Welcome to our 11th annual TD Cowen Commercial Communications Infrastructure Conference. My name is Greg Williams. I cover cable, wireless, telco, and fiber. Joined in this session by Dave Schaeffer, CEO of Cogent . Dave, thanks for joining us.
Hey, Greg. Thank you for hosting me. Maybe most important, I'd like to thank the investors for hanging around here at the end of the conference and not doing something more fun than listening to me run my mouth. I always want to thank TD Cowen for a great venue. You guys are committed to this space.
Yeah, we sure are. It's been 11 years, so it's great to be here. I wanted to start with, your shares are down meaningfully over the last week, you know, since earnings especially. One of the reasons is the calls on your stocks. You know, this time last year was about your commercial real estate business, but from what I understand, it's more about your tax situation. Maybe you can just contextualize the situation and where we are in this process now.
I always get criticized by Colby for a long answer, but it is necessary in this case. When I founded Cogent, I put initial capital in. I owned 23% of the company and had raised $500 million of capital. The telecom crash happened, and in negotiations with my investors, my 10 PE and venture investors, I agreed to voluntarily reduce my stake from 23% to 0.7% of the company, give up my liquidation and preference ratchets in exchange for getting access to all of the capital to buy distressed assets. In the period from April of 2001 through December of 2004, we did 13 acquisitions. We acquired companies like PSINet, Allied Riser, and we dismantled those businesses and reassembled them into what became Cogent. Cogent is not a roll-up. It is the piece parts of other companies.
As a result, when Cogent went public, my equity position was down to 0.5% with the public- equity dilution. My deal with my investors was that I could reconstitute my position by taking 100% of my compensation in stock, and I would do that over time. In fact, I rebuilt my position to 11% of the company.
What did you pay in stock?
Paid in stock and never received a dollar of salary. Again, it's all in the public filings. For the first couple of years, that was OK. As the stock appreciated, it got more expensive. I went to the Board and said, "I need to have the ability to pledge those shares to pay taxes." I have a tax basis as of the first of the year of $155 million that I had paid taxes on, so almost $80 million in card tax that I paid. That was the sum dollar value of my pledged position. I had to contribute stock to be able to do that. I kept a pool of stock unpledged and a pool pledged. As the stock came down, I went to the Board and said, "I would like to, rather than sell, put more stock into the pledge pool."
I did not anticipate the run that we had and effectively got wiped out of my position. I'm a big boy. I get that. That was the bargain I made. The confusion around my real estate. I have never pledged a share of stock for real estate. I have a portfolio outside of Cogent of 42 office buildings. I had office buildings before I had Cogent, and I built that business up, and it has been under tremendous pressure. Over the past 20 months, I injected $152,340,000 of actual cash into that portfolio to negate the decline in value. The portfolio at pre-pandemic levels was 53% LTV. Today, it's probably above 90% LTV. While I've reduced the debt load from $578 million to $422 million, the face value of the assets has fallen from $1.1 billion to below $500 million.
As a result, all of my lenders are looking for incremental equity injections. That is going to be very difficult now that I don't have stock or another source of capital.
You do have shares in Cogent today, but they're just restricted. Is that right?
They are restricted. They will unrestrict over time, and they will represent about 1.2% of the company. I guess I'm better off than I was 28 years earlier, but not 28 years' worth of work. Again, that's the hand I'm dealt.
Gotcha. Thanks for clearing that up. The other issues, I guess, were the EBITDA miss, the waves miss, maybe the lack of data center sale process, and maybe a risk to the dividend. To that last point, your dividend yield is over 13%. At this point, would it make sense to maybe temporarily pause the dividend and focus on a buyback?
We have returned almost $2 billion of capital to shareholders. I am very comfortable that we will be able to delever and grow our return of capital. We have predominantly done that through dividends. We have 52 consecutive quarters of growing the dividend, and we have returned about $250 million through 10.9 million shares of bought-back stock at an average price of $24.13. Prior to this recent sell-off, I could have pointed to a double-digit return on the compounding of those buybacks. I'm going to go back to the first two points that you made around misses. I don't want to sound defensive, but I disagree with some of that statement. First of all, we were extremely clear in our previous earnings call that our revenues would decline for the quarter sequentially.
While Cogent , as a company, does not give quarterly guidance as a practice, because of the complexity around the Sprint transaction, we have added incremental KPIs, and we've given incremental disclosures. We said that we will turn cash flow or revenue growth positive sometime in this quarter. Whether it'll be sufficient to be positive for the full quarter is hard to determine. I think we will return to the positive growth that we had before acquiring Sprint. To just refresh everyone's memory, if you looked at Cogent pre-Sprint, we had a 17-year public history of a compounded growth rate of greater than 10% with no M&A and an average of over 200 basis points a year of margin expansion. We acquired a business that was 40% the size of our business. It was declining at 10.6% a year, and it had negative EBITDA margins of almost 80%.
With that, we actually accelerated the rate of revenue decline by terminating certain negative margin products in bad locations. That revenue decline is mostly behind us. For eight quarters post-transaction, we have been able to sequentially grow EBITDA, and we grew EBITDA on average over those eight quarters, better than $5 million a quarter underlying. In the most recent quarter, we had a 200 basis point expansion in EBITDA margins. I'll be honest, I don't know many companies that can have that kind of EBITDA growth with negative revenue. For us going forward, we will accelerate that pace of EBITDA expansion by growing top line. There are still some further cost cuts.
I just want to talk about the moving around from EBITDA to just the waves business. It seemed to be off to a slow start, but you noted that several hundred installs are there, but they're just not yet built. Help us with that understanding. Who orders a circuit and then sits on it for this long? I'm just trying to understand that process.
I caught a little bit of one of your other interviews, and I think that CEO would have said the same thing, which is customers are expecting two things: one, long delays in installation, and two, a high failure rate in those installations, which may result in the circuit never being installed. I would suggest you or any investor do independent channel checks of customers who buy from the current supply base and ask two questions. What percentage of waves you order never get installed? For those that get installed, how long does it take? For Cogent, there are really four pillars to our competitive advantage. Two of them we've talked about because they're under our control and they're leading indicators. The first is where you can sell the wavelength. You can't service the market if you're not in those buildings.
We are in more carrier-neutral data centers than our competitors. At the end of the quarter, we were in 938. That's an objective metric that we could put out there that could be verified. We've actually sold waves in 428 of those 938. We have a guarantee to provision in 30 days. The customers do not believe us. We have been installing, and customers are anticipating we're going to behave like our competitors. Some customers take the waves right away, but many of them are dependent on ordering cross-connects, patch panel assignments, LOAs, pluggable optics, supplemental power. There are a variety of reasons.
Right. On point number two, before three and four, I just want to understand. I understand that you're a CIO of a company. It's a lot of coordination, and then the waves just came in too fast. Is that the right way of thinking?
That's effectively what faster than I expected. Too fast is the wrong term. It's just faster than I expected. Now, against Cogent, we are still not installing as many as we need to. We need a bigger backlog. That is building at a rapid pace. You know, the fact that we delivered 27% sequential revenue growth in waves and 149.8% year-over-year growth, to me, it's kind of hard to say that's a miss. I may sound paranoid, but those are pretty good numbers.
You mentioned that you have several hundred of these installed but not built waves in the 2Q call. Can you give us a color of what percentage of these orders will take, and then you can bill them? Like, is it half over the next quarter?
I can't answer that because it's in the customer's decision. We have tremendous data in our IP business because we've installed hundreds of thousands of ports. What I can tell you is we guarantee 17 days. We average nine. 62% of all orders request a delay in installation. 10% of all orders are eventually forced built, meaning we don't give the customer any more extensions, and you just start paying because you have a binding contract, you've taken it, you've pushed the date out twice, you now have to take the service. For wavelengths, we are too new and don't have enough credibility to have that kind of market power. I would suspect a year or two from now, after we've got tens of thousands of waves installed and customers know and can count on us, we can implement the same kind of discipline, and this problem totally goes away.
I think as we're building credibility and changing market behavior, we need to be accommodative.
Right. Can you tell us how July wave and build installed built?
What I'm not going to do, which I made the mistake last time, is giving people interim numbers and getting walked into something like guidance. I am not going to do that.
We're not touching that one. How about the.
Good try.
I try. How about the confidence in hitting that $20 million wave runway by year end? I mean, you feel like.
Very high.
OK, even with July and August thus far, you're feeling confident there.
I'm not going to give you specifics for months. What I can say is we ended Q2 with about 1,500 waves and a revenue run rate of $9.1 million for the quarter. We effectively have to double that for the rest of the year. It basically says if the RPU stays the same, we roughly need to be at 3,000 waves and go somewhere between around $20 million and $25 million of revenue run rate exiting the year. We have enough backlog, enough provisioning capability, and enough new orders being signed that I feel comfortable with that. I also want to emphasize that while we give these KPIs as leading indicators, at some point, they're going to go away, and investors will measure us solely on GAAP revenue.
Sure. The competition, Lumen's here, Zayo's here. You mentioned that you have a breadth of service, and you have provisioning times that are very quick. You price competitively. I feel like they're not sitting on their hands. Zayo announced the 400 Gbps wave capable network. Help us with the competitive reaction in the environment now and what you expect going forward.
You should always be paranoid about your competitors. They will react. I think the architecture of their networks is substantially different than ours and gives us some advantage. We have more sites where we can deliver 10 Gbps, 100 Gbps, and 400 Gbps in all directions. There are still probably about 8% of the vectors, meaning from one site to another, where we are not 400 Gbps enabled. That will be done probably by the end of this quarter, that remaining 400 Gbps cleanup. To date, 79% of our orders have been 100 Gbps waves. Whereas for the embedded industry, the mix is 55% 10 Gbps, 40% 100 Gbps, and just around 5% 400 Gbps. The final point I want to make is on our competitive advantage. As I said, there are four pillars. I gave you two. You mentioned the other two in passing. I don't want to expand on that. The third one is price.
We have to be a more competitive player. I heard Steve at lunch and he said, look, there's enough here for everybody.
You said the 10 Gbps and the 100 Gbps were getting more price pressure, but the 400 Gbps is where it's at, where you're strong in price?
That's hard to say because there's such a small install base of 400 Gbps. The reality is on a per-bit-mile basis, 400 Gbps is cheaper than 100 Gbps. It's just buying a bigger thing. It's just like with internet ports, 100 Gbps ports are cheaper than 10/100s or a 400 Gbps.
Yeah, you're buying in bulk, essentially.
You're buying in bulk. It's Costco. The final point, which I think we need to demonstrate, and there's no way to give this to investors up front, is around quality. You can have the first pillars everywhere, fast, at good price. If the quality is not there, you're eventually going to fail. The way we dominated the IP business, and both Lumen and Zayo are here, they're competitors in the IP business, but they're both much smaller than Cogent in that business. We have been able to do that by delivering the quality we do, the ubiquity, the speed to deliver. On wavelengths, we architected a network with only one purpose. I want to be clear. Cogent's running two completely independent networks. There's an IP network that's global that spans 125,000 mi, all based on IRUs.
There is a domestic North American, U.S., Canada, and Mexico network that is solely dedicated to waves.
On point number three, just to go back, pricing. In some markets, you're discounted at 20%. Sometimes it's 35%. I think in the IP transit world, in some data centers, you're half the level.
50%.
50% of the competition. How did you figure out that 20% or 35% is the right place to be as you think about winning subscribers or winning circuits versus pricing it appropriately?
I don't know. I'll be honest, we don't know the answer.
Sure, you're leaving money on the table if you're priced too soon.
Our pricing algorithms are generally set to a 20% discount to our competitive intelligence on a route-by-route basis. We then, in some cases, sell at full list, which is a 20% discount. In other cases, apply incremental discounts. Our pricing grid is based on length, port size, and contract term. We're discovering where we need to be. My goal is not to destroy the market, but to capture market share.
In capturing that market share, is there a concern that maybe you won't win the big deals, but you'll win like the singles and doubles because some of the bigger deals require larger RFPs encompassing other products? Can you win? What does a typical sale look like? Multiple circuits, and can you compete on the large logos that I think Zayo and Lumen would win?
We have done deals as small as one wave, one 10 Gbps wave, to customers that have taken several hundred 100 Gbps waves and even dozens of 400 Gbps waves. We have provided waves to a handful of hyperscalers, which are the largest buyers in the market, but they're still the minority of the market. It's important to understand that when a hyperscaler buys, they could have three different use cases for those waves. They could need waves for AI training. There's a whole bunch of buzz around that. They could use waves to just support their internal network for data replication. Third, they can use waves for their content delivery models. Whether there's a Netflix or a YouTube streaming, Google doesn't come and tell us, oh, this wave is going to YouTube and this one's going to the AI training part of the business.
You can't really see what it's used for.
You can usually figure it out, but it's somewhat irrelevant to us. They're a customer. They need a bunch of waves. We know that their use case is legal. You do what you want with it. It's kind of our same, you know, kind of hands-off approach to transit that, you know, we don't question what our customers do. I mean, we sell transit to the RNC, but we also sell to the DNC. We're not blue or red. You know, we sell to everybody.
I wanted to switch gears and talk about the data center sale process. You mentioned to me on the call, I think that it's too early to come off that targeted $10 million per megawatt valuation for the data centers. What's the rationale there? Is it because you believe you'll receive a lot more offers than the six LOI offers today? I'm just trying to understand why it's premature at this point.
A couple of the offers we have are at that price. It kind of says you don't negotiate against yourself, but you also need to give those counterparties the ability to actually either prove their ability to perform or not. I think we started this process assuming no data center conversion and sale other than taking 45 of the original Sprint facilities and putting a small retail footprint in them. We have subsequently increased that to 130 facilities that we are offering retail colo. Cogent today has 187 data centers, 2.1 million sq ft of data center space, and 214 MW of power. That's our footprint today, and today we're less than 14% occupied. We've got tremendous surplus. In the data centers that we identified as non-strategic, we've cordoned off 24 facilities, 1 million sq ft , and 109 MW of power.
We have no idea what we're eventually going to get for these. I've really tried to be careful to make sure investors don't put a number into their expectation. Second, it's not a recurring revenue business. It's an asset disposition. I think there's value. We've had over 160 parties express interest. We've conducted tours for over 70 of them, I believe. There are still over 50 doing work. There are tours going on literally almost every day. In the early part of this process, we were gathering requirements from the market. What needs to be done for you to be a buyer? That literally ranged from, we'll take them as is, to I need you to have liquid immersion, cooling, and cabinets in place so all I have to do is bring GPUs and deliver service. Why continuum there?
We ultimately coalesced out of that input a set of standard requirements. The most significant of those was the need to convert the DC plant to AC. When we did that, we are spending almost $100 million in that conversion process. That is complete. Those centers are ready to go. We looked at the current market, and the data is sketchy, and it's skewed towards large campus-built hyperscale facilities. We thought we were at about a 40% discount to the market. If we have to take a 60% discount, we will. We are motivated to sell these, but we need to have counterparties that are ready to perform. The facilities have two limitations. One, there's zero revenue with them, which is difficult for some private equity buyers. For the users, these are still relatively small and relatively low power density.
The typical facility is about 5,000 sq ft and 5 MW of power on six or seven acres. I've heard people say, if it's not on a 200-acre plot, I'm not interested. If I can't have 1 GW or more of committed power, it's not for us. That's not what these are.
Got it. You mentioned earlier, you've messaged to not put expectations of a data center sale. Just to be clear, if you don't sell any data centers, you can still honor the dividend. If you pause the dividend and do like a buyback, that's a different thing. Right now, you don't need to sell any data centers. You can delever and still honor the dividend. Is that right?
We can grow our aggregate return of capital and delever simultaneously. Our leverage peaked.
Without a monetization of anything.
Without any asset sales, whether it be IPv4 or data centers or dark fiber, any of the non-core assets do not need to be monetized in order to increase the return of capital. We have been committed to returning capital. We've returned almost $2 billion of capital. We've bought 10.9 million of our shares back, which represents about 22% of the float. We have 52 quarters of growing the dividend. The underlying EBITDA in the business is growing. On our LQA versus LTM, we're actually a half a turn of leverage lower. If nothing happens but the passage of time, that just clock ends up delivering us a half a turn. That is insufficient. We need to grow our EBITDA. If you look at the eight quarters post-transaction, we've averaged just over $5 million a quarter of underlying EBITDA growth. 100% of that growth has come through cost cutting.
We probably have another couple of quarters of cost cutting initiatives that will help us, but we must return the top line growth, and we must return the growth from high margin products. Whether we sell transit, which is 88% on net, wavelengths, which are 100% on net, or IPv4 addresses, which don't use the network, and all of those carry very high contribution margins, we will be able to delever. While we don't like to give guidance, we said for this one time, if you take where we were at peak in Q2 of 2005, by the end of 2026 or six quarters from now, we will delever down to about 5x . That's still above our comfort level. We had hovered around 3x leverage for a full decade. When the pandemic hit, our leverage spiked to 4.2x.
When we initially did deal with T-Mobile because of the front-end loading into payments, we delevered down to 2.5x. When those payments stepped down, our leverage has ratcheted back up. We've been able to mitigate some of that through the underlying EBITDA improvement.
Just back on the data center, it's good to hear that if you didn't sell one, but you very well could. Can you provide more color on the new LOIs? You had four LOIs, and then you added two more.
Two more.
Are those two more same, better, or worse than the preceding four? Do they support the $10 million valuation?
One did, one didn't. One was a lease, and one was a buy. I would say the willingness to put real, meaningful money at risk is no better in these two LOIs than the four previous ones.
Can you talk about that, the financing? These are parties that are interested, but there's no committed financing in place.
Sometimes there's committed financing, but not the willingness to use it until they get revenue.
That's what I was thinking. Because some of these parties from Artex are big private equity firms, what's really stopping them is looking for.
A tenant.
Yeah, getting a tenant in place. OK. Any updates around the process itself? Last month, we spoke with your Head of Data Center Sales, with your permission, so thank you.
Greg gave us his name out. I want anybody to call him who's interested.
Yeah, it was a great call. He noted the process is like rolling admissions. Can you help us with how long the rolling admissions window will be open, if you will?
You know, until we get a real offer that is binding, we're motivated to sell them. You don't sell them if you close the window to buyers.
Yes. Earlier, we discussed the company's liquidity position, juggling shareholder returns and deleveraging. I guess you noted that you're good in terms of a dividend and honoring that and the deleveraging. How do I think about deleveraging from this perspective? You said it really depends on growing your waves business from here, right, for the most part? Net can grow, but waves is the 100% margin. That will really help you in terms of.
Incremental IP leasing is also important, as is incremental transit. I think we're going to have growth from all three. I think in percentage terms, the wave is definitely going to be the fastest grower. In absolute terms, because for transit we're already at 25% of the market, it's much harder to go from 25% to 30% than it is in waves, where we're at 1% of the market and we want to go to 25%.
You mentioned IPv4 address leasing too. That's all margin, if you will. How about selling them outright still? You're sitting on millions of those that you can potentially sell. You're waiting for the right pricing environment, is that right? Typically, are you waiting for AWS and Microsoft to come back into the market? What are the prices today, and when do you think they'll come back in? I know it's tough to.
You should ask them that question. Although they are generating billions of dollars of EBITDA off of the addresses that they have bought, I think there's three pieces. One, I am not sure the market is deep enough to take the size inventory that we would bring to market. If you look at all of the transactions that have occurred, they're for much smaller volumes than we would be bringing to market. Even though there's a liquid market, it may not be big enough to absorb enough that would mean anything to Cogent. Two, we currently are leasing about 15 million out. We've got about 2 million that we've given away for free for historical purposes. That leaves 21 million that are fallow. We would be happy to sell a significant portion of those if we could find a buyer at a reasonable price and take a reasonable volume.
The major volume buyers have historically been those two names you mentioned. There are two public exchanges where addresses are bought and sold. You go look at them just like you look at your stock ticker. You'll see trades, but they're of low volume. I think we are motivated to delever more quickly. It is not fundamental to the value thesis. If we sold addresses, if we sold data centers, it is meaningful on our balance sheet, but it is a one-time event. The real value creation is the ability to produce recurring free cash flow. That is where we're going to build value. Within the assets we acquired from Sprint, there were two that could do that. Wavelengths are a business we understand and can do that. I don't believe Cogent has the wherewithal, people, or expertise to fill up all of this data center space.
It's just not a core strategy, and our sales force doesn't possess those skills. That's why we concluded it should be in the hands of someone else. What we're focused on is building recurring free cash flow growth.
If I hear you correctly, it sounds like you have a stronger propensity to lease the IPv4 addresses than sell them outright?
Sell them outright if there was a buyer for volume at a reasonable price. Today the market's in the low $40.
Do you have a reserve price? I mean, in general, when you were selling the last time, wasn't it like the $50?
I would probably sell in that zip code. I don't want to price out.
You can't sell $40 with the bulk that you have, you're saying.
I would sell if it was meaningful. I don't want to sell $500,000 at that. Okay, and generate $20 million. That's kind of like, OK, that's a yawn.
Right. OK. Moving on to Netcentric, finally talking about your legacy businesses.
Which is only 88% of revenue.
Yeah, 35 minutes into this. Your traffic grew 9% year over year in the second quarter, and it grew 8% in the first quarter. That's a stark contrast from your typical, you know, call it 20% growth in traffic. That's two quarters in a row. Is this rate sort of the new normal for now? I know AI, we could talk about that, augmenting it, but that's not happening yet. Help us with the traffic growth or the slowdown now, and how long the traffic growth will remain here, and what could pick that up?
I partially answered that question already, in that we are 25% of the market. It becomes harder to outpace the market growth significantly as your percentage of market share grows.
The law of large numbers.
Not large numbers, but market share.
Yeah, got it.
Two, the market itself has slowed down. If you look at, for example, OpenVault data, it's 7% growth. That's looking at the world from the download of the customer. If you look at things like Telegeography and the Visual Index, they're also triangulating to about 7% traffic growth. The Internet is not done growing. It's gone through waves of growth, and I think there's at least one more wave to ride, probably many more. I think we will gain share. Our revenue growth has been better in part because of the mix of locations and customers generating the incremental demand. If we sell a lot of IP transit to a large hyperscaler in North America, we get the very lowest price per bit. If we sell a much smaller amount of bandwidth to a regional access player in Cambodia, we get a much higher price per bit.
What we have seen is a gradual rotation away from the U.S. and Western Europe to the rest of the world, and that has been helpful to our revenue growth. We've also seen a market rotation now towards some smaller buyers. We've had quarters where all the growth came from a couple of large buyers. That's not as true now.
Right. You've got maybe larger buyers, but you have the internationalization of the internet, if you will. How long do you think that trend will continue where you had that pricing power before, even they have streaming in Cambodia?
For example, we're going to be entering the Indian market later this quarter. It only took us eight years to get a license. You know, it's going to be eight years before someone comes behind us. If you look at the other U.S.-based dominant players, they're actually contracting. I heard one player at lunch say he sold his European business. He spun it out. I know another player has sold Latin America and Europe and never really invested in Asia. Cogent is operating in 57 countries, 302 markets, and 1,870 data centers. That is a far bigger footprint than anyone else. It just gives us more addressable market.
Right. With the last minute, I wanted to flip over to corporate. On net adds for buildings, you added 18 on-net buildings in the second quarter. I think it's the lowest second quarter I've looked at in the past 20 years or so. Obviously, you're doing the CapEx to refurb your data centers. I'm thinking that's the answer to this question. Can you elaborate on this slowdown? Is this the cadence or the new normal as folks are laser-focused on your second, third quarter, and fourth quarter CapEx right now?
The CapEx and principal payments on capital leases will come down to an annualized run rate of about $40 million on capital leases, $100 million on CapEx. We have definitely slowed down the rate of multi-tenant office building additions. We're in 1 billion, 170 million sq ft for two reasons. One, the impact of the pandemic. Two, we're in most of the buildings we want to be in. We are still adding third-party data centers. We have slowed down the rate of new country and new market expansion. We are adding them, but at a slower pace. Third, we did put a lot of capital into connecting the Sprint network to the Cogent network, reconfiguring, and then optimizing that footprint both for waves and for IP. Ultimately, we've got to deploy capital where the ROIC is substantially above our cost of capital.
Got it. With that, we're just about out of time. Thank you, Dave.
Thank you very much. I did get it done in time.