Cogent Communications Holdings, Inc. (CCOI) Earnings Call Transcript & Summary

Thank you. Good afternoon. So I am the telco high-yield or lev fin, rather, analyst at Deutsche Bank. I've been here a little over a year, and this is my second year with Dave. He was gracious enough to be here last year, and he's repeating his attendance again, and we're grateful for that. Dave Schaeffer, CEO of Cogent Communications. We will have a sort of standard fireside chat. I'll lead off with some questions and a discussion, broadly, and then at the back end, leave some time for questions for you guys. Dave, is there anything you want to say at the outset? Or...

Well, I'd like to thank you for hosting me today. I'd like to thank the investors, Deutsche Bank, for a great venue. I was kind of sad to hear it's going away after this year. They were going to a new home. But again, I want to thank everyone for taking the time and the interest in Cogent, and I'm here to answer your questions.

Fantastic, fantastic. New home new location, but Dave Schaeffer will be there, hopefully.

I hope so.

Great. So I thought we'd sort of kick off because, Dave, you're obviously very visible with investors, and you do a fantastic job communicating the story. But I think maybe more on the debt side, the 2 deals you've done and particularly with the IPV4 securitization notes that were put in, I was hoping you could maybe walk through the structural changes, the sort of formation of Communications Infra and Cogent Communications Group and sort of the transfer of assets and maybe just spend a few minutes on kind of describing how that sort of story unfolded.

Yes, absolutely. Happy to. And a full disclosure, Deutsche Bank was one of our underwriters for that transaction. So Cogent has had a 2-tiered capital structure: a holding company, which is a publicly listed company that issues a dividend; and an operating company, Cogent Communications Group. All of our debt had been held at Cogent Communications Group level. There are traps to restrict cash and allow for the upflow of cash from the operating company, the holding company, which is then freely usable for either dividends or for share buybacks. We have tried to be transparent with our debt investors since 2010, when we did our first debt offering, that the primary purpose of our debt offerings is to facilitate return of capital to shareholders. Cogent has grown its dividend for 48 sequential quarters consecutively. We have also purchased about 23% of our float back in the open market. Now that's over a 15-year period. Last summer, when we completed the acquisition of Sprint from T-Mobile, we knew that we acquired an uneconomic IRU that Sprint had entered into and had been consistently appreciating. The counterparty to that IRU was Verizon. We had the opportunity to buy out of that IRU at a 12% discount per year, and it had 3 years remaining. It had a current cash payment of approximately $49 million a year for the next 3 years. We initially had planned to come to the high-yield market, but at that time, market conditions were not conducive to launching a new offering. We then met with a number of banks around asset-backed securitizations. We were first pitched on the idea of doing an asset-backed securitization against our fiber network. We concluded that, that was not practical. Based on the geographic span of the network, it would just be too daunting of a task to try to perfect security interest in those right of ways. And then looking at portions of our business, we had a very small portion of our business, approximately 4% of revenues, that were coming from leasing IP addresses out. So maybe a bit of background for generalists on what that means in plain in English. When the Internet was created, there were 3 foundational protocols that allow the Internet to work. The first is TCP/IP, the protocol that allows 2 computers to talk to one another. The second protocol is BGP, or Border Gateway Protocol, that allows 2 networks to communicate with each other. And the third is a unique numbering scheme. It was ultimately settled that there would be 2 to the 32nd power or approximately 4.3 million addresses that would be issued for the Internet. And this would have been done in the early 1990s, actually late 1980s, when everyone thought that would be way more than it could ever be used. Well, as it turns out, those addresses became a scarce resource, and they now have monetary value. Addresses are routinely bought and sold on open exchanges. In 2011, the first addresses were sold at about $4 per address. Addresses peaked in 2023 at about $60, and today, routinely trade for the low 50s. Cogent has 37.9 million addresses, probably the third-largest holder of addresses in the world. We began leasing out those addresses in 2015, and had leased approximately 12.5 million of our 37.9 million addresses out. We wanted to place those addresses into a subsidiary that could raise asset-backed funding. In order to facilitate that, we injected additional addresses from outside of the credit group, approximately 9.9 million addresses, that we had acquired from Sprint. We gave the operating company a blanket IRU on fiber across the Sprint network, and then finally, a master lease on a number of data centers. We had a third-party appraisal done, and that appraisal showed that there was approximately $365 million of incremental collateral going into Cogent Group from Cogent Infrastructure, the entity that owns the Sprint assets. So Cogent Infrastructure then raised ABS funding of $206 million against the 8,000 customer relationships that we have and nearly 12,000 contracts, representing about $35 million at the time in EBITDA. And then that cash was then available to allow us, in conjunction with other cash, to buy out that unfavorable IRU that was bought out at a 12% discount, saving the company nearly $40 million in annual cash payments.

Appreciate that. I think the structure -- just one of the other aspects of it is that the cash flow stream also ended up benefiting -- there were some losses that were transferred, I'm guessing as a result of the IRU that you mentioned, between Group and Infrastructure. And therefore, in the most technical sense, the bond issuer is at sort of more of like a $600 million LTM EBITDA positive. Is that correct?

That is directionally correct. So ensuring that when we acquired Sprint, we would not disadvantage our bondholders, knowing that we would have to come back to the credit markets on additional financings, we made sure that the subsidy payments from T-Mobile of $700 million in cash went into Group, the borrower, and we excluded some of the losses associated with the operation of the network, which sat at the sister company level, so adjacent to Group, Cogent Infrastructure, under Holdings. Infrastructure had no operations. It held assets that it purchased for $1, and it absorbed some of the operating losses. So we were careful, again, to make sure that the bondholders would recognize that we were not diluting their collateral base.

Right. And that's a nuance that just, I think, for the benefit of the audience, one thing we've seen is that the rating agencies still continue to view leverage on a consolidated basis, but in theory, you have enhanced the interest coverage and the leverage profile at the bond issuer as a result of the transfer of the losses.

Well, it actually was a bit surprising to us that when we raised incremental debt in each offering, our ratings actually were increased, and it was because of the enhancement of the credit quality of the borrower.

Okay. You did get a bump, I think, from Moody's and -- both from Moody's and S&P.

Both Moody's and S&P. Both raised their ratings on Cogent as a result of these transactions.

But perhaps not commensurate with the actual cash flow that kind of resulted at Group.

I think that's probably right, but we were not overly aggressive in trying to lobby for some additional enhancement to our ratings. We were pretty comfortable that the ratings that we have reflect appropriately the nature of the business.

Okay. Fantastic. Just in terms of like sort of reporting going forward, will we, on a quarterly basis, see Infrastructure revenue and EBITDA and Cogent Group revenue and EBITDA separately in either press releases or filings?

They will be separated in the footnotes, but not necessarily in the press release, which is related to the holding company. So the holding company has both Group and Infrastructure. In Infrastructure today, the only operations are the assumed losses as well as the ABS revenues. And then in Group, all the operations exist independently. We are required, as part of the asset-backed securitization, to annually produce an audited financial for that entity, so that will be also footnoted in our annual filings or K, but probably only briefly described in the Qs.

Okay. Great. And that's actually a good segue, just talking about the losses that you inherited from Sprint, right, regardless of just kind of ignoring the boxes and the fact that the losses were -- some or maybe all the losses were transferred into Infrastructure and...

Some were.

Some were. On a consolidated basis, when you acquired the Sprint assets, you had said that the run rate at that time was about $180 million negative EBITDA.

$190 million actually, at closing, and that corresponds to the last filing that T-Mobile did. So when we began the process of acquiring Sprint Global Markets Group, the burn rate of the business was actually negative $300 million of EBITDA and a revenue stream of $565 million. We identified a number of noncore products that needed to be end-of-life, certain other customer relationships that needed to be terminated and employees that needed to be separated from the business. During that intervening period between signing and closing, T-Mobile continued to both support the burn and implement a number of cost savings objectives. That resulted in the burn rate going from negative $300 million to negative $190 million at closing. Revenue was $490 million at closing, and T-Mobile's expenditures between signing and closing on this business were approximately $500 million. Now subsequent to closing, T-Mobile is supporting us with a $700 million cash payment. $350 million of that is in the first 12 months. The second $350 million of payments are spread out over the subsequent 42 months. Those monthly subsidies stepped down from $29.2 million to $8.3 million a month in June of '24. In addition to those 2 streams of payments, there is also a blanket $100 million indemnification from T-Mobile for any unforeseen liabilities, and then finally, there was a pool of $25 million available for employee severance. We did, in fact, use that pool, and it was fully funded by T-Mobile. It actually was funded to the tune of about $28 million. The extra $3 million were some retention payments that T-Mobile had given certain executives to stay on through the transaction. But T-Mobile did fully fund those. Now they did hit Cogent's EBITDA as an expense and then were added back, adding to some of the accounting complexity related to the transaction. It's highly unusual to buy an asset that had a cost basis of some $20.5 billion for $1, and then buy an operating business that then you're paid $700 million to take. So it did end up creating a little more accounting complexity than we would have liked.

Fair enough. And of course, obviously, the $700 million is meant to offset the -- you inheriting the negative cash flow stream. So sort of sticking with that and just ignoring the IP Transit payments for the moment, at the time when you closed the deal last year -- in fact, when we spoke a year ago, you had said that you would get that $190 million negative EBITDA down to $80 million within a year and 0 within 2 years. Where are you on that time line? So what do you think the run rate losses are today? And when can you get that?

We are actually slightly ahead of plan. We identified $220 million of cost saving synergies that we would achieve over a 36-month period. In the first 15 months of closing, we have achieved 62% of those savings, and we have got the annualized burn rate below $80 million as of last quarter. And we anticipate by the time we report Q2 of '25, the burn rate of the acquired business will be down to 0.

Okay. Now if -- let me take a slightly different sort of approach to thinking about the Sprint acquisition, sort of the glass half empty bondholder perspective. Let's just say there's no upside to wavelengths, right? Clearly, we know why you bought this strategically. There's a very strong equity story behind buying an asset that has a $20 billion total invested capital base. You bought it for $1. You have T-Mobile that's agreed, from a financial perspective, to make these IP Transit payments to kind of gradually work your way out of the losses that you've inherited. And if, as a bondholder, you look at the last full year EBITDA that you reported, so it was $236 million in 2022, if I'm not mistaken.

Correct.

And if there's no sort of wave business coming behind this, there is some element of the Sprint business, I think you've talked about, that the $490 million in revenue that you inherited, you're obviously trying to work your way down out of unprofitable contracts. What -- if you ignored the wavelength business, what kind of base Sprint off-net business would you retain? And at what margin?

Yes. So the acquired business came with the transit payments that you described. Those transit payments allowed our total EBITDA for '23 to be $352 million.

I want to keep the transit payment calculation out of it. Just kind of your downside...

Well, transit payment is a contractual obligation from an investment-grade issuer.

Understood, but I'm trying to get to the actual cash flow capability of the business you acquired.

Right. So the business that we acquired is a shrinking business. Part of that reduction in revenue is as a direct result of our elimination of gross margin negative noncore products, the elimination of certain customer locations and the elimination of some international markets where we and Sprint did not have appropriate licenses. As a result, we had initially expected that business to stabilize at somewhere around $440 million to $450 million. It will probably stabilize below that due to the need to prune these unprofitable access customers. The margins associated with that business were anticipated to stabilize at about a positive 20% EBITDA margin or roughly about $85 million to $90 million in EBITDA, and that would be within 4 years of the transaction closing. While the revenue will probably be below that due to the fact that there were these less profitable access customers, we actually will still achieve the same level of EBITDA. So we anticipate that based on current run rates, we're probably closer to $400 million of stabilized revenue, not the $440 million that we had projected when we announced the deal. But we also anticipate that the margins on that revenue will probably be higher at around 22% or 23% as opposed to...

Low 20s, [indiscernible] $80 million to $90 million on a $400 million basis.

Right. So we're right on track for the EBITDA from this transaction. And while I know you said you wanted to exclude the transfer payments, the way we thought about it in derisking this transaction was that we would consume about $400 million of the $700 million in subsidies to fund the losses and fund the efforts to fix that business to the best extent it could be fixed. It is still not nearly as robust a business as Cogent's legacy business, which has organic growth and superior margins. But if you just looked at it on a stand-alone basis, ascribing 0 value to the physical network that we acquired and saying we did not produce any wavelength revenue, the combined company would still end up with $300 million of net incremental cash that it did not have, and it would end up with between $80 million and $90 million of incremental EBITDA it did not have prior to the transaction. Now what makes this complicated is we have combined our billing systems, our customer care systems, our network management systems, so it is difficult break out which customer came from where, and it's even more complicated by the fact that we have migrated much of the off-net revenue to on-net. It is one of the key ways that we have achieved the improvements in margin. But in trying to minimize risk, we clearly looked at 2 discrete transactions, the first of which is picking up the operating business, its associated losses, in a subsidy payment, and then the ability to turn those profitable. The second is taking the fallow assets, and the assets that were acquired were quantifiable: 19,000 route miles of intercity fiber owned, another 1,200 route miles of metropolitan fiber owned, a 3,800 mile bad IRU that we had to buy out of, and 482 fee simple owned buildings that comprise 230 megawatts of power and 1.9 million square feet. Now many of these buildings are either too remote or too small to have much conversion potential, but we have identified about 100 megawatts of existing power and 1 million square feet that are suitable for data center conversion, and we've begun marketing those facilities. The final asset that physically transferred were an additional 9.9 million IPV4 addresses, so our base went from 28 million to 37.9 million.

Right. That's very helpful. So -- I think you sort of answered my question, sort of thinking at a run rate -- excluding the wavelength business, excluding all these ancillary assets, which obviously have significant value, but ignoring all that for the moment and looking at just the pure cash flow capacity and ignoring growth for a minute, it's not unreasonable to tack on, let's say, $90 million of EBITDA, which would take 4 years to get to. On the Sprint business...

And $300 million of cash.

And $300 million of cash, but that's sort of in aggregate, right, the $300 million.

That's correct.

But I'm talking about coming to a run rate number, so $236 million, you reported in 2022. And I understand I'm looking at it apples and oranges, but let's call it mid-300s EBITDA, if you did not get any upside from wavelengths or anything else plus the $300 million in cash.

That's correct. That's correct, but it did retard our growth rate because we picked up $400 million of nongrowing revenues and attached those to a $600 million revenue business that had grown at a compounded rate since going public of 10.2% a year.

Fair enough. Fair enough. That said, the mid-300s EBITDA gets high-yield investors excited and interested. Maybe not equity investors, but it's kind of framing the downside, if you will, to some extent. Now obviously, the story is much more complicated, so I think this is a good segue to go into how the integration is going. Let's start with the physical side. You talk about what you need to prepare the -- what you need to do to prepare the network for wavelengths, and -- can you sort of give us an update, just purely not -- excluding the numbers for a moment, what the physical elements are that you have needed to go through over the last year to get the wavelength opportunity and hand it off to the sales team, if you will.

Yes. So the Sprint network was built to carry long-distance voice. It terminated typically at a Sprint tandem switch site 15 miles from downtown, including a facility here in Phoenix, about 15 miles from downtown. That facility was only interconnected to the regional Bell company with a lit circuit to terminate voice calls. What we have done is physically extended the Sprint network to touch our metropolitan network in approximately 110 markets. That physical extension was completed in February of '24. The second thing that has to happen physically is that these buildings were full of old telephones, which is we have to remove that old telephone gear. There were 22,500 cabinets of telephone equipment in these buildings that was obsolete and not in service. We have removed a little over 16,000 as of the end of the quarter. The third thing that we had to do is reconfigure our metropolitan networks in order to accept wavelengths. So while we had an extensive 18,600-mile global metro network, about 13,000 of those miles were in North America. They connected to 800 carrier-neutral data centers in North America and roughly 1,870 skyscrapers, and we needed to segregate the skyscrapers from the multi-tenant office buildings, and reconfiguring those networks and then direct wavelengths solely to the data centers. We are about 3/4 of the way through that physical reconfiguration. The fourth thing that we had to do was we have to deploy a transponder shelf, literally a 2RU, so 3-inch tall unit, that you put in each rack of a data center to allow customers to plug wavelengths in when they buy from you, and we're about 3/4 of the way through that deployment across those 800 facilities. And then finally, we need to deploy about 110 reconfigurable add-drop multiplexers that would direct the wavelengths to the correct data center. We are also about 75% of the way through that project. By the end of 2024, so in 3 months and a week, we will be done with this reconfiguration. The outcome of that reconfiguration will allow us to provision a wavelength from any North American data center to any North American data center in 2 weeks with only 2 field deployments. It will allow us to support 10 gig, 100 gig and 400 gig wavelengths. The permutation of possible combinations is truly astronomical. It's 800 minus 1, so 799 divided -- factorial divided by 2. I mean that's trillions of possible combinations. You can never pre-provision all that, but you could build an infrastructure that then allows the field team to just deploy inexpensive pluggable optics to then turn up each wavelength as customers require them.

Great. I think we're sort of -- I have a lot more questions to ask, but I think I want to open it up to investors. Any questions from the audience? Okay. Well, I'll stick to -- so you've kind of explained where you are on the physical integration in order to turn this into an opportunity, and you've already been recording wavelength sales are at a very small level.

About 800 wavelengths have been sold through the end of last quarter. We have about 2,700 in a funnel backlog.

And so how do you see that playing out over the next couple of years in '24 into '25 and '26?

So we will continue to build the backlog. We will custom deploy only a handful of wavelengths, because until this reconfiguration work is done, it is a very manual process. Once that work is done, we should be in a position with our current resources to deploy about 500 wavelengths per month out of the backlog and continue to sell. If we are more successful in that, we can add additional field resources and provisioning teams and scale that further. But the goal is within 5 years of closing, so that would be run rate of May of '28, we would be doing $500 million of wavelength business. And because that wavelength revenue is on-net, it carries a 100% gross margin contribution and $0.95 of incremental EBITDA.

Fantastic. Unfortunately, we are out of time, and I could continue, Dave, this conversation for another hour, but we do have to leave it there. Again, I thank you for attending our conference.

Thank you all very much.

Talk to you next year. Thank you.