffer, the founder and CEO, had the vision that nothing is differentiated. The internet is the lowest-cost network of all of them, and everything should ride over the top.

When he started the company in 1999, that was not consensus. It's now consensus, right? For the company that had that vision 26 years ago, back when everything was kind of a special-purpose network, that was pretty disruptive. It was a maverick mentality in this market. I think he used that to buy out all the competitors on the cheap, tear them apart, and reassemble them into something that was growth-oriented, scalable, and efficient—which, if you think about telecom, basically nothing describes telecom that way.

All the wireline companies have been suffering forever. Cogent has the highest rate of organic revenue growth in all of wireline. All the growth went to wireless just from subscriber growth, but once you achieve maturity in subscribers, what growth is there? If you have the right competitive dynamics, you can maybe squeeze out a little price, but look at what happened in residential broadband. You add another competitor, and suddenly the economics get impaired for everyone. Everyone is on an all-you-can-eat, unlimited plan. There's nothing left.

Dave had this radical view at the time that the internet was just going to take all the share over time, over decades. He went about assembling purpose-built assets just to address the growth parts of the market. He said, "The internet's going to win, so that's the growth part of the market. Let's build a network that's purpose-built just for that application."

In a nutshell, that's Cogent. Now they've tacked on Sprint's wireline business, which is the oldest fiber network in the United States. It was the first one that was built nationwide. He's deploying the same strategy that he did with the internet to this optical transport, or wavelength, market. It's the same playbook: taking a distressed asset from a distressed seller desperate to get it off their books, repurposing it, and reconstituting it into a growth story with purpose-built infrastructure to support that growth market.

At a high level, that's what Cogent is. Dave Schaeffer is the smartest guy in any room—a certified genius. You don't question his logic and his brainpower on this stuff. He's also ruthlessly parsimonious. He runs a very tight ship. If you visit the offices, they're Spartan. They drink water out of Styrofoam cups, and they drink bad coffee. That's all they offer.

That's perfect. Let's start with the legacy business real quick, just so people can get a feel for what they do.

It's funny because when you review the financials and the earnings releases, you're like, "Oh my God." Especially after the Sprint acquisition, which we'll obviously talk about: "These are the most complicated sets of financials I've seen." Maybe not in my lifetime, but probably this month or this quarter. They're pretty complex. The press release runs 20 pages, and there are lots of add-backs and other adjustments.

But then I was struck by something I read in an expert call to prepare for this, with a former manager. He was just like, "It's the simplest business you've ever seen. They sell basically 2 products," right?

My friend Nichetti [?] did an interview that struck me. He said, "Look, these are the most complete, hardest-to-replicate fiber assets in the country on the legacy side." Am I right that the legacy side is basically a ring of fiber around a metro area, connecting to office buildings? You've got a law firm that has 10 floors, and they're going to come in and give you fiber and say, "Here's 100 gigs. Here's perfect, pristine, super-low-latency fiber. You have nothing to worry about when we plug it in." Is that the right way to think about it?

They basically have hubs in every city, and then they run rings around the hub. I think it averages 3 or 4 buildings per ring, so they have a lot of built-in redundancy. If you talk about 3 buildings on a ring, maybe there's a cut at Building 1 going from the hub to Building 1. The other side of the ring can still provide service, so they have low latency, very high reliability, and healthy redundancy.

They architected the whole thing to be super-efficient operationally—to execute on and to maintain. One of the nuances that Dave picked up on early and leveraged for Cogent's advantage was that you don't need to own this fiber. They have 377 dark-fiber providers around the world that run the legacy Cogent network. Now that's going to evolve because they're going to put some of the Cogent network on the Sprint network.

Historically, they've leased the fiber from others because there was so much excess fiber from the telecom boom in the late 1990s and early 2000s. There's been a lot of competition in the metro area, so it's been relatively easy for them to build these rings outside their hubs that connect them. They have foundational long-haul fiber—historically leased, now owned—that connects each city to the others and to all the different hubs. That's how they assembled the network. They have long-haul fiber that connects the big cities and all the different hubs.

Then they run rings from the hubs, and from there they can connect to every other network, essentially. I think they reach 98% of IP addresses globally running that playbook. So it’s kind of cool: you can reach the entire internet, or offer the entire internet, from really just 3,400 buildings.

But if you’re talking about the data center business, it’s like 400 buildings. You can reach 98% of all global IP addresses from that pretty narrow footprint. If you think about Lumen or someone else, they connect to tens or hundreds of thousands of buildings, and they bring the same ubiquity of service and access to the global IP network, but they do it on a much, much bigger footprint, which is obviously much more expensive to maintain.

Sticking with the legacy business, when you talk about connecting a giant office tower or whatever, obviously, they’re not doing home broadband. But Comcast—I’ve talked a lot about cable on this podcast—Comcast, Charter, or, if you’re looking at the legacy telecom, I’m in New York City: Verizon FiOS. You can get one of the legacy telecom providers; if you’re a big building, they can connect to you, too.

So why do building owners choose to do this? Why does this exist? Why isn’t the cable company, the local cable company, just connecting all these and saying, “Look, we’ve already got all this fiber laid out to connect the homes. It’s just a little bit of an extension to go connect office buildings”? Why is Cogent so successful in capturing these large, enterprise-style customers?

Again, the strategy is to pre-build a purpose-built network and then scale it. The telecoms have ubiquity of service. The cable companies don’t have ubiquity of service: they may cover the residential neighborhood but not the downtown business districts all the way with 100% ubiquity.

The answer is that there’s always going to be an incumbent telecom in every skyscraper. There’s usually at least 1 other competitive provider; it could be a cable company. But there’s always going to be Cogent in the really big ones.

The really big ones average, I think, around 450,000 square feet and 50 floors. So they’re going for the cream of the crop. It’s 10 basis points of the building count but 11% of the space in corporate offices in the US. They’re really going for outpunching as much as they can on each building.

Cogent pre-wires every building through the elevator shaft, or the riser. That costs about $10,000 a floor to pre-wire. It’s strange because it doesn’t feel like this should have happened, but the telecom incumbents don’t pre-wire.

So if someone on floor 44 wants service and they’ve never connected to that suite before, they’ve got to run cables up that riser. It takes time, and it costs capital for them to do that. They just don’t have this kind of scalability; the network isn’t as scalable as what Cogent already has sitting in the ground.

When a customer on floor 44 wants service, Cogent says, “Great, we can give it to you.” They guarantee 2 weeks, but they average 9 days. If AT&T or Verizon receives the request, it might take them 90 days, and it might cost you some capital to wire that floor.

They might charge you—it’s almost a construction project.

Exactly. It’s not that they can’t do it; it’s just that Cogent has it already pre-built, and that’s what makes them disruptive.

It turns out that most of the time, you can buy services—you can buy a T3, which is, I think, 45 megabits per second of symmetrical, dedicated access—from AT&T or Verizon. They might charge you $1,000 a month for it. Cogent will charge you around $700 a month for 1 gigabit of symmetrical, dedicated service.

They often price at par but give you many multiples of the capacity.

I’ve also heard their networks are dedicated. AT&T might promise you, “Hey, you’re getting 100 megabits per second,” or something like that. That’s a great thing, but it’s on a shared network. So if floors 44 and 45 both start downloading 20 Netflix shows because they’ve got 50 employees or something, you might experience a lot of lag.

Whereas with Cogent, your strand is your strand. So you download, and you’re not going to have any latency.

That’s especially true for cable companies. Cable companies are HFC, right? They have fiber to the node and then coax to the last mile. All the coax that feeds into that node is shared infrastructure on that one strand of fiber, that fiber node.

So it’s not always a problem, but sometimes it can be at peak hours.

That’s great. I think we’ve done a great job explaining the legacy side. I’ve got so many questions to get to, but we’d be doing a disservice if we didn’t quickly discuss the big deal here: Cogent announced that they bought the legacy Sprint network from T-Mobile.

Why don’t we quickly give an overview of what they bought with the legacy Sprint network, how that really made the financials complex, and why that might be setting up the opportunity here?

You’re right. The financials are as messy as they can be. It’s partially, I would say, a communication issue, but we’ll get into that. The deal was to buy the Sprint wireline network from T-Mobile. This is a network that T-Mobile got as part of the Sprint merger.

How much did Cogent pay for the Sprint network that they bought from T-Mobile?

They paid $1 for the network, and they’re getting paid $700 million to take on the money-losing services business that was attached to that network.

The story there, as far as I can understand it—and I’ve heard a couple of different versions of this, so I’ll tell the one that I think is the most accurate—is that Sprint Wireless was using the Sprint wireline backbone to run the wireless business. When T-Mobile bought Sprint, they migrated everything off that network, and it took away the anchor customer from the network.

Then you have a very unprofitable set of external customers that were no longer subsidized by the 1 massive internal customer. That’s why it suddenly looked like there were huge losses, even though it was serving its purpose for Sprint and then, initially, for T-Mobile. You turn off the anchor customer, and suddenly you’re burdened with all of that bad revenue and too much opex against it.

T-Mobile was motivated to get this off the balance sheet because it wasn’t strategic to the wireless business. It’s tiny compared to the whole enterprise. They weren’t using it, and it was burning over $300 million of cash flow—more like $400 million or $500 million of cash flow per year—when they started this process.

They wanted to go with a partner who could basically not just shut it down. I think they were worried about the PR around the Sprint-T-Mobile merger. If you sell this thing off and then that buyer fires all the customers and all the employees, it’s kind of bad press for T-Mobile and maybe adds some scrutiny from a regulatory perspective.

If somebody’s going to buy it, fire all the employees, and fire all the customers, what are they going to do with the network? It’s not like they can cash-flow it. Maybe they’ve got alternative uses, but that’s a tough sell.

If you were selling it to Amazon or Microsoft, who wanted a network, they could repurpose the whole thing for themselves. We’ll get into that later, I’m sure. But that would be a logical thing to do if you were a strategic buyer that wanted the fiber—the long-haul fiber network—because there are so few long-haul networks out there, right?

Dave comes in and says, “Okay, we can restructure the customer base and the cost structure.” They had mapped out, as far as I understand, before they signed, day by day, every wholesale contract they had in the book on the cost side. They mapped out, for 3 or 4 years, when they could turn off these unprofitable, really expensive contracts that they had on the cost side.

They also knew where they overlapped with off-net providers. Let’s say it was a customer that Sprint was servicing with an off-net provider. Maybe they were using Verizon to service that customer, but Cogent was in that building. They could switch it from Verizon, which has a low margin, to Cogent, which has essentially a 95% margin.

They knew where they could switch from off-net services to on-net services, which immediately was massively accretive. That, I think, has gone pretty well. It’s really opaque and difficult to discern in the financials on a consolidated basis, but Dave is legendary at this stuff. I don’t worry about that at all. I think they’re making good progress there.

He’s got stories that I’ve heard over the years about how, in the 2000s, he’d bring people into an auditorium after he bought a company and say, “Welcome to Cogent. You’re all fired.” Or maybe he’d keep 1 or 2 people, and that was it. He’d just take the assets, repurpose them, and move on.

So, you know, he's doing—not as draconically this time around, because they need to retain some of this workforce and maintain the physical network. They're still going to retain a bunch of this business; it's not going to turn the revenues to zero. I think one of the challenges in reporting the financials is that he had communicated early on that this business was going to bleed off a little bit of revenue, but then stabilize in the $400 millions. I think it's going to stabilize $100 million-plus below where he had communicated, but he hasn't really said it publicly all the way.

If you put owner-oriented people in the same room and look at the contracts that he's been looking at, I think we would all agree that those revenues should go away because they're negative margin. That's why you've seen negative top-line growth at a consolidated level for several quarters in a row sequentially, but you see EBITDA going up. If you turn off $100 of revenue but turn off $200 of cost, you would do that all day long if it's a direct cost associated with that contract. I think there's a bunch of that going on.

This acquisition has 2 components. There's a services business, which is not interesting. It's kind of a necessary evil to swallow that and turn it from a very money-losing business into a slightly profitable business, but there's not a lot of juicy terminal value attached to that business that they acquired. The whole reason you take that on is to buy the network for $1 and repurpose it, invest in it, reconstitute it, rearchitect it, and turn it into a growth engine, kind of like what Cogent has done with the internet business over the last 20 or 25 years. I'll pause there, but we can jump right into that if you want.

No, that's great. So, look, at this point, hopefully listeners at least have an overview of the legacy Cogent business and the Sprint business. Maybe we haven't fully drilled into the long-haul fiber there, but at least they know what they bought for the Sprint business and what they paid. They paid $1, and they get $700 million in payments from T-Mobile. We didn't even talk about the $500 million in IPv4 addresses, all this sort of stuff.

But let's pause there. I want to ask my favorite question. We've done the overview; hopefully, people know what Cogent is. When you look at this company, yes, the financials are very complicated. The story with Sprint is moving, but this company is very well covered by the sell side. More than that, Dave, this CEO, really smashes the conference circuit. You can go on any transcript you can find, and he is fantastic at laying out the story. He tells you what's going on, he tells you the complications, and he tells you why we're doing it and what we think the long-term value is.

He's laid out $500 million of EBITDA in 2028, and the cash flow from this business is very good. So, you've got a CEO whose company is very well covered. There's complexity, but the CEO is laying it out very well. When I put all that together and say, "Cogent is spoon-feeding everyone this," what is the market missing that you think you're seeing that makes this a risk-adjusted alpha opportunity? Or what is the market just skeptical of that you're not as skeptical of?

This is a great question. I think Dave's communication has not been great about this. He's a great communicator, but he hasn't been great about communicating the play-by-play of how this would go, and he's missed a few pretty big milestones along the way. Dave's communication around this deal has been to map his experience in transit, the internet business, to what he expects for this wavelength business. That mapping has been incorrect the entire time, and he's known that. His mentality is basically, "It's the biggest data set I have. I don't have data on how this should go. I'm going to use my playbook from before and just assume we'll map to that." It's been wrong.

I've had a lot of conversations with him about this, and he's like, "Well, I don't have the data, so what do you want me to do?" I was like, "How about you don't give specifics that you're just going to miss?"

This is my view: the hardest part is behind us, but it's not yet visible to the market because it hasn't translated into the acceleration. The acceleration was communicated, or he expected it to show up, about a year ago. It didn't really happen because the network wasn't done yet. If you take an owner-oriented group of people and say, "Should we do all this custom engineering in a nonscalable way to get the revenues in earlier, or should we finish the scalable, purpose-built network so that once we're done with it, we can just fly with this business?" I think most people would say, "Take the 6 months, let's build it, make it scalable, and let's grow over the next 5 to 10 years." I think everyone would choose that.

The problem was that he communicated that they should be at around a $100 million run rate a year ago, which they missed badly. I mean, very badly. A miss rounds to a 100% miss. It's embarrassing, and it's because he thought—again, this is mapping transit to waves—that if they connected the first 50 data centers, the top 50 to 100 locations where they do most of their traffic in the transit business, to the wavelength network and did some preconfiguration, that would be enough to attract $100 million of run rate.

The reality is there are way too many permutations in the wavelength business. You have to do essentially all of it in order to make it scalable. I think they found that there's just too much custom engineering and bespoke work on a customer-by-customer basis to validate doing all that work and pulling resources away from building the scalable thing.

He was mapping it to the internet business, where if you just connect the top 50 to 100 locations, there's a strong 80/20 rule in that business. That did not apply. If you look retrospectively at the data, I think the number is that they have wavelengths going between 329 data centers, and it's still a small business. If a small business is serving 329 endpoints, you're way beyond the top 50 to 100. His mapping was incorrect.

His mapping on the second thing that he missed badly was, "We can install 500 a month. We'll have that capability, and it should turn on pretty quickly." Again, the mapping was to an internet mentality, where internet customers just plug in kind of anywhere. It's not that big of a deal, and they're used to transacting in a pretty quick modality.

In wavelengths, they're not used to that whatsoever. If you do customer channel checks with fiber engineers, network architects, and buyers of these products and services—you hear this from all of them. I've probably done 30, 40, or 50 of these calls over the years. The feedback is always the same. Nobody is consistent on their installation times. They say 6 months, and it's 9 months; they say 9 months, and it's 13 months. They don't even know where their network is.

You hear all the horror stories about Zayo. Someone asks for service in a data center where Zayo is, and Zayo says, "We don't have service there." The guy is FaceTiming with the Zayo sales rep, saying, "I'm looking at your rack. You have service here." That's what causes the delays, and we'll get into some of that later, maybe. They're used to horrible service where it's not turnkey.

When Cogent goes from 18 months of "We can't install you, and we can't promise you anything" to "You flip a switch, and we can install you within 30 days," that's all well and good, but nobody believes you, number 1. Number 2, nobody transacts in that market on that timeline because everyone is used to these significant delays. Again, he's mapping his expectations to the internet.

In both cases—the $100 million revenue ramp and the customer ramp—they're incorrect mappings because he's drawing on the only data set he's got. He knows it's been the wrong one the whole time, but he doesn't have anything better.

From my perspective, why do I think it's different? What's my view on everything? If you ignore his commentary and communication on all the things that map to the internet and think about it from first principles, we have a working wavelength network. It's novel. There's never been anything built like this in all of telecom. It analogizes most closely, for me, to Starlink launching a full constellation in one shot, and it's unloaded like that.

The risk is behind us because the network works. It's there, it's more ubiquitous, it's got tons of capacity, it installs faster, and it has a lot of different attributes that customers care about. This thing should play in the market. Forget about timing for a second, and it should work out. That's my view.

Let me forget about timing for a second and ask you: I believe the dominant player in this space is Lumen, right? I think calls have revealed that Lumen and Zayo are the dominant players, and I think Lumen has the overwhelming market share here. When you say they've launched the Starlink global network and launched it in one shot, what is the difference between what they offer and what Lumen offers? To my dumb-dumb understanding, it's like, "Hey, cool."

Maybe they can get new customers hooked up a lot faster, which is great and should let them take share of net new business. But it's fiber from D.C. to Seattle. How much of their network is Starlink versus the fiber that's already been laid out?

Yeah. The thing is, you have to remember that you can use Starlink; it's 1 player. You can also get connectivity from Viasat or EchoStar—allegedly. First of all, it's telecom, so there are competitive products and competitive capabilities, and they're relatively undifferentiated. I would say wavelengths are probably 1 of the more differentiated products in the market, so we'll get to that in a second.

Fundamentally, you have fiber connecting all these different facilities. It's theoretically possible, of course, and it happens all the time, for others to provide these services, too. They're not the only ones that can deliver these services. The difference is in the purpose-built nature of this network.

Everyone else has heterogeneous products and services running on top of a common fiber architecture. They have a mix of legacy products; they might still have voice, video, and other things running. Of course, they have modern stuff, too. But that mix of businesses, older hardware, newer hardware, older customers, newer customers, old cost structures, and new cost structures—it's a hodgepodge of all this different technology running on top of glass, all this glass distributed across the country into all these different facilities and between these different cities.

I think the difference here is that Cogent basically has none of the legacy stuff because there's essentially no business on this network. They put the modern, newest technology on it that's scalable and efficient. That alone is pretty nice to have, but it doesn't really mean anything competitively.

Okay, so great: you have the modern technology. Lumen also buys all the latest technology that exists for its network, so it's not necessarily differentiated.

I think what makes it differentiated is that they preconfigure enough of this network just for this application—for wavelengths—in a way that makes it much more scalable. How do you gain market share? What are the dimensions of value that customers care about? Price is the last one. I would say it's important, but it's the last one. First, do you bring diversity to my network?

These are nonexclusive connectivity contracts. This is a nuance that most people miss when I talk to other investors about Cogent. A lot of them default to the residential broadband analogy, where it's either Comcast or AT&T, and 1 of them wins that contract. That's not true in this market. They multisource everything for resiliency and diversity.

A lot of it might be a data center with $1 billion of NVIDIA GPUs in there, depreciating over, let's call it, 12 months. So that's $100 million of depreciation every month. If that data center is off for a month—or for a day—because a $25,000-a-year wavelength is offline, that's a disaster. They'll provision 4 of them if there are 4 different ones that can absolutely go down. That can add a 9 of resilience to these.

I love that quote.

So I think they'll win share just because of that. You don't need the purpose-built network in order to do that, but it makes it a much better business for Cogent because they can offer that 9 of resilience for essentially almost 0 incremental cost to Cogent. The free-cash-flow characteristics of Cogent's wavelength network are better than everyone else's because they can do it in a very capital-light way and in a very operationally light way, whereas everyone else is taking longer. They might take 6 months; they might have to send all these field technicians into the field and physically rearrange the network in a way to provision that service.

Cogent has it prewired. It's just like the office buildings: AT&T can pull fiber up there, and it might take them 90 days, and they can give you a gig of service, but Cogent can do it in 9 days because it's prewired. It's the same analogy over and over again for Cogent.

It's not technically that they're doing anything that's impossible for others to achieve. They're just doing it in a much better business model, and they're doing it in a way where it's possible that the time to install is competitively differentiating. I would say from my conversations with people who buy in this market that the consistency of delivery is much more important than the actual time.

This is a very technical sale. You're going from a specific building on a specific route, where there's a very specific latency, to another specific building. It's not plugging into the internet. Arranging all of that and having a customer get ready for that takes more time than 9 days or 17 days. Usually, they're not ready on Cogent's timeline. Cogent can do it, but I don't know if it matters so much as saying, "When we give you a date, we can deliver on that date as promised."

I talked to a former Zayo network operations guy, and he said the best statistic he ever saw for on-time delivery within 30 days was 50%. The industry average is more like 10% to 20%. So if Cogent's at 100%, that can be—never mind just adding a nine of resilience. If you can guarantee service within that time frame, that's potentially disruptive to the market.

As we said, Amazon might bring in $100 million of GPUs based on thinking they're going to have a connection from this data center to that data center. If you're delivering it 50% of the time, well, cool—even if it's 4 days late, $100 million of depreciation is a huge issue. If Cogent is guaranteeing 100% and they can trust that, then the network's just a lot more efficient.

So, let me go back. I know we said timing, and we've got a ton of other stuff to talk about, but I want to posit something to you. I think 1 of the reasons the opportunity might exist is that last summer, when a lot of my friends who got interested in the name were talking about all the stuff you were talking about, right? They said, "Hey, the Sprint integration stuff—we're getting long in Q2 of 2024, Q3 of 2024, because we're about to hit the switch, right? Sprint is about to go from a huge money loser to breakeven and eventually profitable. We're about to see the fruits of all the labor. We're hitting the switch."

Not that the stock price tells everything, but if you look at the Cogent stock price from the summer to the end of the year, it goes from $50 to $80, and then, as we're talking today, it's basically gone round-trip, right? It's $80 back to $50. I think the reason is the Sprint network hasn't flipped; it hasn't filled up for a lot of those reasons you're talking about. But I think it's also more than that.

The data center opportunity—they got a bunch of data centers in the Sprint sale. People were thinking, "Hey, there's like 130 megawatts of data center power in there." They were thinking, "Hey, look at what's happening with AI. Look at what CoreWeave is saying: there might be $1 billion of value here. We might see that by the end of the year."

I think Dave was even pushing people in July. He went to a conference and said, "We're going to sell this thing." You and I are sitting here in June 2025. They say they've got an LOI, but we still haven't seen any movement. So I think people are just starting to question not just the timing, but the fact that they haven't really delivered any of the stuff we're talking about. I think the story is just getting hairier and hairier. What would you say to that?

I think it's completely fair. He's been late on everything. I think back in August of last year, they communicated on the Q2 call that they were going to invest $100 million of capex into this network and into the data centers and get them ready.

I think he had hoped, honestly, that he could sell a really crappy shell—a house that basically had plans but nothing else: no kitchen, no cabinets, no flooring, nothing—and get top dollar for it. I think he realized that the price for that was significantly lower than what he thought. If you invested the $100 million, there would be a pretty easy payback. I think we're kind of nearing completion of that $100 million investment.

It's kind of a weird process if you think about the sales process. They're offering it; they're basically saying, "We'll accept any kind of offer you want to bring us. If you want the whole portfolio, we're interested. If you want a partial portfolio, we're interested. If you want a single facility, we're interested."

If you run that process and know that not everything is ready yet, and some are at varying degrees of completeness, it's kind of odd. It's not really an auction. How do you compare a whole portfolio at $5 million a megawatt to 2 partial portfolios, somewhat overlapping, that might be at $8 million a megawatt? How do you sort through these different offers? I think they're just going through that process.

The thing that I worry about—I talked to Dave about this a couple of weeks ago—is that Dave is so focused on maximizing value. I told him, "I'm worried that you might have an offer for $500 million for the whole portfolio, and you might hold out for 18 months to get $550 million." That's not a great use.

That's not a great setup for equity guys. He's allowed the balance sheet to get overlevered, which we can talk about. He's put in capital; there are both opex and capex and balance-sheet issues explaining why leverage looks extra bad right now. I call it levered but overlevered, but not risky, because he could, in the end, probably take one of the low bids out there pretty easily and raise at least $300 million to $500 million—not overnight, but pretty close.

I think he had a thought that the AI market was so hot—it was like a shark-feeding frenzy—but not for those facilities. These are not AI compute centers, right? These are edge facilities. Most of them are around 5 megawatts; a few are bigger than that. They're not going to be filled up with tons of GPUs and run for AI compute.

I think there are plenty of operators in the market that have an appetite for these kinds of facilities, especially because they often are adjacent to or closely situated to some of the bigger compute centers out there. I think the whole data center market has been Ashburn and Silicon Valley, these Tier 1 markets, when it was the telecom footprint. I think the AI footprint is going to be very different because it's going to have to migrate to places with much more space and power availability at much lower cost.

There's some overlap there, but I think Cogent's footprint is actually okay for being adjacent to these new compute centers that are in historically secondary and tertiary markets for the data center market, in terms of data center topology. For instance, I've heard that the Cheyenne facility has, I think, 2.5 megawatts, but it's apparently one of the more in-demand facilities because there are big proprietary facilities by Microsoft and Meta very close by. There are a few of those examples in the portfolio, but I think it's kind of just an awkward process.

I think he thought they could sell into that momentum last year, but people wanted things closer to turnkey, not where they had to go spend a year retrofitting them to what they needed. So I think Dave recognized that: “Okay, maybe we can get $250 million from these as is, but maybe if we spend $100 million, we can get $700 million. So we should go do this.” I think that's kind of how I think about it for myself.

Let's talk about Dave real quick. If I've done 3 expert calls on this, you've done 300. The consistent thing I hear from people when you talk about Cogent is, “Dave is Cogent.” This is his baby. He's got, what, 40 direct reports? It borders on micromanagement.

Everyone I hear says, “Look, there's one thing you need to know: Dave is the smartest man in any room he's in.” He's a fantastic entrepreneur. You can talk about the taxi wherever you want to go. On one side, you've got this literal genius who has a PhD at, what, 15 or 17—I can't remember—and graduated high school at 12, if I remember correctly. He's a literal genius and the best entrepreneur a lot of people have ever worked with, running this company.

On the other hand, I would say, hey, the history—not just of telecom, but in general—of the smartest man in the room with a lot of leverage does not end well. That would apply in 2 ways here, right? It applies to Cogent, as we mentioned: I think it's overlevered on the stats because of the Sprint, but I don't think it's necessarily risky. But it's also overlevered on a personal basis, and we can talk about that.

The history is: leverage plus the smartest man in the room does not end well. It ends in tears. I would point to Dish, Charlie Ergen at EchoStar/Dish—that has not gone well—and plenty of other examples. That's just the telecom one. So I hear these 2 dynamics, and then I see that you've got a CEO who, for the past 12 months, has stepped on a rake multiple times and just missed and dismissed.

My biggest worry when I see this is, this is the Dave show. When I see smartest man in the room plus leverage, I actually get nervous. I'd love to just talk about all aspects of Dave.

Yeah, I think you summarized it well. I've known Dave for 12 or 13 years now, and I've talked to him I don't know how many times over those years. He's brilliant. He's obviously brilliant. I think he's extremely insightful, and you can see that in Cogent's organic revenue over the last 20 years compared with everyone else in this market.

He's a little bit dumb for choosing telecom to be his playground.

I had that thought. I was thinking, man, if that guy had just gone into SaaS in 2002 or something, why didn't he choose that?

I think part of the reason telecom appealed to Dave is that he's competing against this sunk-cost base of legacy providers that never really acknowledge the reality of their situation. He could use his talents disproportionately to expose and exploit them over time. I think he loves competing against these old legacy companies that can't do much because they have a cost structure and a capital base that were supporting prices that were true 20 years ago but are now 1% of what they were 20 years ago.

I think that's why he's chosen that playground for himself and stayed, because he really loves disrupting it in slow motion. I asked him a couple of years ago, in an interview that I published on my website, how this had gone relative to expectations. He thought that if he offered the best service at the lowest price, he would get 25% market share within a couple of years, but it took a lot longer than that. It took decades.

He thought he could disrupt a lot faster, but there's a lot of inertia among these customers. I think it's been not so slow that he would move on from it, but not fast enough to be as disruptive as his hypothesis would have indicated. I see this a lot with some of these entrepreneurs and owner-operators: they're naturally more aggressive than most people are. I think Dave is a pretty aggressive guy, but I think he also has a lot of outs. He gives himself a lot of outs with Cogent.

What's funny is that he's kind of a forced seller because of his commercial real estate portfolio. Let's discuss that, because I think the first question anyone asks is: you say Cogent, they pull up the insider transactions, and they see the smartest guy in the room has been a consistent net seller of shares for the past 18 months. Do you want to discuss what's happening there?

Yeah. Silicon Valley Bank collapsing was really the catalyst that forced him to start selling, because all the regional banks started pulling out of commercial lending or rebalancing their books.

The way he's described it to me is that he had a portfolio worth about $1.1 billion. It's probably worth about $600 million now. He had, I think, around 50% LTV on it at the time, and now it's, I want to say, around 65%.

So he's using his stock sales. First of all, he takes no salary; he just gets stock and dividends. He has around 4 million shares and gets about $16 million in dividends, mostly return of capital, if not exclusively return of capital. So he's deferring his taxes personally on that a lot.

But when he sells, the way he tells it—and I don't want to air all his dirty laundry—is that the banks are kind of forcing him to pay down the loans' principal on an accelerated basis, or else they may force him to do an LTV test, which would trip a covenant and then he would lose the assets, essentially. He's paying these off as slowly as he can with stock sales, but when he sells stock and he has a negative basis in it, he has a big tax obligation, too. So it's kind of picked up the pace, especially in the last month or 2. It looks really bad.

I totally agree. What's funny about it to me is that no one is more motivated than Dave to monetize the data centers and get the leverage question off his plate. If he brought in $500 million—even if he thought he could get $1 billion—and got $500 million of proceeds from a data center sale, it would probably solve a lot of, or ease a lot of, the concerns about the dividend strategy, the leverage ratio, all of that.

It would probably send the stock higher, right? Which means he would not have to sell as many shares to accomplish what he needs. No one is more motivated than a forced seller to get the stock higher, and yet he seems to be still optimizing for the long-term value. Or you could look at it the other way and say, well, maybe not. Maybe he is optimizing for the short term and it's still failing. So that's the worry.

Totally valid. Totally valid. But knowing Dave, I would question that logic or that conclusion.

I understand the logic. It makes total sense. I don't know if I agree with the conclusion.

I completely understand. And look, that's the tough thing. But again, the biggest risk to me is the smartest man in the room with leverage at Cogent. I don't think the leverage at Cogent is too bad, though.

I want to talk about capital allocation, but leverage on an office portfolio resulting in foreclosure—I see that and I'm like, “Oh, yeah. That is the story 9 times out of 10. It doesn't end too well.” But I don't want to—I think we've discussed that enough. Dave actually talks about it a little on the calls, too. You can see how he ended the Q2 call.

Let me ask one more question, though. This does not relate, per se, to Dave's leverage, but to the capital allocation, right? You have a company here that is quite levered. I think the headline is 6.6× leveraged. I think that will peak in Q3 and then come down, maybe sooner if they get an asset sale done.

They're paying out a large dividend. The dividend yield, as we speak, is roughly 10%—it rounds to a dollar per share per quarter on a $48 stock. They've grown it 54 quarters in a row or something crazy. But I guess I would ask you: Does the dividend strategy make sense? You have an owner-operator who still owns almost 10% of the company.

They bought back a little bit of stock after Q1. Would it make more sense to just shut the dividend off, delever, use the excess cash flow to buy back stock on the cheap, and then, when this works—this could really work? I would also sprinkle in a little bit of, “Hey, is the capital allocation and dividend strategy being run for Dave's cash flow needs, not for what's best for the company?”

Yeah, it's a good question. I think whether it's a levered buyback or a levered dividend strategy, they both require the John Malone playbook. It works until it doesn't, and then it's a really bad thing. I've learned that the very hard way several times.

I would say Dave's twist on the John Malone playbook was that he was doing return-of-capital dividends, so deferring the taxes—not necessarily adding to the buying power in the stock in the market to boost it higher, but returning the capital anyway in a tax-efficient way to shareholders.

I think fundamental to the view of them continuing on this policy is the belief that revenue acceleration is around the corner and the cash flow growth of the business is going to support that. He says it's not in their business plan to monetize these data centers, but obviously the board's not going to greenlight a $100 million capex program for something that's not in your business plan. It's obviously in the business plan to monetize these; otherwise, you wouldn't have spent the capital, right? You would have just taken the price a year ago at whatever the number was, even though it was lower than they wanted.

I don't think Dave makes bad capital allocation decisions on that front. He doesn't speculate with that much capital—$100 million of data center retrofits. So the way I think about it is, I think we're sitting at peakish leverage.

We're getting hit because of the capex they've had to absorb for the Sprint network, which isn't yet monetizing the wavelengths to the level it should—and will—as well as for the data center retrofits. On the other side, you've had to absorb all this negative opex from Sprint Core, and then you have this $50 million to $60 million opex item that's going to bleed down through the end of 2026.

I wrote about this on my website. He's disclosed it 1 time on a public call, but I've talked to him about it over time since it's shown up because it's really hard to make the numbers reconcile if you don't have that in your model. So I think we're getting hit on higher net debt than they should have, which should normalize, whether through data center sales or cash flow going forward. Then you're getting hit on opex, so you have an elevated numerator and a depressed denominator affecting that net leverage.

If you fast-forward a year or even 6 months and they haven't sold anything, wavelengths aren't accelerating, I would expect the board and Dave to revisit the dividend strategy. I doubt that they would cut it—sorry, eliminate the dividend, suspend the dividend. I think they would cut it to a lower level for some period of time to get it under control, bring more operating cash flow into the business, and delever like you talked about.

I think he's pretty married to the dividend-growth path because it's something he's promised people, and I think he really values that consistency of so many quarters or years in a row—54 quarters in a row of dividend growth. He knows a lot of investors depend on that and expect that, and a deviation from that would be pretty damaging to his credibility and reputation.

I've talked to him in the past about this, and he's said he's been willing, if the opportunity presented itself, to suspend the dividend temporarily and just plow everything into buybacks if the opportunity was there. I was like, “Dave, you're borrowing at 6.5%, and your dividend yield is 8.5%. It's crazy when that's true. You should do this; this is when you should be buying back stock, right? You're taking out a lot of dividend liability if you buy back the stock today at these kinds of valuations.”

He gets it, but it only happens when people have a concern about the balance sheet. So if he were to buy back stock—they just did a refi, right?—$600 million of new notes to refinance $500 million of 2026 notes. If he took the $100 million or $80 million of excess proceeds or something and did a buyback with it, he would just lever the balance sheet even more.

Do you want to do that optically when you're already at this kind of leverage, even if it's very accretive in the medium term? I would say probably not if it were me. But if Dave knows the wavelengths are ramping and they're going to be exiting this year with a $100 million run rate, then it would be a great trade.

So, just on Dave: Dave is about 70, I think. As I mentioned earlier, whenever you talk to anyone associated with the company, Dave is Cogent, right? I've heard people say it borders on micromanaging, which generally means micromanaging, and I don't mean that in a bad way. There are plenty of fantastic companies run by founder-entrepreneurs who are absolute micromanagers. But he's 70—what is the endgame here? The titans aren't living forever, but there has to be an endgame. What is the post-Dave endgame?

Is it, “Hey, 4 years from now, this company looks great to an infrastructure private-equity firm, and they sell to private equity. Dave maybe stays on for another 2 years, gets some PSUs vested, and then moves on”? Or how do you see this company post-Dave evolving?

I think he'll sell the company before he leaves, and when he's out, he's out. He's not going to stay on and run it for someone else. So I think the endgame is that he will sell, and there are so many digital infrastructure players out there.

This is a pretty unique asset. In the hands of someone else, it probably would need to be split up. I would split the wavelength business from the internet business.

Oh, interesting. They've been arguing that there are synergies between the 2 sides, but you think they'd be easily split?

I think the fiber network, if it were to trade on its own—let's say Cogent Infrastructure, which is how they structured this deal—I think that probably has very high strategic value to, I don't know, anyone in tech, not really any of the other telecoms. It has high strategic value to the tech community because it's one of the 5 or 6 long-haul networks that's out there.

Yep.

You can see Microsoft—or Amazon—saying, “We're going to control our own destiny. We own this network. We never have to rely on someone else saying, ‘Hey, we'd like you to connect Seattle to Chicago,’ and then they say, ‘We don't want to.’ Amazon owns it. Amazon does it.”

Right. Right. Exactly.

So I think that physical network has strategic value. I think it would probably be worth $5 billion plus on its own. The way it's been reconstituted, I think nobody wanted to take it on with the Sprint business.

Of course, Dave really had to disentangle this really crappy company. It was a mess, and it's like disentangling it, reconstituting it, and making it into something that's actually attractive. If it's attractive to CCOI and it works, which it does, it'll be attractive to others, which is why I think the wavelength business is going to start scaling. Timing could be anything, but I think we'll know before the end of the year how it's going for real.

The internet stuff is more of a niche business. These are smaller TAMs. The wavelength TAM is multiples larger than anything else they've been operating in. It's narrow, US-only or North America-only. It's physical infrastructure that's pretty scarce, and I think that has really nice value on its own, on a standalone basis.

The corporate, cream-skimming internet business and the corporate skyscraper business—maybe it's a private-equity-owned business that could cash flow, but will it ever stand independently? I don't know. The transit business probably couldn't be rolled into one of the other big guys.

But maybe that has strategic relevance, importance, and value to a tech company, too, because Cogent is a tier-one internet company. That means they get to peer with everyone settlement-free, so all of their internet traffic is free globally. If Amazon got that, or Microsoft, Google, whoever, you could justify it. I’m sure they pay a lot in internet costs, so if they get to trade traffic settlement-free with everyone globally, maybe that’s interesting to them. Plus, they get some customer traffic on that, too.

All of their businesses are Moore’s law businesses, right? The cost to deliver a bit of data over the internet has historically gone down with Moore’s law—30%. Now, this has always been countered by the fact that data demand goes up, right? I’m never going to bet against Americans’ ability to watch more TikTok videos, but video is pretty well penetrated. AI hasn’t really resulted in a spike in data usage yet, although that might change with inference. I’d love to talk about CCOI as an AI beneficiary as well.

One of the things that my really smart friends who do what you do—concentrated investing and deep research—say is, “Hey, outside of the messiness, this is the thing that’s really hard for me: You have a Moore’s law business, so it’s always deflationary, and I worry about what happens if that data usage just doesn’t always go up.” What if one day you go to someone and say, “You’re paying us $1,000 per month. Last year, you were getting 10 gigs. We’re upgrading you for free to 25, but we’re going to keep charging you $1,000 per month”? What happens when you go to someone and they say, “Actually, don’t bring me from 25 to 50. Keep me at 25, and I’m going to pay you $500 per month because I don’t have any more data usage”? How do you think about that deflationary risk, if I explained it well and all of that makes sense?

I think, depending on the end market, there’s a different answer. First of all, Cogent has been the one leading a lot of this price deflation in the market. They’ve been the price leader by far, and they have a standing offer to undercut anyone by 50% on the wholesale business. The contracts work in a way where there’s a volume toggle. There’s a volume and price toggle, so if the volume doesn’t show up, the prices don’t deflate. The price deflation you’re talking about is reliant on volume growth being really healthy. When it doesn’t show up, the price compression is a lot more moderate.

I actually just talked about this with Dave last week because they’ve had 2 quarters in a row of sequentially flat traffic growth, which hasn’t happened, like, ever. They’re at 8% or 9% year-over-year growth in traffic, which is as low as it’s ever been. I think they may have had 1 quarter after Megaupload, but when I looked at my numbers, it was 9% year-over-year. In the worst quarter ever, with the most disruptive customer turned off, they still grew 9%, and that’s kind of where they are now.

I think we’re in this maturation phase of streaming video. There’s a lot of mobile video consumption, which means your bit rates are constrained. You’re not going from 4K to 8K TVs en masse, which means all of this IP traffic has historically been driven by human attention—whether it’s consuming text content originally, then pictures, then video, then higher-quality video, HD, 4K, whatever. When you’re on a mobile device with a small screen, you don’t need to keep upgrading the video quality because you can’t see the difference anymore.

It becomes linear. You had a multiplicative effect of having more minutes of use per day times a higher bit rate, or an increasing bit rate. I think that story has, to a large degree, played out. If you believe that there’s metaverse-type stuff on the way—VR, AR—you get a pretty big multiplier effect.

I’m a believer that as we go into this world of AI, there’s going to be a lot more machine-to-machine chatter, where machines exchange information without a human involved. It’s not a minutes-of-use-times-bit-rate-per-minute equation anymore. It’s going to be more completely disassociated from the human, or at least largely disassociated from human engagement. If we get into that world, then I think there can be much bigger long-term volume growth in that business.

I think there’s 1 layer of protection in place, which is that the volume toggles in the contracts prevent all this massive price deflation if the volume growth doesn’t show up. That’s in the wholesale business. In the internet provider business, with the skyscrapers, I don’t think anyone is motivated to cut price, and I think pricing there has been relatively stable. Cogent has been offering $700 a month for essentially all of time. They’ve upgraded the circuit size, but I think those will be manageable curves. That’s more of a utility, an all-you-can-eat service.

Can I just come back to the wholesale real quick? You mentioned Cogent has a standing offer to undercut anyone by 50% on price. I always thought this related to Moore’s law, but it was an interesting risk factor to me. That works when you’re the small upstart competing against AT&T, Zayo, or whoever it is, right?

These are fixed-cost businesses, right? You lay the fiber, and then there’s basically no other cost as you throw people on it. Eventually, if you take enough share, you’re going to provoke a competitive response because AT&T is going to say, “We have no one left on our fiber. We have to start selling fiber basically for free just to try and start filling this thing up.” How have they been able to sustain that 50% offer, and do you worry about a competitive response, particularly if data usage, as we just talked about, really starts flattening out?

Again, this is a multi-source product. Nobody single-sources their transit provider in the wholesale market, so there’s some amount of inherent balance in that market because of that dynamic.

I think the way it works is that they already have all the customers. The logos are already there. There are some new logos that pop up here and there, but they can still win new business from Amazon by saying, “Let us help you on these paths through the internet. We’re lower share, so if you give us more share, we’ll undercut the price you’re paying to AT&T, Lumen, Arelion, whoever it is, by 50%.”

On the growth side, you’re right: When you’re small, it’s disruptive. What I’ve heard is that the offer is still out there, but it’s being used less and less in the market. When they’re at 25% market share, they’re starting to emulate the overall pie more and more, and there’s less reason to be that disruptive. I think it’s a great strategy to go from 0% to 25%. It’s probably not the strategy you need to go from 25% to 50%, if they can get there.

I think they are starting to be less drastic on the price aggression in that business. I’ve picked up on that a little bit over the last couple of years, especially as traffic has started to slow down.

One of the cool things they’ve done is that they analyze what’s happening on the internet network. They don’t know with perfect vision, but they can analyze traffic in such a way that they can say, “Netflix is sending a movie from Los Angeles to Paris, and the way it’s routed today is Netflix to AT&T. AT&T hands it off to Cogent for free because it’s a settlement-free peer. Cogent takes it transatlantic and terminates in Paris. Orange takes it from Cogent to the end customer.” Let’s say that’s the workflow.

Cogent will say, “We see that there’s traffic going between Los Angeles and Paris. Orange is a customer because they’re on a tier-one network. AT&T is not a paying customer. Why don’t we go to Netflix and Orange and say, ‘You guys are both paying for this traffic. We’ll cut your price on both sides’?” Let’s say the customers are paying $200 in total for that, just as rough numbers. Cogent is making $100 of it, and AT&T is making $100 of it. Cogent will say, “Why don’t we structure this so we can make $150, and you guys both pay $75 instead of $100 on each side?”

They’ve been doing more of that, which is a clever way to steal share. It is cutting price, but it’s very accretive to Cogent. If you look at 5 years ago, about half their traffic was originated and terminated with customers—a paying customer to a paying customer, using Cogent in between. Now it’s 75% to 80%, because they’ve gone in and done some of that data gathering around the traffic patterns of the network. They can offer up pricing but also take more revenue.

When they’re cutting AT&T out, would Cogent then go build that last 100 feet or so to the Netflix facility to cut them out, or is there a different—

No, they’re already there.

They’re already there. Okay, they’re already there, so they’re just—gotcha, gotcha. Let me ask one last question, and then, again, my notes: I’ve got so many notes, and we’ve already been super generous with our time. I’m kind of like Charlie Day smoking the cigarette in that famous meme. I’ve got so many notes, but I want to ask one last question before I just let you talk about anything.

One thing I think investors might say is, “Hey, Andrew, this is great. It’s really interesting and founder-led, but you’ve got a hairy story with a lot of debt.” Somebody might say, “Is this really worth the brain damage?” I think they’ve said 2028 EBITDA will be about $500 million, and there’ll hopefully be cash flow between here and then. But $500 million of EBITDA in 2028—as you and I are speaking, the enterprise value is, let’s just make it easy, $400 million: $2 billion of debt and $2 billion of market cap, right?

Somebody might say, “Hey, we’re playing for 2028. That’s 8 times EBITDA in 2028.” Is there a lot of upside here? These guys are talking a lot about fundamental research and taking a big position. Is there that much upside when you’re talking about a telecom player? AT&T and Verizon trade for 7 times EBITDA, 7.5 times right now. The same with the cable players—the cable players are trading at 8 times 3-year-out EBITDA. Is there that much upside here?

Yeah, it’s kind of a question of what growth looks like between here and there. Obviously, if they get there, there’s a lot of fundamental growth. What does it look like when you’re at that launching point? I think what I would say is, if you believe in the fundamentals of this network—if you believe that the Wave network works, that it’s scalable, operationally scalable, and financially scalable in the way that it should be, in the way that it’s been communicated, and in the way that I believe it actually is already showing, it’s just not showing enough to really move the needle on the numbers yet—I think there’s no reason that the revenue should stop there. There’s no reason that it should stop there.

There’s also the top-down versus bottom-up math. The $500 million is a very top-down number that Dave has given. The bottom-up math would suggest much higher than that. A lot of people I talk to—other investors—are penciling out Cogent achieving over $10 a share of free cash flow. In the context of an under-$50 stock, that’s more interesting.

You get to collect a lot of dividends between here and there. If you look at it on a total-return basis, and you believe that they won’t cut the dividend, you’ll collect all those dividends. They’re going to pay out over 100% of free cash flow when they get to $10 a share of free cash flow, because that’s what Dave does. Are they paying a $12 dividend? Are you trading at something like a 25% dividend yield on an organic-growth story in telecom with high recurring revenues, high incremental margins, and all the reasons to believe that it should continue scaling from there? That’s how I would articulate the upside. It’s not hard to get to big numbers when you look at it that way.

The question is, how do you look at the downside? If we’re wrong, what does that look like? I think it’s plenty cheap if you look at it on a free-cash-flow-per-share basis and they achieve what he’s laid out on a bottom-up basis.

Let me just push back on that slightly. I think the only real pushback I have is that this is 15% debt to equity, right? If we’re using the numbers, a lot of that upside is the fundamental growth. Again, I think the 2 pushbacks would be that a lot of that upside is them going from about $350 million in trailing EBITDA to $500 million, and people might say, “Hey, they’re over a year behind. They’ve kind of stalled out if you’re just looking at the headline number.”

A lot of that is T-Mobile payments going away, but they thought they’d be further along. They would admit that they thought they’d be further along. So I’m betting on an organic-growth story, and that’s in question based on the past year to 18 months. A lot of that free-cash-flow number that Aaron’s throwing out is because this is quite leveraged.

You get the organic growth, which I’m questioning, and you’re getting a decent bit of leverage on top of that. So, yes, the free-cash-flow-to-equity story looks great, but when you start doing it on an EV basis and factoring in that risk to organic growth, it doesn’t quite look the same. You really need things to work out well there. I guess that would be the one pushback.

I understand. Yeah, it’s leverage. I don’t disagree with that. But leverage cuts both ways, and they’re kind of hurting from it right now.

I think the interesting thing about Cogent is that it’s basically just a call on whether this business is going to work or not, right? If it doesn’t work, there are reasons to believe there’s some downside protection from the 4-asset portfolio, whether it’s IP addresses, the data centers, or the fiber network. If they can’t build a Wave business and they sell the fiber network to Amazon or Microsoft, or just spin off the long-haul network and say, “It turns out we don’t know how to address this market properly ourselves, so we’re going to put it on the auction block,” let’s say it’s worth several billion dollars.

First of all, the path between here and there would be really painful. You could say the fundamental value would give you plenty of downside support, but the actual trading dynamics would be pretty terrible, because it would be an admission that the growth doesn’t show up for Cogent.

A Cogent dividend call?

Yeah, yeah. Then you’re wondering, “Okay, if they can’t make it work, who thinks there’s value here?” It may be that there’s very limited downside, except you will eat a lot of downside as an owner of the stock while that happens. So I’m well aware of that.

I think reasonable people can disagree at this juncture, because the stock right now is essentially a levered call on whether this Wave network is real and whether it’s going to attract customer demand. If so, the leverage will work in your favor, and that free-cash-flow-per-share number will probably be the driver. If it doesn’t, then that leverage becomes a bigger anchor on the stock and a big problem for the stock.

It’s funny, because the tracks are pretty well worn in these end markets. It’s not like you’re having to reinvent the wheel here. They reinvented the network strategy. This is a novel network strategy that they’re putting into the market.

But the wavelength market is large. It’s growing. It’s one of the healthier things in all of telecom—one of the growthier things in all of telecom. There’s plenty of demand for it. I think they would pick up 15% or 20% share just to add a nine of resilience for a lot of customers. They could just do the same thing that Lumen and Zayo are doing, but add a different path, and they would probably pick up a decent amount of share.

The fact that they’re doing these other things—90% route uniqueness, faster installation times that are consistent and reliable, and lower prices—all of those things will help them get there faster and extend them far beyond what the steady-state market share would be for just another layer of resilience.

We’re sitting in a place where the Wave opportunity has not developed as advertised by Dave and not as communicated by Dave. It’s taken longer. Everything has taken longer. Everything, I would say, except the Sprint synergies has taken longer.

The data center sales have taken longer. The Wave revenues have taken longer. The Wave network did not take longer. I think we are there. They hit the things that were most in their control.

I was about to say that. Yep.

The things in their control, they hit. The things outside of their control have been slower. Maybe they should have known that, because things in telecom are slow. But what they could control, they basically executed on plan, it seems to me.

Yeah. Last question, and then I’d love to just get your thoughts. You mentioned the fiber network. If they said, “We’re not the ones to run it,” they could put it up for sale. Have there been any other recent transactions selling wholesale fiber networks? I mean, you have the Zayo-Crown Castle thing, but that’s not a direct comp to this.

There’s only—I mean, the long-haul networks in the country are Zayo, Lumen, AT&T, Verizon, and then you have Cogent. So there are 5 of them. There are regional ones that can kind of play for a lot of the country or a big chunk of the country. You have Windstream, Frontier, and Crown Castle—or now Crown Castle is going to be part of Zayo, so not anymore.

But we haven’t seen anything quite like this come to the market, because they’ve all kind of been rolled up into those big companies.

Perfect. Look, you have done such great work, and I’d love to— you’ve been super generous with your time. I’d love just to get your last thoughts. Again, I’m like Charlie Day with notes all over. You have done such great work, and I’ll link to some of them. You’ve got these super-extensive notes on Cogent on the website. My favorite would probably be the most recent one, “Investor FAQ,” back in April, but there are tons of them.

Anything that I should have been asking or thinking about, or that listeners should be thinking about, that we didn’t hit so far?

No, I think you’re spot-on with the questions. I think these are all the right questions to be asking.

I think the biggest challenge for everyone is: Why is everything late? It’s the most valid question there is. The thing that has really blown me away, if I take a big step back and set aside my annoyance with the timing and the path of this—which I’m very annoyed by—is how excruciating it’s been to live through Dave’s bad communication on this. He’s such a great communicator, so the optics are just really messy. I hate every earnings release because it’s really impossible to discern what’s actually going on.

It’s hard to pick apart the numbers. Like you said, it requires a lot of brain damage. At the end of the day, every single story has to come down to this: every single good equity story has to revolve around some nice growth story. There has to be something there. I’m not a sum-of-the-parts guy. I’m not a deep-value investor looking for cigar butts. I wouldn’t be interested in that. If this is just a matter of selling off IP addresses and selling off data centers and making 20%, 30%, 40%, I would not be interested whatsoever. I’m here for the growth story. There’s no question about it.

From my perspective, what they did last year was insane. The biggest risk to Cogent was what they did last year with the network. They popped so many data centers last year. When I’ve talked to people about Windstream, Frontier, and other wholesale wavelength providers, they say, “In a good year, we’ll do 20 data centers. We’ll add another 20 data centers to the network.” Cogent did something like 700. It’s just built differently.

There are all these ways you can articulate why, but the way they build scalable networks is just different from everyone else, which is why I analogize it to Starlink. It’s built with a completely different architecture in mind. It’s launching ubiquitous service from day 1, which is pretty crazy, and it works. There are big customers. They can turn on 10 terabytes of capacity between point A and point Z within 30 days, on a timeline and with a consistency that nobody else can even touch. The hyperscalers are buying it.

The rumor was—and I think it’s confirmed at this point—that Amazon bought a lot of capacity between 2 data centers. The most differentiated route in Cogent’s network is Seattle to Chicago. Everyone, if you talk to Cogent and do a customer check, wants that route because that northern east-west crossing is very underserved and kind of the only game in town when it comes to that path with that latency.

Amazon has bought out an entire line of fiber from them already on that route. Who knows if there’s more to come, but that was their first order: to buy out an entire route of capacity, an entire strand of fiber capacity. The story I heard about that was that Amazon tested every single circuit for latency, round-trip time, all of that—425 of them or something—and they passed with flying colors. Their KPIs and SLAs are better; they’re hitting all their SLAs or better.

If the network is performing and has been rigorously tested by the biggest of the biggest and best in this business, the network risk is gone. It’s behind us. So now it’s an operational risk. It’s a sales-execution risk, and Cogent’s sales force—you know, there are a lot of stories about 6% or 7% monthly churn in the sales force—but the guys selling to these customers are not those guys. They’re not the high-churn guys.

These are the guys who have been at Cogent for 10, 15, or 20 years. If you look at a lot of the senior people at Cogent and the biggest sales leaders, these are very seasoned people. They know these customers really well. They can compete against the biggest and the best of the other companies as well. These are the strategic-account, premier-account sales reps—not the guys who are churning through calling every law firm in New York trying to sell internet service.

So I think the biggest fundamental risk, to me, was that the network was not going to work as designed because it was novel. Now we kind of know that it works. That’s pretty cool. The question is: What can you do with it? Can this Cogent team go monetize and execute on it the way it deserves, given what it is?

I hear the feedback from the other guys that they’re freaking out that Cogent has this, because if you’re Zayo and Cogent can turn on 20 terabits in 30 days while you can do it in 9 months, what’s going to happen to your business? It’s not just what’s going to happen to the growth of that business; they can also steal your share pretty easily, right? It’s not going to be a big, heavy lift for anyone to steal your share. So I think it could be disruptive to incumbent revenues and disruptive to growth at Zayo.

Zayo, which—if memory serves—I just pulled up their proxy, got bought out for 15 times EBITDA or something. So you’ve got a company that might be—I don’t want to say dead because they’ve got a ton of physical assets in the ground—not dead, but a company that’s probably going to be giving up significant share as things go forward to Cogent.

Yes, it was 6 years ago, so a completely different world, but 15 times EBITDA. And you go and look through their proxy peers: 25 times, 25 times, 22 times, 20 times, 10 to 20 times. Like I said, we’re trading at 8 times 3-year-out EBITDA, if you believe their numbers, and you probably think they’re higher. What’s the brain damage worth? Every peer is 15 to 20 times. So you’re talking about double the multiple I talked about, and, by the way, all that accrues to equity on a significantly levered story.

Anything else we should be chatting about?

I don’t think so. I think we covered it pretty well. Look, I appreciate that it’s a lot of brain damage, but it’s going to simplify. The cool thing is that if it works well, it should simplify to a really nice growth story with high incremental margins. The Sprint impact on the financials is going to decline. Hopefully, it will be a negligible impact by next year, and all the financial characteristics will start turning up as they should. If this works the way it should, that’s what should happen.

Then it should be a pretty compelling growth story in telecom, and it does feed the AI ecosystem. Who’s a buyer of wavelengths? It’s not like wavelengths are a very specific, highly technical connectivity service. It’s not just plugging into the internet; it’s facility-specific, from one specific facility within a city to another specific facility. When you’re using wavelength protocols to transfer data, these are massive files. You need guaranteed latency and connectivity. It’s a higher-end service and a bigger end market, and it is what a lot of the hyperscalers and the AI industry are going to grow with when it comes to their connectivity needs.

Everyone wishes they had dark fiber, but there are only a handful of dark-fiber providers. Even if Lumen is going to open up the network to sell dark fiber to these guys, everyone still needs multiple providers. It’s not a sole-source market. One thing we haven’t touched on is whether they can monetize their dark fiber, because that’s another potential monetization angle on Cogent.

I am skeptical that they’ll get anywhere material on that because I think if we see Amazon buying out an entire line of fiber on the first order, I don’t think, if I were Cogent, I would be motivated to sell out any dark fiber if I could light it up for customers and get much more monetization out of it. Maybe here or there they’ll do some, but I probably wouldn’t expect any material monetization of dark fiber, because if it’s true that there’s that much demand for the wavelength network, you would want to hold back that inventory for the wavelength business.

I mean, Aaron, correct me if I’m wrong, but dark fiber is—you sell it to Amazon, so it’s a one-time sale, and now Amazon controls it. You don’t get that ongoing revenue, whereas with the wavelengths, they’re basically renting, right? So you’re switching between: Is Amazon going to own it, or are you going to own it and Amazon just rents it from you, and you get the ongoing revenue stream?

Well, the way it would work is that it’s usually a big upfront check, and then you get ongoing maintenance. So maybe a nationwide contract would be $500 million, but then you get $25 million of annual maintenance on that contract. There’s some ongoing recurring component, but if you did a nationwide wavelength, if you monetized all the wavelengths in that nationwide network, that would be probably $1 billion of revenue or something. So there’s a big value gap between dark fiber and wavelengths.

Perfect. Perfect. Cool. Aaron Chan from Recurve put some great stuff on Cogent and a bunch of other stocks. I mean, you only have 4 or 5 stocks, but there’s great stuff on Royal Caribbean, which I’m always fascinated by. Ever since the pandemic, I saw someone call themselves a rat addicted to cocaine when it came to cruising.

I've been obsessed with the cruise line, obviously Carvana Co., but this has been awesome. Thank you so much for coming on for a second time, and I’m looking forward to the third.

Thank you so much, Andrew. I enjoyed it. Thanks.

A quick disclaimer: Nothing on this podcast should be considered investment advice. Guests or the hosts may have positions in any of the stocks mentioned during this podcast. Please do your own work and consult a financial adviser. Thanks.