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

So I think you can kind of just kick it off. I guess, Dave, again, congrats on the smooth transaction. What's going right? What's going wrong?

Well, fair enough. First of all, I'd like to thank you for hosting me today. Thanks BofA. I thank the investors for taking the time to meet with us. And we've been paid a compliment in the beginning saying it's a shiny new asset. Oftentimes, one of the alleged issues with the Sprint network is its age. And like some things, old does not necessarily mean bad. And in the case of the Sprint network, we've been very happy with the quality of the assets that we acquired. We, in fact, did an appraisal on the fiscal assets, that's 485 buildings as well as the 19,000 route miles of intercity fiber, 1,200 route miles of metropolitan fiber, that came in at approximately $1 billion. We paid $1 billion for that asset. So I think we did well. I think -- what's going right? We definitely are happy with the asset. We're happy with people that we acquired. What is not going as well as we would like, I think data transferred over from T-Mobile onto our systems was a slower process, in part because some of the cyber security concerns that T-Mobile has had because of some of its brokers. I think a second area of concern for us is the fact that the demand for buying services actually turned out to be stronger than we anticipated, which is great, and our ability to provision is not where we wanted it to be. Just to remind investors, we are taking an asset that is effectively a fine asset that was originally bought for long distance voice. It was the nation's first transcontinental fiber optic network built in the late 1980s for a cost of about $20 billion to terminate in 45 tandem switch offices and that network has nearly have been actively utilized from nearly 20 years [Technical Difficulty] switch facilities and repurposing for physical network to sell optical transport services or wavelengths and in order to do that, we need to connect those kind of locations back to our metro networks, and we need to reconfigure those so we can sell wavelength services in 800 carrier-neutral centers across the U.S. It's very recently provision in about 35 facilities with a 60-day provisioning window and through provision in about another 210 facilities with more like a 120-day provisioning window. Our internet goal is to be able to provision in 17 days or less. So we have a lot of work to do. So the good news is more demand than we expected. The bad is we're not filling that demand as quickly as we'd like.

So let's [Technical Difficulty] to the things that are going right. So one of the things I think that when we look backwards your track record in M&A and this is actually, I think, a slide that you put up in the very early phases of the Sprint transaction announcement, maybe you were on this conference in Boca. There was a slide that showed [Technical Difficulty] of the company you bought it. And how many people were left inside the organization when you were done integrating it? And this transaction seems to be a bit of an anomaly because you keep waxing according to how great the people are and how long their tenure has been. And it seems like you might be getting a little soft on us. So can you tells us a little bit about what's going on with the opportunity to create 3G through head count reduction?

Yes. So there's a lot of areas to create synergy headcount reduction as one of those areas. Again, just to remind investors, this business at peak had [indiscernible] employees. When we started working with the business, there were 1,800 still in the business. During the time we began negotiations with T-Mobile and signed a deal [Technical Difficulty] the business, the head count was reduced to 1,320. Between signing and closing, the head count was further reduced from 1,320 to 952 at closing. We anticipate probably about another 150 employees with the company as a result. I also think composition of the employees is going to change. Prior to the acquisitions, Sprint had 950 employees and only 51 literally were in sales. And I put that qualifier in because none of those 51 had a actual quarter, we then had actually sold new customers for nearly 15 years. And as customer relationship managers, they have existing relationships that they manage. There were 1,396 customers, that is individuals formed, but we didn't go out and hunt for new business. Compare that to Cogent to the acquisition, we had 1,100 employees, 150 had direct quarter responsibility, another 200 in sales support roles, so 750 and only 250 in operations. So some of the things we've done, some of the sales people have left the organization that came from Sprint. Some of them reassigned to operational roles and we actually continued to hire salespeople. Now we today have over 650 quota-bearing reps, and that's still not enough out of a total employee population of about 2,000 employees. So our goal will be to continue to create value out of the network and operations teams in Sprint to sales organization. And lot of the individuals will have quite the same responsibility. A criticism of Cogent has been all transactional sales model, we typically turn about 5.2% of our sales force a month, I don't see that materially changing. We've a quota, it's black and white of a ramp. And at the end of that ramp, they either make their numbers, but that's just not the right way for them, we have softened our criteria.

Just on maybe on the topic, we're generally -- one of the way we kind of model the company really is hiring, train quota-bearing sales reps productivity for rep and obviously, that was a struggle. Even in the aftermath of the pandemic, people still had kind of cushion and there wasn't really a huge amount of enthusiasm for work. What does the funnel look like for you for inbound sales, trainees, conversion to quota bearing? And then how are you working on productivity at the margin?

So I think if you ask employees, they've had an appetite for work, but not an appetite to go to the office of work. I actually disagree with that. I tend to think you stayed at home, you probably were not as productive. So in the pandemic, we kept hiring at the pre-pandemic levels, but our sales force turnover because of the strict discipline around productivity actually increased from 5.2% a month to 8.7% at peak and we backed down last quarter to 5.4%. So pretty much to pre-pandemic levels in terms of turnover. Our model is to continue to hire people and grow the sales force on an aggregate basis of somewhere around 7% per year, number of quota-bearing reps. We have a very formalized training program that lasts a month. There's two more months of a ramp and then you're expected to be selling. Now sales is a process. And in any process, you need to go through the steps. The most important step here is building the funnel and that means having enough legitimate opportunities to work. Our sales force that is successful have adequate funnels, those who fail, don't do the activities. And to be successful at Cogent means to do high-volume outbound selling. We expect a new rep to do 100 cold calls a day. Now the majority of those cold calls don't result into sale ever, and many of them result in delayed sales. What we have seen over the past 6 months is a gradual improvement in sales cycles shortening. So our corporate growth rate actually has continued to improve and is positive. So to remind investors, prior to the pandemic, our corporate business where we are selling directly to end users was growing at 11% year-over-year, and it had that average growth rate for 15 consecutive years, with only 2 negative quarters during the great financial crisis. Pandemic hit and the growth rate went from positive 11% to negative 9%. Today, we're back to a 1% growth rate year-over-year in that business and improving. The second part of Cogent's organic business was its NetCentric business. That had average 9% growth prior to the pandemic. Going into the pandemic, it was actually growing below trend line at 3%. Pandemic hit and we actually had the best growth rate in our corporate history. Our growth rate shot up organically to 26% year-over-year. This was 44% of our revenues growing at 26% a year, great result, it was a result of a rapid transition from linear television to streaming. That growth rate has moderated some, but is still low double digits, around 10%, 11% and we anticipate that NetCentric growth rate to remain above trend line for the foreseeable future. The pull-forward in streaming that occurred in the U.S. is now occurring in the rest of the world on a delayed basis. So our NetCentric business is continuing to outperform historic trends. And our corporate business is improving, but at a slow pace. The acquired enterprise business that came with Sprint, the second asset that we bought was a business that was burning nearly $1 million a day. So when we consummated the transaction, there was $560 million of run rate, negative $300 million of EBITDA and $30 million of CapEx and negative $330 million of cash flow on a $560 million business. The way to fix that is jettison gross margin negative products. We have done that to rightsize the sales force and most importantly, optimize the network. We believe we will stabilize that business at between $440 million and $450 million in revenue, and we'll get it to be a 20% positive margin business by applying a disciplined strategy around products, selling 4 basic products; Internet access, VPN services, colocation and optical transport services. But the real upside and growth driver going forward will be the ability to take the Sprint network and sell optical transport services. We think we will be able to grow that from its current annualized run rate at closing of $8 million a year to a $700 million run rate within 7 years, fairly linearly. We will report transparently to investors, number of wavelengths, amount of revenue. We are seeing more demand than we expected. But as I said, the installations are longer. We should even this out and should be able to hit a $100 million run rate within roughly 14, 15 months of closing $80 million in the first year.

So that was the answer to the hiring question?

That was it. There were a lot of answers with why we do need to continue to hire salespeople, and we don't need to hire more operations people.

And so is the hiring of salespeople, is that really a conversion of the Sprint bodies from what they're doing today into salespeople or replacing those bodies with new blood that's sales-people centric?

There have been -- of the 51 people that came in the sales -- with a sales title from Sprint. The majority of them are no longer in sales at Cogent. Some have left the company, some have been converted to operational roles. So there are probably about low 20s number that are still here that I think are embracing our sales model, but most of our sales growth is coming from hiring new candidates from the outside.

Okay. So the other thing that you said was working with the merger was the asset. At the same time, you said that the connection from the 45 nodes back to your metro network was the chokepoint to enabling wavelength sales, among other things. And that, that was going far more slowly than expected. Can you elaborate a little bit on what the chokepoint is?

So I'm not sure it's going slower than what we expected. What is frustrating is the demand has been better than we expected. So we have more wavelength orders that we cannot fulfill. That's kind of a unique experience. Normally, we're out creating demand and chasing customers, we're actually getting orders that we have to tell customers on a delayed install. So to convert the Sprint network into a wavelength, sellable network, 4 things have to happen. The first you had mentioned is the physical connection of the Sprint backbone to a metro footprint. We're probably about 70% of the way through that. We expect to get all of those connections done probably by the middle of first quarter of next year. The second thing that has to happen is our metro networks need to be optimized for wavelength sales. So today, we have roughly about 1,150 physical rings around the world sitting on about 20,000 route miles of metro fiber. In the U.S., each of these rings has a mix of multi-tenant office buildings and data centers. To optimize for wavelength delivery, we need to segregate those 2 types of buildings on to separate rings. This is a span by span effort that's underway. We, I think, have I think as of yesterday, there were 136 of those U.S. rings in process of being optimized. We will get all of that done between now and the end of next year. The third thing that has to happen is we need to put transponder shelves in all of the carrier-neutral data centers that we are connecting. That allows us to then provision a wavelength by just plugging in a transponder at each end. I know one of our competitors in response to our provisioning goals gave a statement at a conference saying they're going to have instantaneously provisionable waves. Well, the only way you can do that is pre-provision everything. You still need to go plug in something. If you've got a pre pugged in, you're going to be very capital inefficient and not optimize your network. So the goal is to create the base structure that is quickly then upgraded on an as-demand needed basis. The final thing we need to be able to do is optimize these routes for where the wavelengths are regened and where they are dropped. And to do that, there are really 2 different architectures. They're not unique -- one or the other in each market, you have to work on a case-by-case basis. But all 4 of these efforts are being simultaneously done. And again, we're going to report every quarter, a wavelength-enabled wavelength provisioning times because of the high contribution margins and this being the primary reason for doing the transaction, it's what we are going to be measured on.

So the -- this linear ramp from kind of $2 million revenues run rate in the second quarter would suggest that we're going to be ramping. I can't remember the math, but it's like you have to kind of go 2, 4, 6, 8, 10 kind of progression. Are some of the provisioning delays that you're seeing impeding that result? Are some of those -- is that linearity contingent on these issues being addressed? Or is your linearity assuming the current plan of addressing all these issues and so therefore, we're not impeding the rate of growth in wavelengths right now?

It's much more of the latter. So on the good news side, the number of wavelength orders that we've received is greater than we anticipated this time. So we have a bigger pool. On the bad news side, we are slower on provisioning than we would like to be, but we are where we expected to be. So if we are successful in bringing down the provisioning times and demand remains as robust as it is, we will actually exceed our goals much more quickly than what we have outlined. What I am not comfortable in representing to investors is how durable that excess demand is and it is a little hard to tell, is this because there's a new entrant in the market and people are frustrated with the current suppliers, is it because we really do have unique routes that are better, and that's why people are coming to us. Or is it because there are new applications that are driving incremental demand. So what I'm trying to do is be maybe a little more conservative on the demand side because that's a little bit outside of our control. And where I'm comfortable we'll meet our objectives is on the supply side and having the ability to have 800 facilities with 2-week provisioning windows.

By when?

By the end of '24.

So let's talk -- are we talking about twice as much wavelength orders, 10% more wavelength orders and order of magnitude more wavelength orders than you expected? Like what are we talking about here?

I would say, at 4 months into the transaction, we probably have a backlog that's over double where we anticipated it being at this point. But again, I don't want to get people too excited, I can't confirm that, that's all demand -- just because Cogent is so great or it's just people frustrated with their current supplier.

I mean, it's not unusual for anyone who kind of jumps into a new market with a new product, but there's 10% of the market hates who their provider is. And if that's -- if there's a, I don't know, multiyear contract cycle that exists 2, 3 -- I mean, so right away, 3% of the $2 billion wavelengths market might be interested in being at your doorstep asking, please let me out of this existing relationship with this guy. So who is it? You -- when you described the pool of demand for wavelengths, you kind of said there's 3 big buckets. There's the hyperscalers, there's the -- for lack of a better term, the ISPs, and then kind of large enterprises that might include maybe some of the large streamers and those sorts of guys?

Organizations like Bank of America.

And Bank of America. So who -- where is the demand coming from? Is it pretty much just -- or rather everywhere? Or is it one camp seems to be the driver of the margin?

I would say, 2 of the 3 are driving the bulk of that demand. That being, the content generators, primarily hyperscalers but also some CDN operators. So companies -- hosting companies that are pushing applications or content out. The second large group are the regional ISPs pulling down traffic and connecting their islands of demand. The enterprise market is the smallest of the 3, and that's what we've continued to see. And there's actually a fourth pool that we were not anticipating in closure, which is a new group of buyers of service and I almost hate to use this term because it's so promotional, but it really is the correct one, which is AI related companies that are not traditional hyperscalers but are looking to do large amounts of computation at multiple sites. So the Internet remains the cheapest, most reliable and most ubiquitous way to move information. There is a significant market for people willing to pay a premium for dedicated transport services for 3 reasons: security; deterministic, meaning you know exactly how long it's going to take from point A to point B; and third, the ability to move very chunky or large amounts of data. The Internet does very well with packetized data and small file transfers. If you're trying to replicate the entire content of a 1 million square foot data center between 2 locations, doing it over the Internet, it's very painful. Doing it with wavelengths is a much better way to do it. That's why the hyperscalers are the most robust set of demand. But there are some industrial companies that are more focused on AI now and have placed orders for wavelengths.

So now that we're kind of -- I know we're still early days, but it's healthy, then we've got maybe at least a data set to work with of orders that are coming in. So you described to me a marketplace for wavelengths that looks like $800 to $900 a wave a month for 10 gig.

That's correct.

About $2,500 a month for 100 gig and that there is a nascent market for 400 gig, which we might think is around $5,000.

That's correct.

So if those are market prices, where is Cogent coming into the market? Are those orders booked at relative to those price points?

So again, off of a small base, the last quarter, our ARPUs were [$1,899]. So it was a mix of 10s and 100s, no 400s. And it is very route specific. There are cases where we will discount. We will discount in response to specific orders, but a little different than we did in entry to the transit market, we don't feel it's necessary to go to the market with a categorical 50% discount to market. I think when customers buy in bulk, they are expecting a discount. When customers buy for longer term -- we had one international carrier who insisted on 5-year term for wavelengths. Well, we gave a 15% discount for locking up that revenue for 5 years.

What would be the normal term, 2?

2 to 3. It's pretty, but obviously be the most common. But this particular carrier actually wanted it for 5 years and actually wants to prepay on install just the way their accounting allows them to treat it as a prepay and you never say no to taking payments upfront.

So just in that again, and maybe you can have -- give us a little bit more historical context given the work you've done. But in the past, we've talked about how the NetCentric business -- that's kind of down this perpetual price decline curve, I think maybe 10% decline, maybe more depending. But if we kind of take those $800, $900, $2,500, $5,000 price points, at what rate are those price points likely falling offset by what rate is...

So I started this process with maybe a bit of a pessimistic view. I had read a number of third-party studies, did some channel checks on our own. And I did not believe that the total addressable market for wavelengths was growing. It was about a $2 billion market, and it was flat. I think as I saw this fourth bucket of demand enter the market, I became more comfortable with the industry studies that have indicated the market is going to grow in dollar value at about 7% a year. As I think about price declines, they have averaged in the 15% to 20% per year, so slightly lower than the rate of price declines for transit, which is average 23% a year. But the issue with wavelength is it is so route specific for some routes that actually go up in price, particularly if the route has low latency characteristics or is particularly unique. So one of the advantages to the Sprint asset is that 90% of the routes are truly unique to Sprint.

The $19 ARPU, what do you think that is discount to market roughly? You said previously, when you did that, it was a 50% discount in wavelengths, where do you think you'll -- your pricing relative to market?

So on the discounting, it's an iterative process with sales. We have built an automated tool that takes 2 endpoints, you plug it into our CRM, it actually produces the KMZ map showing the exact route with 1 meter accuracy and produces a price. Customers naturally don't usually get the first price and there's a negotiation around that. Typically, if we need to discount off of that list more than about 10%, we are looking for some competitive offer; show us an invoice of what you're paying today, show us a competitive offer and we will guarantee to beat it. We have not seen many of those. We think that our target pricing was established realistically against the current market prices.

Okay. Got it. So all good on the wavelength front then. Market is flat to potentially growing, pricing coming down, but mix improving. You've come to market with pricing that people are pushing back on. Backlog is twice what you thought it would be. Current course and speed, we can still hit all the targets we've put out despite the fact that we would love to be provisioning faster if we could. And the next 15 months will be kind of pretty much the meat of getting to the point where you can do what you would really like to be doing at a national and a global level. Any -- presumably no one in the competitive landscape would be looking to price their back book differently. But is your presence in the market, do you detect affecting anyone's position in the front book as they go to market?

So the market is dominated by Lumen and Zayo, Lumen being by far and way the largest. They have appeared to be going in a different direction, focusing more on managed enterprise services both in their public statement, but probably the most important place to see that is in their sales comp plans. So they actually compensate sales people more for selling those services than they do selling wavelengths. Sales people are very Pavlovian, they follow the incentives. And for that reason, we think we have a good chance of capturing a very large market share quickly from someone who is not particularly focused on this market. And I'm going to play back the tape 15 years ago when we were a relatively nascent player in the transit market and at that time, Level 3 was the dominant transit player and they kind of said, "Well, transit is not an important product to us, Cogent is not really a credible player." And we have gone from a couple of percent market share to now 25% of that market surpassing Lumen on third-party research for number of routes, amount of traffic carried, number of connections, size of connections. By all of those objective measures, we have been the dominant player in transit. And their response had been to hold pricing, not be ultra-aggressive and to incent customers to buy other products. I don't want to be complacent, but it appears that's what they're doing with wavelengths as well. In the case of Zayo, there's a different issue. Zayo -- and we are a customer of Zayo as well as a supplier to them of IP. And we've got a great relationship, but they struggle to have accurate records of their inventory, having done 49 acquisitions and securing either wavelengths or dark fiber from them is challenging. And again, I just encourage people not to talk to us as a single customer, but talk to other customers in the market. Now I know they're working on correcting those inventory problems, but there's still some work to be done.

So just to understand the market better and your 2 key competitors. So this -- it sounds like there was a change in Lumen's behavior. Is that coinciding with the change in management there? And I'm just wondering why the change in Lumen and the defocus? And then secondly, on Zayo, they brought Steve Smith in a couple of years ago. So it's under the kind of new change in management there, do you think they're becoming a stronger competitor?

So 2 very different questions, Ana. So first of all, I think the desire to sell higher-level managed services already was started under Jeff Storey's administration and probably accelerated by Kate Johnson when she came in. So there's a greater emphasis, but it was something that was already in the works. Secondly, we were not in the wavelength business. So it's kind of something that was orthogonal to what we were paying attention to. But it does seem that both the composition of the sales force and the compensation of the sales force is incenting them to focus on enterprise accounts and somewhat harvest the existing wavelength business. So I can't predict the future, but it appear to be going in a different direction. In the case of Zayo, I think it is different. Zayo was not nearly as dominant in the wavelength business and a large part of its wavelength business was focused on a very niche market, low rates. It's actually a market that we are not focused on for high-speed trading. That seems to be the bulk of where they have been focused. They have a problem that's probably more severe than Lumen. Lumen was a series of acquisitions, but not 49 acquisitions. And as a customer of Zayo, I know we are often frustrated in buying dark fiber from them and they can't deliver it because they don't know where it is. Now they've hired third parties to come in and help them fix their inventory management system. They're putting in new software to do that. I think Steve is a great leader. I just think it's a very large task that he has in front of him. And it's not clear to me that the institutional knowledge and the data even exists to reconstruct their inventory. One of the huge advantages to Sprint was the fact there were no acquisitions. It was all organically built, they actually spent too much money maintaining records and data. They overengineer things. They disregard profitability. That's an issue that we're trying to address but it actually left us with a very good sense of where the network assets are.

So we just spent 40 minutes talking about 1% of the business.

Let's talk about the other 99%.

We have 0 minutes left. I think we did hit some of the run rate. Do you want to say just very quickly, anything that we need to know about Corporate or NetCentric that's going to surprise us in the third quarter?

I think the corporate business is going to continue to slowly improve, but do not expect it to rebound to pre-pandemic levels over the next several quarters. I think the NetCentric business will continue to outperform long-term expectations and will continue to perform above the run rates that we have had to date. We're seeing good growth in traffic, good growth in demand. We feel pretty good. And on the enterprise side, which is an acquired customer base, I think the 3 takeaways are: one, we're taking away products that customers have been used to, that are gross margin negative, and they're not happy, but they understand it. Two, I think those customers are more durable than we even thought they would be, because the cost of shifting to another provider is so large. And third, and this has been a little surprising to us, the lack of other global enterprise alternatives. So Lumen used to have a global network. It's now retreated in North America. AT&T & Verizon had global facilities-based networks have pulled back. Orange Business Services, BT, T-Systems, all had global facilities networks are now pulled back to our home market. That actually has left Cogent in this unique position of being the only global facilities-based network. Now we had the network to sell transit. But because that network exists, we're also able to layer on MPLS and VPLS services, and I think that will help us more than we expected in the enterprise space.

Okay. Great. We'll leave it there. Thank you, Dave, for joining us. We appreciate it. Thank you, everybody.

Thank you, everyone.

Thanks. Next up in this room will be Uniti.