Good afternoon, everybody. Thank you very much for joining us. I'm Dave Barden. I head up U.S. and Canadian Telecom and Comm Infrastructure Research for Bank of America. Thank you for joining us. I'm really pleased to have with us today Dave Schaeffer, the Founder, Chairman, CEO of Cogent Communications to have a conversation. This is a very interesting day for this meeting to be happening, so I'm really looking forward to this conversation. But in order to set the table, there's also been a lot of other stuff going on, and I'm really pleased to have with us the host with the most, Jill Hall, who's our head of mid-cap strategy to hopefully maybe set the table for all of us on a couple topics.
One, I think, is just the general craziness in the market. And then second is just how maybe small cap TMT, which has been maybe ignored in the shadow of the Mag Seven, might have an opportunity to shine now that there seems to be rotation going on. So I don't wanna... Maybe I'll hand it over to you, Jill. Thank you so much for being a part of our conversation today, and I'll hand it off to you.
Yeah. Thank you. Thank you so much for for everyone for joining, and and for David for hosting, and Dave and all of our other companies for for joining. I'm Jill Hall, head of small and mid-cap strategy. I also work with Savita Subramanian on our overall U.S. equity strategy and quantitative strategy. So just wanted to hop on. Good mid-cap conference we've had going on the last day. Continues tomorrow. Opportunity to hear from a lot of different companies. Feel free to to reach out to me if I can help anyone with anything on the the smaller mid-cap side, connecting you with other analysts, signing up for other sessions.
Also just wanted to give, as Dave said, the market volatility, just give a couple of high-level comments on what's going on, opportunities within that, how this impacts midcaps, TMT, et cetera. I'd say the overall volatility that we've seen, it's. One could argue that by historical standards and some of the signals that we're looking at, it could have been expected, in that we were sort of due for a correction. Even when you look at, you know, just S&P 500 back 100 years, typically, these 5% corrections happen about three times a year on average. 10% corrections happen about once a year. We haven't had a 10% pullback since last fall. Volatility typically picks up from July to November in election years. So usually you see about a 25% rise in the VIX.
We've also typically seen the yield curve be a good long lead signal of what volatility does, and that was suggesting volatility could bottom about midway through this year and then pick up. So all of that said, it, you know, could be a volatile near-term period for the market for several more months now. But with volatility, that brings opportunities. And, you know, certainly there's some uncertainty over the weakening macro data recently. Prior to that, we had seen, obviously, a big rally around expectations around Fed cutting, which is particularly relevant for small caps, given their rate sensitivity. So a lot of refinancing risk for smaller companies, a lot of rate sensitivity for longer duration small caps and non-earners.
I think now the focus has sort of shifted to fundamentals, you know, not only given the macro, but given the fact that... You know, we're seeing a broader shift within the market in terms of, as Dave pointed out, Magnificent Seven had been leading the profits growth story. They've continued to surprise to the upside. Now you're starting to see growth for those companies slow down, and they're also investing a lot in CapEx, these big hyperscalers, relative to their earnings. And now, finally, earnings for the rest of the market are expected to pick up. And for large caps, the other 493, that started to happen. For small caps, the story has been pushed out a bit.
So I think what investors are now focused on is evidence of, you know, companies that can show that the profits are recovering, you know, shifts in estimates or guidance. I think fundamentals will be a big focus. I think also relevant to, you know, TMT, as Dave said, I mean, number one, just this broadening out of the market. If profits broaden out, then performance theoretically should follow. And, you know, a lot of the Mag Seven stocks have become very crowded and expensive relative to the rest of the market and also, you know, small caps relative to large. A lot of these small caps and parts of TMT have been underowned by investors.
I think dividend stocks have also been sort of a stealth underperformer within small caps and often a way to hedge regime risk. So that's been another interesting one where, you know, as profits recover, dividends tend to follow. And you know, we were in a period for a long time where price returns kind of drove the market, but historically, dividends were a larger share of investors' total returns, and we think that could come back. So I think that's another relevant one for TMT, and we've now seen the S&P 600 small cap index have a higher dividend yield than the S&P 500, and kind of an investor focus on dividends, even for small caps.
So lots of crosscurrents going on, but bottom line, we think it'll be, you know, a stock picker's market, and so we're really happy to be hosting this conference. And, you know, our analysts here at BofA cover about 1,000 small and mid-cap U.S. stocks, so definitely lots of opportunities for Alpha within the small cap index, even if we see the rally stall near term. And so with that, I can turn it back to Dave and Dave. And if anyone has anything I can help with or questions on the broader mid landscape, definitely don't hesitate to reach out to me. Thank you.
... Thanks, Jill. So I appreciate the insights and the comments and all the strategy work you do, and being a part of this conference, so thank you for letting us do this. And so Dave, man of the hour, thank you for being here. I really appreciate it.
Well, thank you, Dave, and thank you, Jill, both for hosting us and giving us this great forum to chat with investors.
So Dave, there's a lot to talk about. Obviously, results are coming out on Thursday. Everyone's interested in an update on the business, and I'm gonna weave this all together. But I'm gonna start with a question that's less about Cogent and more about the market. And so we have been watching one of your biggest competitors, Lumen, kind of unfold a story in the last couple weeks in slow motion. First, a kind of an AI quid pro quo type of deal with Microsoft, which is the CEO, Kate Johnson's alma mater. Then, on the Corning call, something about, you know, reserving two years' worth of fiber capacity, kind of suggesting that there was a deal that was brewing, some reason why they believed they needed this much fiber capacity. And then, you know, the...
I guess, the surprise announcement, the culmination of this slow-motion narrative last night was that there was a $5 billion, and I'm not gonna. I've been dealing with people all day about the syntax, so I'm gonna use the wrong precise word, but a $5 billion deal with unknown counterparty or counterparties, you know, to do something in AI. And you know, everyone on this call who has an interest in this has read the press release, and so I won't go through it, 'cause I wanna focus on your views, Dave. What... A $5 billion fiber deal sounds like an awful lot. And, you know, some people might view it as transformative for the industry, might view it as transformative for Lumen.
We'll talk a little bit in a second about what it means potentially for Cogent, but what do you, when you wake up this morning and read what this is, what do you think it means?
Okay, so maybe let's step back and understand three big trends in our industry. The first is the deployment of fiber as a replacement for other transmission medium. Back in the late 1990s and early 2000s, literally $several hundred billion of capital was deployed that facilitated the internet, but resulted in little actual revenue. Created tremendous economic growth, enabled what we know now as the public internet, and facilitated the burgeoning of dozens of business models and applications that have transformed society. AI is building off of that. AI is possible because of the huge amounts of data that were collected at virtually no cost over the internet, and then the ability to process that data through large language learning models to create predictive pattern development.
That is effectively what AI is today, and that has the ability to drive another leg of productivity improvement and a transformation across society. The second key issue has been the migration of all services and telecom to over-the-top, using one common IP network. And then the third trend has been the proliferation of fiber to endpoints to be able to allow everyone to take advantage of the full capabilities of the internet. I think the announcement that you're referring to from one of our competitors leverages all of these trends. It may not result in a significant amount of profitability, but will further enable the deployment of AI large language models. Because of the constraints of power and data center space, the training of these models occurs in a distributed fashion.
Many of the hyperscalers are, in fact, building proprietary data centers or data center campuses where they can get affordable land and available power. I think a key component of this announcement is going to be the construction of new fiber from existing endpoints of pre-existing fiber backbones to these brand-new developed sites. Cogent has chosen not to participate in that market because there is too much monopsony power on the part of those hyperscalers who are constructing these facilities. You spend $10s of millions, maybe $100s of millions, building fiber to these newly constructed data centers with only one customer in those facilities, and little opportunity to harness additional revenue streams. I think a key component of these announcements has been the fact that one company decided to hire another company to outsource that construction, utilizing their existing right of ways.
The second will be the purchase of fiber that is surplus along the existing routes that are in place today and are unused, so selling off excess inventory. Whether that is going to generate significant operating free cash flow is yet to be seen. So Cogent is the largest carrier of internet traffic in the world. We carry approximately 25% of global internet traffic. We do that over a network where we have purchased fiber from 328 different suppliers around the world, just as any hyperscaler could go out and purchase that capacity. With the advances in wave division multiplexing, you don't need a large number of fibers to have virtually unlimited throughput and support any of these AI training applications. Our model has been to purchase that fiber to connect to major traffic aggregation.
We connect to 1,680 carrier-neutral data centers in 54 countries around the world. This allows us to access roughly 98% of global internet traffic. What we have chosen not to do is develop extensions of that network into these proprietary locations, where all of the market power resides with the purchaser, and there is only one customer. So, you know, where we sit in this ecosystem is two places. One, we're connecting those 1,680 data centers, we're carrying a quarter of the world's internet traffic, and we are connected to over 8,100 other networks that make up the internet. That gives us a unique position in helping aggregate that traffic for these training models. The second place that we participate is we, about 15 months ago, acquired the Sprint Global Markets Group network.
It was a 19,000 route-mile inner-city fiber network that had been built at a capital cost of $20.5 billion between 1982 and 1991 to carry long-distance voice. That network had been maintained but had become unoccupied. As Sprint shut down its long-distance voice business, the network quickly became a liability. We were paid $700 million in cash over 54 months to take over that network from T-Mobile. What we're doing is repurposing that network to provide optical transport services. So Cogent had historically only sold internet-based services. Optical transport services have three characteristics that the internet cannot replicate: They have defined latency, they have defined endpoints, and they are completely secure. Those attributes are valued by companies that are doing large amounts of data replication.
The cost to move those bits, however, is more expensive, generally 2-2.5 times more per bit mile than just using the public internet. We are entering a market to sell wavelength services or optical transport. What makes our services maybe different than what Lumen has announced is we took a former voice network, are in the process of re-engineering that network, and making it so it can provide any data center to any data center optical transport capability, and being able to provision that with a 2-week interval. That puts us in a very unique position.
We have begun selling those services, so they are analogous to the transport services that Lumen announced, with two key differences: one, the endpoints, and two, at least from what appears to be what has been indicated in the cryptic press releases, that this is a dark fiber sale, and they're looking for the counterparty to light that fiber. Hopefully, that background, David, was helpful in you kind of framing this question.
No, thank you. Thank you, Dave. So I think that I had a follow-up question, and I think that the answer is in that answer, but I just wanted to maybe fine-tune it a little bit, which was that, you know, that your perspective, and in the press release, you know, Lumen talks about how the majority of the $5 billion is gonna be both received and spent in the first three to four years of the relationship with whoever it is or the whoevers it is.
I guess I agree with your idea that, you know, a lot of this is gonna be building new stuff to new places, that these places may not even exist yet, and that's why it takes three or four years to build this stuff, which it wouldn't take if these places existed. But the question I'm getting is, when you acquired Sprint, and you mapped out a number, and we'll get to these things in a bit, but you mapped out a number of opportunities that the Sprint asset brought to you, and you talked about one of them is the Lit Wave opportunity. And you sized that opportunity at $2 billion today.
I guess the question I've been getting today has been: Does this fiber deal represent an evolution of the marketplace, which is where people no longer want waves, they want full fibers? And then does that diminish your opportunity set, or does this fiber deal potentially represent, you know, an expansion of the market? Maybe it's not the wave market necessarily, but it's an expansion of the market. It's not cannibalistic, it's additive. I think I know the answer, but I just wanted to clarify it.
Yeah. So, there are three ways to move bits. The cheapest, most ubiquitous, and easy to use is the public internet. That dominates the amount of traffic that is carried globally, and that will continue to grow because of the application-agnostic nature, as well as the efficiency of the internet. That $2 billion North American wave network opportunity has been growing at modest single-digit rates. It is today about a $7 billion global market. About $3.5 billion is North America, and of that $3.5 billion, about $1.5 billion is metro wave, so basically from one point within a given market city to another, and then about $2 billion is intercity. The third and most expensive, or more expensive way, is to buy wavelengths. Now, those attributes that I talked about earlier may justify that incremental expense.
And then there is a fourth, and even more expensive way, which is to buy dark fiber. Now, why would you buy that dark fiber? One, you may want to go to a site where none of those three other alternatives is available. If there's nobody there, you can't connect to it, it's kind of useless to you, so you either have to build the fiber yourself or get someone to act as your construction outsourcer. That is what is happening in a lot of these new-build environments. The second piece is, once you connect to fiber that is already in place, do you want your own dedicated fiber as opposed to a wavelength? And there, it's a question of control, it's a question of: Are you comfortable with your counterparty's ability to maintain it and grow it, and their financial capabilities?
And then finally, do you view this as something as long-term strategic? And the answer probably varies among each of the hyperscale operators. There will be some of them that do elect to buy dark fiber on backbone routes, but there's nothing new in this press release. That's been going on for 20 years. You know, just as Cogent has gone out and bought its backbone fiber, 91... Well, we bought roughly 72,000 route miles of fiber, and we own 19,000, so we operate 91,000 route miles of inter-city fiber. But that kind of shows that there's plenty of inventory out there and plenty of existing counterparties that wanna sell it, and most of those counterparties have lost tremendous amounts of capital, generating virtually no return on their deployed capital.
In fact, in many cases, those companies have gone bankrupt once or even twice, and are still struggling, which is why that inventory is readily available. So I think there are buyers out there, but there are also a lot of sellers in the market. I think what we fill with Wavelengths is a segment of the market that wants those three attributes of a dedicated network, but want them on a flexible, scalable, and ubiquitous feature set and location inventory, as opposed to going out and buying dedicated routes for 20 or 30 years, obligating yourself to large maintenance payments, without necessarily having a lot of confidence that the counterparty that you're paying those payments to can continue to provide those services.
Got it. So you, you don't wanna say right here, right now, Dave, that you're about to jump into AI fiber and have your stock go up 300%?
Well, first of all, I have no idea what the term AI fiber means. I know what a piece of fiber is. I have a sample in the room next to me, and, you know, the fiber basically comes in two different varieties. You had the initial deployments of single-mode fiber that were deployed in the eighties and nineties. It was then replaced by non-zero dispersion-shifted fiber that became popular in the late nineties and early 2000s, and now all of the fiber that is being deployed has reverted back to that standard single-mode fiber. And the reason for the two different types of fiber was the optoelectronics that sent signals down that fiber. Initially, all of the signals were asynchronous. They then standardized on a non-coherent synchronous signal.
That technology did not maximize the optical spectrum in the fiber itself, and in the mid-2000s, the market transitioned to coherent transmission. With coherent, the non-zero dispersion-shifted fiber turned out to be sub-op, and the difference is, am I sending the information down the fiber as a coherent blob of information, or am I sending it incoherently as a set of disjointed pulses, rather than a single coherent transmission? And the non-coherent... or the non-zero dispersion-shifted fiber was developed specifically to prevent chromatic dispersion, the spreading out of the signal. So think of it as you look at light through a prism, and the colors bleed together, but then they kind of bleed out even further.
What is happening in a non-zero dispersion-shifted fiber is the fiber has a certain topology and a chemical injected into it that compresses that optical signal back together, pushing that dispersion together. When you put a coherent signal in that, it actually works very poorly. It actually does the exact opposite of what you want it to do. It takes a single coherent ball of information and distorts it or spreads it out when it doesn't want to distort or disperse on its own. So it is for this reason that the fiber that's been deployed since 2006, 2007, has all been back to standard single mode, and today, when you get the largest throughput on that fiber, it is all done with single mode fiber and coherent optics.
So Dave, I'll just assume that everyone... Thank you for all that, by the way. That's helpful context for the industry dynamic. I'm gonna assume that everyone on the call or here at the conference knows a little bit about what Cogent is all about. I guess, you know, one of the things, Dave, you know, one of your core businesses, the corporate connectivity business, you know, you've told this story, you know, it was growing 10%, it got hit by the pandemic. It's been kind of leveling out and coming back as we've been experiencing or finding the new normal with return to work, and all of a sudden, in the last three days, we're now talking about the possibility of recession.
So could you give us an update on, you know, kind of what, you know, how the Cogent corporate business has been developing? And give us a picture of, if we were worried about recession, should we be worried about this unit again?
Yeah. So Cogent, prior to acquiring Sprint, only sold internet-based services, and it had two customer bases. It had corporate end users and other service providers. Our NetCentric, our wholesale business to other service providers, represented 97% of our traffic, but only 40% of our revenues, and that business tended to be fairly volatile. You know, when, you know, Netflix came along and everybody started streaming, that business grew at double-digit rates, and then when we reached saturation on that application, it levels off, grows at much more modest rates, then the next application drives growth. The much more stable business for Cogent, up until the pandemic, was our end user corporate business. So in the U.S., the end market for business internet connectivity is about $9 billion. It's actually been relatively static for the past 15 years. Cogent focused on the best part of that market.
We went after any business that was located in a skyscraper in the central business districts of a major North American city. We connected 1 billion sq ft of skyscrapers to our network that represented about 1,860 buildings. So we start with a universe of approximately 4.1 million commercial buildings in the U.S. and Canada. Of those 4.1 million buildings, 3 million of them are single-tenant buildings that don't make sense. Then, we looked at the remaining, roughly 1 million buildings, and realized that we needed to pick the largest of those buildings, but also buildings that have a diverse customer base. So the average building we connected to is approximately 41 stories tall, 550,000 sq ft, and had 51 discrete businesses in it pre-pandemic. These buildings tended to be located in the downtowns of major metropolitan markets.
We were growing that business at 11% year-over-year for 15 consecutive years by taking market share. Customers switched to us in those buildings because our service was installed 9 times faster than our competitors, we were 3 times more reliable once we installed, and we were delivering 30-60 times the throughput. A great story. The pandemic hit, and we went from being a 11% year-over-year growing business in that segment to -9%. With us coming out of the pandemic, we have recovered, but only partially. That business today is growing at between 2% and 3%... not the 11% it was growing historically. We have picked the best buildings and the best markets. We have weathered several other recessions, the 2002, 2003 recession, the great financial recession, and in both of those cases, our corporate business continued to grow. The pandemic was different.
People's work patterns changed. Many companies are reevaluating their aggregate real estate footprint for office workers, but that business is continuing to improve slowly, and I think even if we go into a recession, that improvement will continue, albeit maybe not as quickly as we went down. You know, the events of COVID were a shock to the system. A recession that we would enter into now would probably take longer and be more diffuse. Finally, the buildings that we connect to tend to be the most recession-proof. They tend to be the best buildings in a city, the skyscrapers in the downtown, the Class A buildings. And the tenants that are in those buildings tend to be fairly well-vetted by the landlords, and much more recession-proof. So I think our corporate business is going to continue to improve.
As companies' IT budgets continue to get squeezed, and a recession does that to all expenses, companies constantly look to shift to greater value, more throughput, higher reliability, lower prices. Those are positives for Cogent's corporate business.
Great. Thank you, Dave. Just to check in, the other side of the house, the traditional house, the NetCentric business, you know, had a very strong boost as, you know, we saw the streaming wars take a toll on the media industry. You know, it, it had a big benefit to, you know, the connectivity business. Now there's kind of some, I don't know, I don't wanna say stagnation in the, in the streaming business, but there seems to be some plateauing, and, you know, rising churn, and some question marks among some of these players about, you know, how committed they wanna be, and should there be consolidation in the industry? You know, what is the... We know where the NetCentric businesses come from. Where do you think it, it's going with the margin?
Yeah. So we saw a step function increase in the percentage of video that was streamed during the pandemic. We went into the pandemic with 18% of all video consumption being streamed. Today, we're at about 49% of all video streamed, so almost half of all video globally is streamed. We're also seeing streaming continue to gain market share, albeit at a slower rate, and we're seeing the globalization of what happened in the U.S. and Western Europe being pushed into more peripheral markets. So even though there has been a slowdown in these more developed markets, we're continuing to see strong growth in streaming globally. Cogent's growth is global. We serve access networks and streaming providers in 180 countries. Now, we only have physical assets in 54, but, you know, whether it's China Telecom, or China Mobile, or China Unicom, all of them connect to us outside of China.
We're prohibited by Chinese law to actually have network inside of China. And on the other side, we sell to virtually every major streaming platform, whether it be Disney, or Netflix, or Warner, or Major League Baseball, or the NFL, or probably hundreds of names you haven't heard of. Whether it's CBS or NBC, all of them buy some upstream from Cogent. You know, I do think the current revenue models of some of those streamers are challenged. You know, streaming broke down the studio barrier to entry that had existed for nearly 100 years, and also allowed for the disintermediation of the aggregators. That has put tremendous pressure on the media business. There will be winners and losers. I don't wanna sound callous, but we at Cogent are somewhat indifferent to that.
I think our net-centric business, which in revenue terms is growing at about 10% or 11% a year, better than its long-term average, better than we were doing pre-pandemic, is continuing to grow at these elevated rates. And what we are seeing is three things: more people streaming. Two, spending more minutes a day using those streaming applications. And then three, as part of the war for those eyeballs, we're seeing the quality of that streaming increase. I have no ability to comment on the quality of the content, but I can comment on the quality of the transmission signal, where we're seeing much higher resolution, much faster action content, that all requires much higher video encoders, and therefore more bits. So, you know, as we move to a world of live event streaming, we will see yet another leg up in the growth of streaming.
Dave, I just got a question from a client, and I wanna kind of shift gears as quick as I can to maybe wrap up on some of the new opportunities like data centers and IPv4. But have you seen any deleterious effect from the de-peering, you know, exercises that you undertook so far this year, I think with Tata and NTT?
So with Tata, we actually fully restored harmonious peering connections. You know, Cogent is the most interconnected network in the world. We run over 29,000 BGP sessions. We directly connect to over 8,100 networks. We only have 23 peers. We buy no transit. As we gained market share in Asia, two of our global peers began to constrict our access to their customer base, so they grew peering connections with us in Europe and in North America, but not in Asia. We took a different approach with each of the two companies you mentioned. With Tata, we actively, selectively de-peered their BPO customer base. The pressure from that ended up forcing them to honor their peering agreement with us, and they established multiple peering connections throughout Asia, inclusive of Indian routes, and today there are no issues.
We restored full connectivity to their BPO customers, and we, in fact, got a full routing table for our customer base globally. NTT has taken a slightly different approach to its Japanese household customers. They have been willing to allow those customers to suffer and get suboptimal connectivity to the internet. As a result of that, we never had connectivity in Asia with them. They continue to have ample connectivity to us in North America, but we've removed their connectivity in Europe. So now, for a residential Japanese customer, if they are looking to get content, NTT has to bring that request bit all the way to the U.S., grab that content, and carry it all the way back to Japan. Whereas, if they allowed us in their Asian theater, we would carry that traffic and exchange that traffic locally. We have not had any similar breakthrough with NTT.
NTT is relatively irrelevant to global traffic. You know, much like of the access networks, the only other major access network that's taken this strategy is Deutsche Telekom. You know, whether it be BT or Orange or Telefónica, we have very harmonious relationships. But those two companies have been willing to allow their residential end user customers to receive suboptimal traffic. From Cogent's perspective, this is a de minimis impact on our global traffic volumes.
Dave, I wanted to say thank you so much for the time. I wanted to talk about a lot more stuff, but I know that for all those that wanna talk about all the stuff that I know that you know that we didn't get to, the conference call is gonna be out 8:30 A.M. on Thursday for the second quarter results, so we'll have a lot more to talk about. Data center sales, IPv4, dark fiber, anything we hear on the Lumen call tonight, that's relevant. So I wanna thank everybody for being a part of this. I wanna thank Jill for organizing the conference. And Dave, thank you for being a part of it again this year. We appreciate you.
Thank you very much, Dave, and I promise to answer every question on our earnings call.
You always do. Thank you.
All right, take care, all. Bye-bye.