Ladies and gentlemen, good morning and welcome to the Cogent Communications Holdings Second Quarter 2024 Earnings Conference Call. As a reminder, this conference call is being recorded and it will be available for replay at www.cogentco.com. A transcript of this conference call will be posted on Cogent's website when it becomes available. Cogent's summary of financial and operational results attached to its press release can be downloaded from the Cogent's website. I would now like to turn the call over to Mr. Dave Schaeffer, Chairman and Chief Executive Officer of Cogent Communications Holdings.

Hey, good morning and thanks everyone for joining today this call. Welcome to our second quarter 2024 earnings conference call. I'm Dave Schaeffer, Cogent's Chief Executive Officer. With me on this morning's call is Tad Weed, our Chief Financial Officer. Hopefully, you've had a chance to review our earnings press release. The press release includes a number of historical metrics that we present on a consistent basis each quarter. Now for a couple of comments on activity in the quarter. In the quarter we had two debt transactions. On May 2, we closed the issuance of our inaugural $206 million asset backed securitization of our IPV4 notes at 7.9%. These notes mature in five years and may be extended for up to a 30-year term. This securitization transaction was the first ever of IPV4 lease revenue. We are the owners of approximately $37.8 million IPV4 addresses. We acquired $27.9 million of these addresses when we purchased PSINet and other acquisitions in the early 2000s. We further enhanced our portfolio when we acquired $9.9 million additional IPV4 addresses in May of 2023 as part of our acquisition of the Sprint Global Markets Group from T-Mobile. On June 11, we closed the issuance of our $300 million 7% unsecured notes. These are mirrored notes. These notes have identical terms to our existing $450 million 2027 notes. We use $114.6 million of the proceeds from this offering to prepay a Dark Fiber IRU finance lease at a 12% discount rate, saving $15.6 million in cash. This IRU Dark Fiber lease had monthly cash payments of $4.2 million a month through 2026, all of which have been eliminated and therefore materially improving our cash flow. Through December of 2026 we are leasing approximately $12.8 million of our IPV4 addresses out for a monthly revenue…

Thank you Dave and good morning everyone. This earnings conference call includes forward-looking statements. These forward-looking statements are based upon our current intent, belief and expectations. These forward-looking statements and all other statements that may be made on this call that are not historical facts are subject to a number of risks and uncertainties, and actual results may differ materially. Please refer to our SEC filings for more information on the factors that could cause actual results to differ. net-centric: Comments on the corporate business for the quarter. Our corporate business represented 45.9% of our revenues for the quarter and our corporate revenue grew by 7.7% year-over-year but decreased sequentially by 4.3%. The sequential decrease was due to the continued grooming of low margin off-net connections and the elimination of non-core products. We had 48,690 corporate customer connections on our network at the end of the quarter. For the quarter, the sequential impact of negative USF on our revenues was minus $1.4 million. Our net-centric business continues to benefit from continued growth in video traffic, activity related to AI or artificial intelligence, streaming and wavelength sales. Our net-centric business represented 35% of our revenues this quarter and grew by 4% year-over-year and by 4.5% on a constant currency basis, but declined sequentially by 0.9%. We had 61,736 net-centric customer connections on our network at quarter end. Our network traffic for the quarter, as Dave mentioned, increased by 1.9% sequentially and was up by 17.4% year-over-year. On enterprise. Our enterprise business represented 19.1% of our revenues this quarter and was $49.8 million. We had 18,356 enterprise customer connections at the end of the quarter and our enterprise revenue increased by 20.8% year-over-year and increased sequentially by 0.9% on revenue and customer connections by network type and on-net revenue. We serve our…

Hey, thanks Thad. I'd like to highlight a few of the strengths of our network, our customer base and our sales force. In our net-centric business, we continue to see significant traffic growth from our customer base. We are direct beneficiaries of over the top video, AI activity and streaming. At quarter end, we have over 1602 carrier neutral data centers connected to our network and 86 Cogent data centers, bringing our total connected data center footprint to 1688, more than any other carrier globally as measured by third party independent research. The breadth of our coverage enables our net-centric customers to better optimize their networks and reduce latency. We expect we'll continue to widen this lead in the market and are projected to add over an additional 100 carrier neutral data centers per year to our network over the next several years. With an extended provisioning cycle, we can sell wavelengths today in 574 locations. We are slow to do that because we are focused on the optimization of our network and therefore the shortening of our provisioning cycles across the entire base. By year-end 2024, we expect to be able to sell wavelength services in over 800 carrier neutrals in North America with very reduced provisioning intervals. At quarter end, we directly connected to 8135 networks. 23 of these networks are peers and 1112 of these networks are Cogent transit customers. This makes Cogent the most interconnected network globally. This collection of ISPs, telephone companies, cable companies and mobile phone operators allow us to reach the vast majority of the world's broadband subscribers and mobile phone users directly. We remain focused on our sales force, productivity and efficacy and continue to manage our underperforming reps. Our sales force turnover rate was 5.6% per month for the quarter, down…

Hi, thank you for taking the question. Dave, just starting with EBITDA Classic, you realized 62% of the targeted cost savings you just touched on, which is likely to provide a tailwind and continue to groom or exit unprofitable contracts. Just help us think about expectations for EBITDA Classic margin expansion. Is that something maybe we can continue to see here off of the 10.4 in the second quarter of 2024? And then separately, you may have heard that a competitor recently announced dark fiber deals totaling $5 billion. In the past you talked about dark fiber and today as a longer term opportunity, perhaps after you stand up and scale the waves opportunity. I guess, first, are you surprised by that announcement and does this in any way change how you're thinking about the speed to market on a dark fiber basis? And how are you positioned to maybe get a piece of that action? Thank you.

Sure. Hey thanks for both questions Sebastiano. So let me take the EBITDA Classic question first. We had significant improvement sequentially in that due to the reduction in our SG&A as well as cost of goods sold. We understand that the IP transit payments from T-Mobile are continuing to step down. In the quarter we had effectively a $20 million reduction in those payments, but our EBITDA as adjusted was down only $9 million due to the improvements in the underlying Classic EBITDA rate. We expect that trend to continue. We understand that in the third quarter we will only be receiving three of the $8.3 million payments, so we have effectively another $40 million of headwind to EBITDA as adjusted. We will continue to see improvement in EBITDA classic. The underlying costs due to these grooming and costs synergy programs, probably at a similar pace to what we saw from Q1 to Q2 and I think you'll see that similar pacing Q2 to Q3 and Q3 to Q4 and continuing on. As we add wavelength revenue, which carries very high contribution margins, we'll see a further acceleration in the EBITDA Classic number due to the fact that those wavelength services are virtually all on-net. And now that's probably a decent transition into the second question. We have been very clear about our targeted market for wavelengths and our strategy of being very aggressive in gaining market share. We think we have a number of competitive advantages, whether it be ubiquity, speed of provisioning or reliability once installed and uniqueness of routes. I actually think what was announced by our competitor was more of a defensive action as a result to our entry into the market. That particular competitor had been very reluctant to sell dark fiber. In fact, on…

Thanks.

And your next question comes from the line of Greg Williams with TD Cowen. Your line is open.

Great. Thanks for taking my questions. Dave, I just wanted to talk about Waves. It came in a little light at $3.6 million, but you did say it will ramp up early 2025 and you're going to reach the 800 data centers by year end. But we've only got 156 locations sold that you can sell waves to by the end of June. You've got a long way to go with less than five months to get to 800. So just help us feel comfortable with ramping up to the 800 wave data centers? And part and parcel to that, the backlog that you mentioned of 2700 circuits, can we apply the ARPU on waves of 1670 to that 2700 so we get a dollar amount to that backlog? Is that a fair way to do it? Thanks.

So, two very different questions, Greg. First of all, we are wave enabled in 574 data centers. As of today, we have actually only sold waves in 156 of those 574. Now, the method that we have to use to provision those wavelengths is very cumbersome and similar to the way our competitors provision a wavelength. The foundational work that we are doing, starting with a clean sheet of paper, allowed us to build a network on top of the Sprint physical assets that would allow us to provision a wavelength in two weeks without pre-deploying capital for transponders, which is a very capital and efficient way based on the number of wave permutations that are possible. So each time we provision a wave today, we take resources away from the network reconfiguration work. We will take that 574 and be at all 800 by year end. So we literally have 226 to go, and we're confident we will make that goal. Two, we are conducting a number of network modifications internal to the network to allow provisioning of a wave to occur end-to-end with only two field dispatches, compared to the industry average of probably six to eight dispatches, which is similar to the way we have to do it today. And with that, we should be able to rapidly eat into that backlog, but also hopefully build an even larger backlog as customers see, we are able to provision much faster than our competitors. Now to the ARPU question. Wavelength pricing is determined by two characteristics, the speed of the wavelength, which comes in three speeds, 10 gig, 100 gig and 400 gig. The larger the speed, the more expensive. The second dimension of pricing is the distance traversed. If you noticed, our ARPU did tick up 2% sequentially in the quarter with a modest number of incremental waves. More of the backlog is skewed towards hundred and 400 gig waves than 10 gig. That probably means the ARPU of what's in that backlog is slightly larger than the ARPU of the installed base. A big part of our effort is to make sure that we can support all of these speeds across the footprint. As these higher bandwidth applications continue to need more connectivity, we are seeing a transition away from 10 gig. 100 gig remains dominant, but I think over the next year or two 400 gig will be the dominant form of wavelengths bringing ARPUs up. And then beyond that, there will be a migration path based on the equipment we've installed to support 800 and eventually 1.6 terabyte waves. The systems that we are deploying are flex spectrum, meaning they're no longer adhering to the ITU grid standards and therefore allow us the flexibility to support these higher throughputs as the equipment vendors make them commercially available. Hopefully that was helpful, Greg.

That was helpful. Thank you.

Thanks.

And your next question comes from the line of Walter Piecyk with LightShed. Your line is open.

Dave, thanks for that explanation on the fiber deals. Just maybe as a counter thought on this, I mean, they needed money, right? They have a free cash flow burn situation. With your TSA payments coming up, your leverage ratio obviously is going to go up. Recognizing that maybe it's single-digit returns, is there still an opportunity maybe to use some of the Sprint dark fiber to do a similar deal except the lower terms, but basically just use it to cash in some money. I mean, it's basically just a construction type deal, right? I mean, sorry Dave, just to interrupt is your thought, I guess, to that over at some period of time that dark fiber will get used, you'll get higher returns, but then it's like time value in terms of when that may or may not happen in the future. Sorry, go ahead.

Yes, so I think two parts to that Walt. We are absolutely willing and expect to sell dark fiber on the routes that we have, i.e. sell our inventory where the return is infinite, because we have a negative cost basis in that fiber. We will do that when we have the resources to allocate to the provision at fiber. The new construction to single purpose facilities has to be weighed against other uses of our capital. I can't speak to what my competitors' alternatives are. I can speak to what Cogent has and Cogent has much better places to go to raise cash. We have over a 100 megawatts and a million square feet of surplus data center space. The current market price for that type of asset is about $1.4 million or about $14 million a megawatt compared to our offering price of $10 million of megawatt. So rather than go out and accept a very low return construction project, it is better for us to take our resources and capital and sell-off some of these data centers that again, we have a negative basis in. As Thad mentioned, we pivoted and did increase our capital spending slightly in the quarter to accelerate the availability of these facilities. Again, to remind everyone, these were not built as data centers, they were built as telephone central offices. Many of them are quite large, but we had to remove telephone equipment and we had to condition those spaces to turn them into marketable data centers and we're in discussions now with dozens of counterparties for dozens of locations. That is a better way to raise cash. We also have our IPV4 inventory. The competitor that you mentioned actually sold off a large portion of their inventory several years ago to raise cash. We did not. We chose not to. Prices continue to appreciate. We've demonstrated our ability to raise pricing on leasing and to securitize that revenue with a very low cost of capital. The fact that we have $462 million of cash on our balance sheet, we continue to grow our dividend. We buy back our stock, and we are deploying capital as fast as we can in the network, gives us the flexibility to say no to low single-digit IRRs. My goal is not to borrow money at 7% to get a 3% return.

That's fair. So Dave, that's a good segue to my next question. I mean, first on the data centers, yes, you can sell them, but I think you grew on-net data centers and your pace was typically adding carry on neutral data centers. I'm talking about I think you typically add 100 a year. You seem well ahead of that pace. Just curious thought process there? And then on the IPV4, similar question. The growth in the quarter was a little slower than last quarter in terms of the number of leases. Obviously you implemented a price increase, but is this the type of pace to expect on those IP leases? And given no sales in the quarter, and you're kind of talking about the lease growth there are sales still on the table. Can you give us some update on thoughts on whether that may or may not happen given the current market or are you just rather add the lease revenue sequentially every quarter? Thank you.

Yes, let me take those in order Walt. On the data centers, to be honest, it's a reactive process. We operate in 54 countries around the world and our goal is to get to every carrier neutral in every country we are licensed in. Now, we measure those based on the size of the facility, the quality of the underlying operator and our cost effective ability to get to that center. With the rapid influx of capital into the data center market over the past year and continuing, we are seeing a much larger number of data centers coming online. So it is likely that for the next year or two we may do over 100 a year just because there are more being built. And then on the IPV4, we are still considering selling addresses. We evaluate the market conditions. We know that the two largest buyers are currently leasing addresses at a significant premium to our even increased prices. As Thad mentioned, we increased pricing from an average of $0.30 an address per month across the base for new sales we went to $0.51 per address and saw a very modest reduction in unit volume and I think that's somewhat temporary. We also are going to evaluate our ability to raise the existing base with the average contract length of 1.5 months. We have a great deal of flexibility in raising prices to existing customers. This is a very sticky business with an annual churn rate of 0.8% per year. So we will look at our ability to lease more addresses at a higher rate and the DCF associated with that versus our cost of capital, versus what we can sell those addresses in the marketplace and we are open to any of those options to help us maximize return to our shareholders. Thanks, Walter.

Just a question on the pace.

Oh, I just got a long queue, Walt [indiscernible].

Your prepared comments that's the issue. The pace of leasing, is it, this is the typical pace or is it going to be up and down?

It's with any service you're going to depend on a specific demand in the quarter, but it does appear that the funnel for IP leasing remains very strong. And even with the increase in offering price, we are seeing no real diminution in demand.

Okay, thank you.

Thanks, Walt.

And your next question comes from the line of Tim Horan with Oppenheimer. Your line is open.

Thanks, Dave. Just following up on Walt. Would you raise prices on the legacy base and maybe timing or if so, why not the IPV4 addresses? And then secondly, it looks like Lumen is talking about a substantially larger market. I think they're referencing $7 billion or so or more for the IP, well, the wavelength market broadly speaking. Do you think the market is that large? And they're also talking about obviously doubling the locations they go to and a much better platform to provision services. Are you worried that maybe they will be a better competitor? Are you seeing them lower prices or quality improve there? Thank you.

Yes. So let me take the IPV4 one first, Tim. We absolutely intend to raise prices on the installed base. We will be doing that later this quarter for a portion of the base and measure the impact on churn. We will continue to monitor the market for the sale of addresses and that is definitely not off the table. The goal is to sell at the highest price possible and I think there is the likelihood that prices will go up. And then finally on IPV4 pricing, it's not clear to us that the increase that we put in place now will be the last increase. As the resource continues to increase in scarcity, we intend to adjust price to maximize return on our inventory. We've streamlined our sales process of addresses or leasing process and normalized our compensation structure. All of those things I think are going to help us see this continue to grow. And while it is only 4% of our total revenues, it did grow at a sequential growth rate of over 4%, quarter-over-quarter. Now to the wavelength market. There's both wavelengths and dark fiber and there is some fungibility for a lot of customers' dark fiber makes no sense because of the upfront capital and the ongoing operational support expense. For others, it may make sense. The global wavelength market is $7 billion as measured by third parties. Unfortunately, Lumen is only in North America, so they can't participate in $3.5 billion it's outside of North America. And then that includes both metro and long haul. We are going to have more locations wave enabled than anyone else. We'll continue to add them to follow up on Walt's question about data center addition, any data center in North America. So that means Mexico, Canada and the U.S. that we add, we will make wave enabled from day one. And we welcome competition. In the transit business, where we become the dominant player globally with 25% of the world's traffic, we competed in a market with 200 players. And in fact, the competitor you mentioned was the dominant player that we displaced to be the largest player. Listen, I wish I could wake up one day and be a monopolist. I'm not. I welcome the fact that we have to win customers every day on the value that we deliver. And I am quite confident that Cogent will have a better value proposition than any of our competitors. And if that means being more aggressive on wave pricing, that is absolutely something we are prepared and expect to be able to do. We are going to grow that business and generate that incremental revenue.

Yes, I guess, well, they're saying there's a potential like $5 billion locked up, but another $7 billion, so $12 billion total. I guess it's at a high level. Do you think the market maybe is much larger, growing much faster than maybe what you thought a year ago because of AI?

I think it's it probably is somewhat larger. But a big part of what they're referencing is new construction projects, meaning building fiber to facilities that never had fiber before. How big the recurring revenue opportunity is in those facilities is yet to be, figure it out.

Thank you.

And your next question comes from the line of Frank Louthan with Raymond James. Your line is open.

Great. Thanks, Dave, a couple of questions. Do you -- I think I know the answer, but just wanted to check. So looking at your network that you have, do you have spare conduit in GMG network that you could sell that would to possibly compete with some of these deals. And then going forward, once you have all of the buildings configured and so forth, and you're set for selling waves, what do you think is a reasonable quarterly run rate for that business once it gets up to speed? Thanks.

Yes, so in terms of conduit, we do not have spare conduit. The Sprint network was built with direct buried fiber in armor jacket, typically 6-feet below grade, versus conduit networks in plastic that are generally along public highway right of way, 2-feet below grade. As a result, we have had 50% fewer splices per mile than those other networks even though the network is a decade older. That gives us better loss and better throughput. We absolutely have fiber to sell, not conduits. And the fiber that we have is capable of supporting all of the coherent technology for 100, 400 and even 801.6 terabyte wavelengths when that becomes the de facto standard. Now, in terms of the run rate for our wave growth, we have said that from five years from acquisition, so that would be by May of 2028, we will be generating $500 million in wave revenues. We believe that is absolutely a realistic target out of the $2 billion addressable market that we referenced into Tim's point. May be we undersized that market a little bit, but we feel quite confident that we will get there. And I think once we have normalized our provisioning and our footprint is complete, that growth rate should be fairly linear.

Okay, great.

Thanks, Frank.

And your next question comes from the line of Jim Schneider with Goldman Sachs. Your line is open.

Good morning. Thanks for taking my question. Dave, I wanted to sort of ask you on the cost control side, delivering very strong cost controls on some of the data center transitions and other things. Maybe could you quantify for us how much more there is to go on this? Any other buckets of cost takeout there may be? But more importantly, with those cost takeouts, is that sufficient to potentially get you to an EBITDA non-classic or with a Sprint payments growth situation in 2025?

Hey, Jim well, first of all, welcome to Cogent. Welcome to coverage with us, and thanks for your question. So, we absolutely have significantly more cost savings to achieve. We have a combination of non-core services, low margin revenue services, and then additional G&A savings. We also understand that the T-Mobile payments were meant to bridge this period of maximum burn are stepping down. We will now be at a run rate for the next 41 quarters – 41 months, excuse me, at the lower rate of $8.3 million per month. With that, we absolutely will be in a position in 2025 when we're on an apples-to-apples basis. So we're looking at quarters where we have three monthly payments of $8.3 million in 2024 versus Q3 in 2025. So it'll be Q3 at the $8.3 million that EBITDA will be growing, and we expect that growth to continue. So, yes, EBITDA will step down in the remainder of 2024 due to these lower payments because EBITDA is adjusted now to the earlier question, our EBITDA classic will continue to improve, but that improvement will not fully offset that step down from $87 million down to $25 million. But after we normalize to that rate, we absolutely will be in a positive EBITDA growth number. And we've achieved only 62% of the targeted $220 million in savings. And our belief at this time is that $220 million will eventually go up. I think there are some additional savings opportunities that we'll be able to talk about in the quarters to come.

Thanks. And maybe a second one, at the risk of piling on waves and drug fiber, but maybe want to just get your sense, can you maybe kind of characterize for us your view on how many and which routes that Cogent has are unique relative to your competitors? And maybe say something about whether your competitor has any roots, which you believe are particularly unique. And if you sort of think about the future landscape of additional deals that could transpire?

Yes, we and our competitors, multiple competitors, serve all of the key North American markets, roughly 100 cities. I think, there's at least three or four dark fiber routes between those cities. Now, 90% of the Sprint routes have no other fiber on them. They were built along these unique railroad right of ways. There may be a parallel path along a public highway, along a pipeline. Generally, most of our competitors built their fiber, or did build their fiber differently along public highway right of way, lower burial distances, and in plastic conduit. The conduit absolutely gives you the flexibility to add more. But, if you pick two cities, just put two pins in a map, and I'm going to just pick two random, say, Phoenix to Cleveland. Cogent would have a unique route, but you could also buy a different route from a Zayo or a Lumen to connect those two cities. Now, it's more likely than not that maybe the other two are not on a unique right of way, but were all three of us going to be able to serve those two city payers

Thank you.

Thanks, Jim.

And your next question comes from the line of Nick Del Deo with MoffettNathanson. Your line is open.

Hey, good morning, guys. Thanks for the questions and appreciate all the detail on this call. I guess first you gave us a little color on some of the revenue and connection declines in your prepared remarks, but I was hoping you could dig a little bit into the corporate trends specifically. I'm just trying to tease out how much of that was a function of bad Sprint revenue and connections rolling off versus what's happening with better Sprint corporate connections and classic Cogent corporate connections.

Yes, so thanks for a question, Nick. First of all, our on-net revenues grew sequentially in the quarter, even though our net-centric revenue growth actually declined slightly in the quarter. So we grew on-net corporate revenues. That was both kind of traditional Cogent business as well as Sprint business that we migrated from off-net to on-net. There are three buckets of revenue that we acquired from Sprint that are not desirable. We have an easy fix for one of those three. That is traffic, that is in buildings that were off-net to Sprint, that are on-net to Cogent and we're rapidly migrating those customers as those tail circuit contracts mature. We also have non-core products that we absolutely want to end the life as quickly as possible. And you saw our run rate on those products declined by about $2 million sequentially. But there's still over $4 million a quarter in that non-core revenue bucket that really needs to go to zero and help us improve margins. And then the final bucket are off-net Sprint acquired circuits that are not profitable, and it's because they're very low speed. They may have been delivered over a suboptimal media coax twisted pair in some cases or fixed wireless were migrating to fiber and we're upping those speeds. Some of that revenue will go away. Most of it will actually transition to a fiber solution. So as we look at the corporate decline, it was really bad business that we are managing off intentionally and the good business is growing. And that I think is part of the way that we achieved that sequential improvement of over 250 basis points and cost of goods sold is by getting rid of that low margin off-net business as well as encore.

Okay, that'll make sense. And one other question on costs and EBITDA. Obviously good progress there. Again, we've talked in the past about there being a fair bit of integration, cost lease exit costs, lease remediation costs, and so on, that you're not specifically breaking out of EBITDA that probably suppress the reported number versus what the kind of real underlying results are, I guess. Could you update us on that and how we should think about that progressing in future quarters?

Yes. And for the past 15 months since the acquisition, we've been spending about $5 million a month on OpEx related to integration projects. We do not break that out separately. We do not add it back to EBITDA because we had the added complexity of the T-Mobile add back and did not want to, I think, overly complicate our accounting. We anticipate that $5 million a month drag on EBITDA to be a three year drag. That was what we said it would take us to go from the negative 190 of acquired EBITDA to, if you just looked at those customers, about $85 million to $90 million of positive EBITDA. I think we are on track to do that. I think that drag will continue beyond this year and then well into next year, but it will start to taper off. We are ahead of schedule in terms of taking costs out, which gives us a, the ability to turn off some of that extraordinary expense sooner.

Okay, great. Thanks, Dave.

Hey, thanks, Nick.

And your next question comes from the line of Bora Lee with RBC Capital Markets. Your line is open.

Hi, Dave. Thanks for taking the questions. Once the wave enablement is done and the internal team is freed up to turn its attention to dark fiber, what would it take operationally to get to a point for the sales force to be able to start offering dark fiber services? And what's sort of a rough timeline for that process?

So the sales force already knows the dark fiber is coming. They've had multiple requests from customers for specific routes. We will actually show this quarter a couple of very modest dark fiber sales in Q3, really to help certain customers in specific situations. But each of these requests is kind of a calculus of do we put in jeopardy the big opportunity to optimize the wave network? By year end, that optimization project will be complete and we can then kind of free up people to do that provisioning. I think in terms of complexity, these are more bespoke, it's less of a standard product, and it is very route specific. That means, the distance, it means the uniqueness of the route and also the total fiber availability on the route. Not all routes are the same. An example would be one of our most in demand routes is our northern route that goes between Chicago and Seattle. The reason for that is it's the lowest latency path between the east coast and the west coast. That's particularly interesting for certain applications. Why is that true? The earth is curved. The distance at higher latitudes is shorter than it is at lower latitudes to hit the same longitudinal line and for that reason, that's a particularly in demand path. We don't have enough market data and haven't actively sold to be able to fully assess that. But I do think next year we'll be in a position to have the wave business kind of working as expected and then pivot resources to start to build a funnel of dark fiber requests and then figure out where the optimal pricing is. But again, we have no intention of building new fiber based on what we view as suboptimal returns but we will sell out our excess inventory.

Got it. And monthly on-net ARPU continues to tick up. In the past you've mentioned that you're benefiting from a transition to higher capacity connections. Just wondering if that's what's driving the ARPU gains this quarter and if so, can you update us on the mix and trends that you've been seeing there?

Yes, that is in fact a key driver. On the corporate side, we continue to see an expansion of 10 gig opportunities and multitenant office buildings. And then on the net-centric side, we are seeing an increased number of 100 gig and even 400 gig transit ports, so all of that is helping drive higher ARPU on-net.

And lastly, just curious about the types of counterparties who've shown the most interest in the data center assets, just given their size, are these enterprises for their own use or national sponsors strategies? If you could provide any color there, that would be helpful.

Yes, I would say there are at least five different categories of potential counterparties that we're talking to. There are Tier 2 data center operators who are attracted to some of these markets. There are private equity funds that are looking for a platform to build off of. There are international carriers who are looking for a U.S. data center footprint. There are hyperscalers who are looking at these as edge facilities and there are AI compute businesses who want greater power density. But understand that there is a time to market advantage of these facilities. So, all five of those types of customers are in the mix today for these data centers.

Thanks Dave.

Hey, thanks.

And your next question comes from the line of Brandon Nispel with KeyBanc Capital Markets. Your line is open.

Great. Hey Dave, thanks for taking the question. Mostly just a housekeeping question. Could you guys provide what the churn rate was for net-centric corporate and enterprise? And then as we look at the reported connection, net adds or losses by those segments, could you give us an estimate in terms of how many are coming from sort of non-core, low margin products or other grooming, so we can try to understand what the underlying run rate is? Thanks.

Yes, so we give our churn rate between on-net and off-net, which is a good way to triangulate to the corporate churn rate. It's roughly about 1.2% per month and net-centric was just a little bit below that this quarter and enterprise was a little further below that. As you saw, enterprise revenues actually grow sequentially. So in terms of churn rate, there were no material deviations from kind of historical norms. Now, what was your second question, Brandon?

I was just hoping you could provide, you know, some adjustments for us as we look at the reported connection losses or gains to help us understand, what is sort of a function of you guys making decisions around, non-core or low margin product versus what the underlying business is doing.

Okay, so on the non-core, you should make the assumption that every non-core churn, both in unit and dollar, is intentional. We would like that business to go away today if we could, but we have contractual obligations that we have to honor and those run through the end of 2026. Again, that's part of the reason why we continue to get transit payment subsidies from T-Mobile through November of 2027. Secondly, for the both corporate and enterprise customers, there are locations for enterprise customers in countries that we are not licensed to do business. Sprint had been selling those through a agency program that we concluded did not meet our threshold for legal compliance. So we are effectively selling the port, but taking the loop and having the customer buy the loop separately. So that's an intentional churn to bifurcate that service and lower the ARPU by taking the loop of that off-net enterprise customer in an exotic market, off of our revenue stream. On the corporate side, it's mostly this very low speed broadband services that we are looking to migrate. Now, if those services happen to be in an on-net building, we're migrating them independent of speed and automatically upgrading the speed, if they are in a location that is very rural. We are in fact, terming those a very large termination. For example, in this last quarter, which was intentional of a corporate customer, was ERCOT. That is the energy management system in the State of Texas. They had several 100 locations, primarily at very rural substations with 2 megabit connectivity, all of which was delivered over copper, none of which was upgradable to a fiber solution. And when we looked at the totality of that particular customer relationship, our business is not providing SCADA backhaul management for 2 meg copper circuits. They're just not profitable. So we in fact are terming all of those. Now, some still have contract term on them, but a big bulk of the churn in this quarter was that specific, very uneconomic customer which actually helped us improve margin. They were actually losing money on those off-net circuits.

Thanks, Dave.

Hey, thanks Brandon.

And your final question comes from the line of Michael Rollins with Citi. Your line is open.

Thanks and good morning.

Good morning and thanks for hanging in there, Mike.

Sure I appreciate you squeezing me in for a question. Just curious to delve a little bit more into the corporate trends, if you can give us a sense of what's happening on the corporate side with respect to unique customers in the buildings and their spending patterns with you, with respect to the core transit services, the VPN, and the types of bandwidth that they're buying? Thanks.

Yes, so we are seeing modest growth in our corporate on-net business, probably in the order of about 4% year-over-year, still substantially below pre pandemic levels. We are winning new incremental customers, albeit the fact that there are less tenants in the building today than there were pre-pandemic. We also are seeing a migration of a subset of customers, not all, but a very, it's probably now several percent of the base that want 10 gig connections. So we went through a transition from 100 to gig. We're starting to see the beginning of that connection going to 10 gig. I don't think that is going to be a dominant product for several years because I think most customer premise equipment for corporate customers are not able to take those connections. But I do see that as an increasing ARPU uplift and corporate on-net services. And then to the VPN question. We saw a fairly dramatic grooming of VPN's coming out of the pandemic as companies were paring back multiple remote offices. We've seen that stabilize and we are seeing our VPLS sales to corporate customers be about a quarter of new incremental sales, which is similar to where it was pre pandemic. So when it dipped, it has started to come back. Now it's dwarfed by the VPN sales that we provide to the legacy Sprint customers that are provided over MPLS. And you know, we actually saw our enterprise business grow sequentially, in part because of those large enterprise networks continuing to grow and increase port speeds.

Thank Dave.

Hey, thanks Mike. Well, I would like to thank everyone. I appreciate your patience on the call and hopefully we answered all your questions. Look forward to seeing you soon, and take care all. Bye-bye.

Ladies and gentlemen, this concludes today's call, and we thank you for your participation. You may now disconnect.