All right, good afternoon. Thanks, everybody, for being here. My name is Frank Louthan. I'm the Senior Analyst here for Raymond James. This is the 46th Annual Investor Conference. I've only been here for about 24 of them, and I think we've had Dave here for most of those as well. Very pleased to have Dave Schaeffer here back again from Cogent. I'm going to go through a few questions and then maybe get some from folks in the audience. Dave, maybe why don't you start out a little bit, tell us a little bit about Cogent, kind of what you do, how you fit into the telecom landscape, who you're like and who you aren't like, and with some of the names folks might be familiar with. Thanks.
Yeah, sure. First of all, Frank, thanks for hosting me again. I'd like to thank investors for taking time to listen and always thank Raymond James for a great venue where we get to talk to a lot of interesting investors. Cogent is a global internet service provider. We have a physical network that spans 91,000 route miles of inner city fiber and 32,000 route miles of metropolitan fiber. We serve customers in 56 countries around the world, and we have two very different customer segments. We serve service providers in approximately 1,800 data centers around the world where we sell bulk internet connectivity. In North America, we sell transport services between data centers via wavelength services.
We serve end user customers that can either be corporate end users, small and mid-sized businesses, or global Fortune 500 companies where we serve their footprint internationally as well as domestically. We're about $1 billion in revenue, approximately $350 million of EBITDA. We have a dividend and have grown that dividend for 50 consecutive quarters sequentially. I think we're oftentimes compared to Lumen and Zayo. I think there are similarities, but substantial differences. We were not a telephone company. We are actually not a roll-up. We are a network that was built from the ground up out of the piece parts acquired in various acquisitions. Between 2001 and 2004, we did a total of 13 acquisitions.
We typically fired the customers, fired the employees, and then repurposed those assets to build the all IP over DWDM network I described, protected at layer three using Ethernet as the customer interface and solely productized that to deliver internet and VPN services. Our business expanded slightly as we added colocation services. Today, we operate 159 data centers around the world that are our own in addition to the 1,800 third-party data centers that we connect to. Our data centers are approximately 197 megawatts of power. Recently, the company has expanded its product offering further to include optical transport. In September of 2022, we announced the acquisition of Sprint GMG from T-Mobile. We have been busily working on repurposing those assets into a network that is optimized to deliver point-to-point services between any data center in North America. Today, we have the capability to serve about 880 data centers.
We can provision a data center-to-data center connectivity product at either 10 gig, 100 gig, or 400 gig speeds, and we can do that in a month or less. That gives us a substantial differentiation to our competitors.
All right, great. Very good overview. Walk us through kind of Cogent today, kind of how you've gotten there versus where you were a year ago and what's kind of changed in the last 12 months.
In the last 12 months, we have continued to groom unprofitable business that we acquired when we acquired T-Mobile wireline business. We were paid $700 million to take that business. In the past 12 months, we received about $200 million from T-Mobile of those $700 million payments. They were really meant to be an offset against the burn that we inherited when we took over that business. We also had to diligently work on taking a network that had sat idle for a decade and turning it into a productive asset. We connected the Sprint network to our Metro network. We reconfigured our Metro footprint. We deployed reconfigurable add drop multiplexers and transponder shelves. Finally, we have been very busy at repurposing some of the Sprint switch sites into data centers.
Okay. When you look at the opportunities that that network acquisition gives you, what are the two or three that are kind of the most, you're the most excited about if we look out a couple of years that'll have the most impact?
I think the most exciting opportunity is to enter a new but adjacent market that is to sell point-to-point services known as wavelengths. We have a more ubiquitous footprint than any other provider. We can provision more quickly. We have unique routes in over 90% of the instances. We can provide detailed mapping along with those routes, and we have a network with no cost basis. The second big advantage that we have comes from the physical buildings that we acquired. We are in the process of repurposing some of those data centers with their surplus power and marketing it off on a wholesale basis. The third opportunity that is new to Cogent is serving the large enterprise market space. In serving those customers, we needed to be disciplined around the products that we sell.
We needed to focus on making sure that each service was profitable, whether it be by speed, network access technology, or location. We have gone through the majority of that revenue attrition associated with jettisoning less profitable businesses. I think from here, we expect to be in a position to grow our free cash flow, which is ultimately the most exciting thing of all.
All right, great. Talk to us about the wavelengths and so forth. It is really kind of the bigger opportunity here. Talk to us about what is a customer using a wavelength for that maybe they were using other services before. That market has really taken off even since you made that acquisition. Talk to us about what people are doing with wavelengths.
Yeah. The cheapest, easiest way to move data is the public internet. In fact, it's about two and a half times cheaper on a per-bit mile than a wavelength. There are three characteristics that a wavelength can deliver that the internet can't. One, it's completely secure. Two, there are no limitations on the size of the packets transmitted. Third, the latency or distance between endpoints is completely defined, whereas in the internet, it's approximated as packets can arrive taking different physical paths. The primary customers for wavelengths are four groups. They are regional networks that are looking to tie their networks together. They are international and subsea carriers looking to extend their networks. They are content distributors who are looking to mirror content across multiple locations. Probably the most exciting application over the past year has been the growth in AI opportunities.
Most AI training occurs in different locations and where the data is resident. The reason for that is there is usually not sufficient power at the data center, which has been storing that data for training and the GPUs necessary to train it. For that reason, companies buy wavelengths to move that data from a repository data center to a training facility and then back again. Because the GPU utilization is the most precious commodity for that training, companies are very reluctant to have those processors running at anything less than full theoretical efficiency.
Okay, great. How would that be different from, say, the regular other customers that folks are familiar with, your corporate customers, your other networking customers, that net center customers that you have historically? What is different about a wavelength customer? How should we characterize that?
They're almost all net-centric. At Cogent, 18% of our revenues come from selling to large multinational companies who are typically buying two services, internet access and private VPNs based on MPLS distributed around the world. 44% of our revenues come from selling to other corporate users that are typically located in the central business districts of major cities where they're buying either internet access or VPN services, actually based on a different technology, VPLS to connect their remote offices together. Finally, 37% of our revenues come from selling to other service providers. These could either be access networks or content generators. The companies that are buying wavelengths are predominantly those service providers. It looks much like our net-centric customer base.
All right. Part of the story last year was getting the 800 data centers set up to sell the wavelengths. You exceeded that number by a little bit, continuing to build. Tell us how that sets you guys up to grow that business and why that was so important to reach that goal.
Yeah. Let's start with the addressable markets. Cogent's internet access market is a $1.5 billion addressable market that in unit volume has grown at about 25% a year, but has been subject to 23% compounded price erosion, keeping that addressable market flat at about $1.5 billion. Cogent has 25% market share. We are the largest carrier of internet traffic in the world. Our end user market is large. It's about $10 billion, but it's highly dependent on having last mile connectivity to be cost-effective. We today touch about 11% of all rentable office space in North America with on-net services. The wavelength market is $7 billion globally. Three and a half billion of that is in North America. That market is split between inner city and metropolitan. Approximately $1.5 billion is local or metro waves and $2 billion inner city.
It is that $2 billion inner city market that we are primarily focused on. There are about 145,000 wavelengths in that market generating $2 billion in revenue. For Cogent, we have distinct competitive advantages, whether it be the ubiquity of our reach, the uniqueness of our routes, the speed to deliver, the accuracy of our mapping, and maybe most importantly, the ability to price at whatever it takes to clear the market. Because not only did we get the network from T-Mobile for free, we were actually paid $700 million to take it because we acquired the enterprise business that was hemorrhaging cash. We have been able to fix that enterprise business to the extent where it's at least cash flow neutral today.
All right, great. Another aspect of the story gets a lot of attention is the data center capacity that you have for sale. If we look out 12 months, how much of that capacity that you have will get sold? Will it be sold outright or d o you think you'll lease it up and get some monthly recurring revenue from that? How should we think about how that'll end up?
Data centers are a small part of Cogent's revenue. We have a total of 159 facilities, and we have approximately 197 megawatts of power in about 1.8 million sq ft. We identify 23 of these facilities that have substantial surplus capacity. 1 million sq ft and 109 megawatts out of that 197 megawatts are data center services that we need to sell to others. We began a process of refurbishing these former switch sites and marketing that data center capacity. We are still completing that effort. We invested nearly $100 million in capital to have that conversion, and we expect to either sell or lease that excess capacity. We have gone to market with a straw man of a $10 million per megawatt sale price or $1 million per year on a triple net lease basis.
That would represent about a 40% discount to where the current market is. Since the demand from AI has mopped up all of the excess capacity in the market, we think that the existing power that we have in place is a scarce and valuable commodity. If we sell them, we will use that capital, put it on our balance sheet, and reduce our leverage. If we end up leasing them, we would take that revenue in a special purpose vehicle and look to securitize it, further giving us access to capital and reducing leverage on the organic Cogent balance sheet.
All right, great. We have touched on AI a little bit here with some of the wavelengths and where your networking goes and so forth, some of the customers, but maybe give us a fuller picture of how Cogent, as one of the largest carriers in the world and your different product sets, how you will really fit into an AI world.
Let's start with the raw material for AI, which is data that is collected over the internet. Over the internet's 30-year history, it has collected and stored about 800 zettabytes of data, and it is increasing that inventory at about 200 zettabytes a year. That data is the material that then trains large language models. You take a GPU and you place it against that data set with a question that you already know the answer to, and you watch the process of asking that question over and over again till you build a neural network that is the pattern that came up with the correct answer. You take that pattern and then project it into new data sets. That is known as the inference phase.
In terms of where Cogent plays as the largest carrier of internet traffic, we are helping people collect more data today than they've ever collected in the past. Secondly, as we migrate past the training phase to the inference phase in AI, most of those inference outputs will be distributed across the internet to the end users. What AI is, is the ability to use capital rather than people to create very agile, very customized process software that had previously been done by large groups of humans in a very rigid manner. The other benefit to Cogent outside of its internet business, but in its optical transport, is the fact that the large language model training is typically happening in locations that have ample power. Those locations are typically different than where the data sits.
Therefore, there's this need for wavelength transport services that benefits the non-IP portion of Cogent's revenue streams.
Okay, great. Talk about a couple of other. Talk to us about the corporate business. How would you assess where that is in its recovery? Obviously, been down since COVID. Talk to us about the corporate business and how you see the outlook there.
Yeah. So our corporate business, where we sell to end users with one or more locations in our on-net footprint and then branch offices in separate locations that are off-net, had been growing at about 11% a year between 2005 and 2020. With the COVID pandemic, that growth rate turned - 9%. It returned to about a + 4% growth rate. In its most recent quarter, that organic growth rate for that corporate business is about 3%. The vacancy rate in our on-net footprint has gone from 6% to a peak of 18% to today at 16.5%. So while the business is improving, it's still been impaired by the hangover from the COVID pandemic.
Okay. What is the outlook for that business? Where do you expect that to go? Is this as good as it's going to get, or can it recover some more? What are your thoughts?
I think it continues to recover at a slow rate. I know recent initiatives to get employees back into the office is a positive, but this data is about a quarter old now. Pre-pandemic, we were running at 97% of workdays in the office for all employees. As of Q3, that number of office worker days in the office was at 63%. We're still running at about a 30% gap in terms of people working from home. I think a lot of companies have struggled with what their final office architecture is going to look like. How many offices, how big will they be, and how many days a week employees will be required to be in the office. I think long term, we expect our corporate business to resume growth similar to pre-pandemic levels since we are only about 25% penetrated in our footprint.
Our large enterprise end user business is different. It's almost all off-net. We're moving about half of that on-net. It tends to be very sticky, and that customers tend to stay a long time, but there's very little growth in that business. In our enterprise segment, we have been rotating out of less profitable products, but stabilizing the revenue for either the VPN services or DIA services we want to continue to support.
All right, great. Let's talk about the IPv4, the business that you have. How important is that going forward? Maybe talk about the growth in that business this year. How much of that is going to come from the cycling through and new price increases versus new sales? How should we think about the trajectory of that business?
We are the third largest holder of IP addresses in the world. When the internet was first architected, there were three foundational technologies: TCP/IP, which is the technology that allows two machines to communicate with one or the other; BGP, which allows two networks to communicate with one another; and then a unique hexadecimal numbering scheme known as IPv4. There are 4.3 billion potential addresses. They were all owned by the U.S. government initially. The U.S. government kept 800 million and distributed 3.5 billion to commercial enterprises. We have 38 million of those addresses. Of those, we have leased out 13 million of those addresses at an average price per address of about $0.48 per address. We have been able to raise pricing and continue to lease out addresses.
Over the past year, we have grown the revenue from that address leasing business from roughly $30 million run rate to a $50 million run rate. We expect that number to continue to grow, and it will be both by further price increases as well as continuing to lease out some of our existing inventory.
All right, great. Talk to us a little bit about your liquidity position as the T-Mobile payments are set to decline and eventually as they go away. Walk us through sort of potential other sources of funds. You've done some stuff in the ABS market. Talk to us about that as that liquidity goes away. What's sort of the glide path there?
Yeah. Cogent's target is to be levered at around three times with a range of two and a half to three and a half. We are above that now because of the capital we're spending and the OpEx associated with integrating the Sprint assets. When we acquired Sprint, we had a payment stream from T-Mobile of $29.2 million a month for the first 12 months, stepping down then for the next 42 months to $8.3 million a month. In months 55 through 58, a working capital drop of about $7 million a month. Those payment streams stepped down in June of 2024. We are at a run rate now of $100 million annually, and that will continue through November of 2027, at which point we then get the additional working capital true up from T-Mobile.
They have made all their payments on time, and we anticipate they will continue to do so. Our leverage has increased to over five. We have a number of methods to get that leverage down. One will be the disposition of some assets, and we have three. We've talked about data centers. We've talked a little bit about some of our surplus IP address inventory and potentially some of our excess dark fiber. We also have been rapidly taking costs out of the acquired business. We have gotten the burn rate on that acquired business down to zero from a business that was burning $300 million annually when we signed our purchase agreement. Finally, we will have the ability to increase our high margin revenue services and therefore our EBITDA.
If we look at the beginning of 2020, well, let's look at end of year 2022, we did $260 million of EBITDA on $600 million of revenue pre-Sprint. With that transaction, our revenue stepped up to $940 million for 2024, or excuse me, 2023, and our EBITDA was $352 million. In 2024, our revenue full year was $1 billion and EBITDA of $348.4 million. We anticipate revenue and EBITDA to be effectively flat for 2025, inclusive of the $100 million payment stream from T-Mobile. We will have cycled through that bad revenue attrition, and we will ramp up our wavelength sales, which will rapidly deliver us. Delivering will occur one of three ways. It will occur either through growth in EBITDA in the organic business. It will occur through the divestiture of certain assets, or it will happen through the leasing of those unused assets as incremental revenue streams.
All right, great. Anybody in the audience have a question you'd like to ask? No? All right. Maybe, Dave, just to kind of wrap up, what are one or two things that you think are most misunderstood by investors about the company, and how would you address those here?
You know, as a public company doing a major acquisition of a declining business, that is a challenging situation. Rarely are companies paid $700 million in cash to acquire an asset that was constructed at a capital cost of $20 billion, $500 million. That's the fact pattern here with the acquisition of the Sprint GMG business from T-Mobile. I think the rotation of the revenue from less profitable to more profitable revenues, the repurposing of the assets from fallow to revenue producing, and the growth in the core underlying IP business are all represented together in our financials. I think it's hard for investors to parse out each of those trends, but I think if you just look at the most recent quarter, we delivered 280 basis points of EBITDA margin expansion on a sequential basis.
The fact that we pay investors a dividend that has grown for 50 sequential consecutive quarters, I think mitigates some of that confusion.
Great. Thanks, Dave. Really appreciate it. Thanks, everybody, for being here. We've got a breakout session after this if you'd like. Thanks.
Hey, thanks, Frank.