Earnings Call Transcript - CCOI Q2 2025

Operator, Operator

Good morning, and welcome to the Cogent Communications Holdings Second Quarter 2025 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 website. I would now like to turn the call over to Mr. Dave Schaeffer, Chairman and Chief Executive Officer of Cogent Communications Holdings. Please go ahead.

David Schaeffer, CEO

Thank you, and good morning, everyone. Welcome to our Second Quarter 2025 Earnings Conference Call. I'm Dave Schaeffer, Cogent's Chief Executive Officer and with me on this morning's call is Tad Weed, our Chief Financial Officer. I'd like to take a moment to touch on some of the key milestones that we achieved in the quarter. As of the end of the quarter, we were offering Wavelength services in 938 data centers at 10 gig, 100 gig, and 400 gig service models. Materially, we also have reduced our provisioning intervals to approximately 30 days. Our Wavelength revenues for the quarter were $9.1 million, a 150% increase on a year-over-year basis and a sequential increase of that revenue stream of 27%. As of the end of the quarter, we have sold wavelengths in 418 locations. We currently have a backlog and funnel of 4,687 wavelength opportunities. We do intend to capture 25% of the highly concentrated North American wavelength market. In the quarter, we completed two significant debt transactions that materially enhanced our liquidity. In April, we issued an additional $174.4 million of debt against our IPv4 securitizations at a rate of 6.646%, which was substantially below our initial securitization rate of 7.924% for the initial $206 million of asset-backed securitization IPv4 notes issued in May of 2024. In June, we issued $600 million of 6.5% secured notes that mature in 2032. This extended the maturity of our $500 million secured notes which were coming due in May of 2026 and provided us an additional $100 million of liquidity. Our EBITDA increased sequentially by 11% to $48.5 million, and our EBITDA margin improved sequentially by 200 basis points to 19.7%. Our EBITDA as adjusted increased sequentially by 7% to $73.5 million, and our EBITDA margin as adjusted increased by 200 basis points sequentially to 29.8%. Our SG&A expenses declined sequentially by $5.6 million and a decline of 27% of our revenues to 25% of revenues. Our IPv4 leasing revenues for the quarter increased sequentially by 6.3% to $15.3 million and this represents a 40.1% increase on a year-over-year basis. Our average revenue per IPv4 address leased in the quarter was $0.39, a 22% increase from the base at the beginning of last year. We have an inventory of a total of approximately 38 million IPv4 addresses. We have continued the reconfiguration of Sprint facilities and added them to our data center footprint. We currently have connected 1,675 third-party carrier-neutral data centers as well as our total fleet of 187 Cogent data centers. The Cogent data centers have an installed base of 214 megawatts of available power. During the quarter, we purchased 230,000 shares of our stock for a price of $11.5 million at an average price of $50.18. So far this quarter, we have purchased an additional 95,000 shares for $4.5 million at an average price of $47.24. Our Board has authorized an additional $100 million buyback program that will remain in place through December 31, 2026. We currently have a grand total of $106.4 million available for the company under its buyback program. Our sales force productivity significantly improved in the quarter to 4.8 installed orders per rep per month from an average of 3.8 orders installed per rep per month in the previous quarter. After considering the impacts of the H.R. 1 tax bill, we do not expect to be a federal income tax payer for at least the next 5 years. Our Board decided to increase our dividend by another $0.005 per share quarterly from $1.01 per share per quarter to $1.015. This represents the 52nd consecutive sequential increase in our regular dividend and a 3% annual dividend growth rate. We anticipate our long-term average revenue growth to be between 6% and 8%. And we expect our EBITDA as adjusted margins to expand by approximately 200 basis points annually. Our updated revenue and EBITDA guidance targets are meant to be multiyear targets and are not intended to be specific quarterly or annual guidance. We're nearing the end of the grooming of unprofitable and undesirable revenue that we had acquired in the Sprint base. As these contracts expire, we expect to return to positive top line growth in mid-Q3 of 2025. We remain focused on selling high-margin on-net services. Our sequential revenue decline improved materially to $800,000 as compared to a sequential rate of revenue decline in the previous quarter of $5.2 million. With regard to our aggregate leverage, it has peaked at this point. We believe that our leverage on an LTM basis will continue to improve. Our leverage inclusive of the payments from T-Mobile under the IP transfer agreement and the net present value of those of $244.8 million should be treated as a cash receivable in calculating our net leverage and represented both on a short-term and long-term basis. Our gross debt as adjusted for the amounts due from T-Mobile on a gross basis was 7.74% at the end of the quarter and 6.61% at the end of Q2. As I stated, we expect these numbers to decline sequentially from this point forward. Now I'd like to turn it over to Tad and let him read our safe harbor language and provide some additional detail on the operating performance in the quarter.

Thaddeus G. Weed, CFO

Thank you, Dave, and good morning to everyone. This earnings conference call includes forward-looking statements based on our current intent, belief, and expectations. These statements, along with any other remarks made during this call that are not historical facts, carry risks and uncertainties that could result in actual outcomes differing materially. Please refer to our SEC filings for more information on these factors. Cogent has no obligation to update or revise any forward-looking statements. If non-GAAP financial measures are discussed, they will be reconciled to the corresponding GAAP measures in the earnings releases available on our website at cogentco.com. In summary of our results, our revenue for the quarter was $246.2 million, representing a sequential decline of $800,000. Our adjusted EBITDA was $73.5 million for the quarter, an increase of $4.7 million, with an adjusted EBITDA margin rising sequentially by 200 basis points to 29.8%. Our adjusted EBITDA includes $25 million per quarter under our IP transit agreement with T-Mobile. We received three payments this quarter totaling $25 million, which is consistent with the previous quarter. Last year’s second quarter saw $66.7 million in payments before they ramped down to $25 million quarterly, continuing until November 2027. Additionally, T-Mobile will provide at least $28 million related to lease obligations, to be paid in four equal installments from December 2027 to March 2028. We categorize our revenues by network connection type as well as customer type, which includes NetCentric, corporate, and enterprise. Our Corporate business accounted for 44.3% of our revenue this quarter, marking an 8.8% decrease year-over-year and a 1.5% decrease sequentially, attributed mainly to low-margin off-net customer connections and the removal of certain noncore products. Our NetCentric business benefited from growth in video traffic, AI activities, streaming, and wavelength sales, constituting 39.5% of our revenue with a 6.8% year-over-year increase and 5.1% sequentially. Conversely, our Enterprise business made up 16.2% of our revenue this quarter, showing a 19.9% decline year-over-year and an 8.8% decline sequentially, primarily due to reduced noncore and low-margin off-net enterprise revenues from the Sprint acquisition. Regarding on-net revenue, we serve 3,529 on-net buildings, generating $132.3 million this quarter, a 6% decrease year-over-year but a 2.1% sequential increase of $2.7 million. Off-net revenue was $102.2 million this quarter, showing an 8.3% year-over-year decrease and a 4.8% sequential decrease, with 26,239 off-net customers across 19,073 off-net buildings. These results were affected by the transition of certain off-net customers to on-net and the continued pruning of low-margin off-net contracts inherited from Sprint and T-Mobile. In terms of pricing, our average price per megabit for our installed base fell sequentially by 11% to $0.17, and down 30% year-over-year, aligning with historical trends. For new customer contracts, the average price per megabit was $0.08, reflecting a sequential drop of 21% and a 34% decrease year-over-year. On ARPU, we reported the following: $506 for on-net, $1,267 for off-net, $2,163 for wavelength, and $0.39 per address for IPv4. Churn has remained stable, with an on-net monthly churn rate of 1.4% and an off-net unit churn rate of 2.3%, which is a slight increase from 2.2% last quarter. Traffic on our network increased by 1% sequentially and by 9% year-over-year. Regarding foreign exchange, approximately 19% of our revenue this quarter was generated outside the U.S., reported in U.S. dollars. The average euro to USD exchange rate for this quarter has been $1.17 and CAD 0.73. If these rates remain steady, we estimate a $1 million positive impact on sequential revenues and about $2 million positive year-over-year. Our customer base is not heavily concentrated, with our top 25 customers representing 17% of revenues this quarter, comparable to last quarter. In terms of CapEx and capital lease payments, our CapEx fell by $1.9 million sequentially to $56.2 million, while principal payments on capital leases saw a slight increase of $0.5 million, totaling $8.5 million this quarter. We are currently integrating the former Sprint and legacy Cogent networks into one unified system, converting Sprint switch sites into Cogent data centers. We anticipate capital spending similar in the first half of 2025 to the last half of 2024, with a reduction expected in the latter half of this year. Our capital spending was $114.3 million for the first half of 2025 and $105.3 million for the fourth quarter. The principal payments on capital leases for the first half of 2025 were $16.5 million, down from $156.7 million last year, which included a buyout of an unprofitable lease for $114.6 million at a 12% discount. Regarding debt, our total gross debt at par, including $605.2 million in finance lease obligations, was $2.3 billion at the end of the quarter. Our net debt, totaling $1.8 billion, is calculated as total debt minus cash and $244.8 million due from T-Mobile. Our leverage ratio, measured under our more restrictive 2027 unsecured $750 million notes, was 6.82, while the secured leverage ratio was 4.2 and fixed coverage ratio was 2.43. Under the new 2032 secured $600 million notes indenture, the leverage ratio was calculated at 5.05, secured leverage at 3.12, and fixed coverage at 3.27. The definition of consolidated cash flow under the $600 million secured notes issued this quarter includes cash payments from the IP transit agreement with T-Mobile, totaling $100 million for the past 12 months. Finally, our days sales outstanding stood at 31 days at the end of the quarter, a slight increase from 29 days last quarter, primarily due to cash receipt timing. Our bad debt expense was significantly less than 1% of revenues for the quarter. Excellent performance there. I will now turn the call back over to Dave.

David Schaeffer, CEO

Thanks, Tad. I'd like to highlight a couple of the strengths of our network, our customer base, and sales force. We are direct beneficiaries of the continued increase in video traffic, artificial intelligence activity, and streaming trends. At quarter end, we were able to sell wavelengths in 938 carrier-neutral data centers across North America with reduced provisioning windows of 30 days. At quarter's end, we were able to sell our IP services globally in 1,862 data centers. At quarter's end, we were directly connected to 8,085 networks, 22 of these networks represent peers and 8,063 are Cogent transit customers. The reduction in networks connected from last quarter was due to the completion of the combination of the Sprint and Cogent IP networks into a single unified autonomous system number, AS174. Some details on our sales force. We remain focused on increasing our sales force productivity and managing out underperforming reps. Sales force turnover was 6.2% per month in the quarter, down from a peak of 8.7% during the height of the pandemic, but slightly above the historical average of 5.7% per month. At the end of the quarter, we had 628 sales reps. Our sales reps include 296 sales force that focus solely on the NetCentric market, 318 sales reps focusing on the corporate market in North America, and finally, 14 reps focusing on global enterprise customers. We expect to continue to provide profitable on-net and off-net IP services to enterprises, corporate customers, and NetCentric customers. We remain encouraged and enthusiastic about the prospects for our Wavelength business. We have a significant wavelength backlog funnel of over 4,687 wavelength opportunities. We have several hundred wavelengths that have been installed but have not yet billed due to customers' inability to accept the services as they are preparing their equipment to receive those wavelengths. And since our inception, we are focused on offering superior service, expedited provisioning, and disruptive pricing. We now have a base of installed wavelengths that are beginning to provide us data showing that our wavelength quality is substantially better than that of our competitors. We expect to continue to monitor this and use quality as a key differentiator in our ability to gain market share. With that, I'd like to open the floor for questions.

Operator, Operator

Our first question will come from Greg Williams with TD Cowen.

Gregory Bradford Williams, Analyst

Dave, you just ended the statements with focusing on quality as marketing your waves. Waves seem to be continuing to be off to a slow start. Did you still target 400 to 500 circuits installed a month by year-end? Second question is just on your data center sales progress. Any additional color on interest you're seeing and perhaps more importantly, at this point, is $10 million a megawatt a realistic ask from the interest that you're seeing? Or should we see a haircut on this target?

David Schaeffer, CEO

Thank you for your questions, Greg. We have always understood that to gain market share in wavelengths, we must excel in three areas: coverage, pricing, and service quality. The quality is assessed in two ways: our capacity to install quickly and the reliability of the service once set up. Initially, we prioritized the installation metric since our wavelength base was minimal. Although it's still small at around 1% of the North American market, it is now statistically significant enough for us to measure performance quality based on outage frequency along each route. Currently, and anecdotally, we experience approximately seven times fewer outages per span compared to at least one of our main competitors, as reported by customers with similar locations using both vendors. We believe these factors are essential for capturing market share. Regarding our wavelength installation pace, we set up and started billing for 147 wavelengths this quarter, but we've actually installed several hundred more that have yet to be billed. As mentioned in previous discussions, we are installing services faster than our customers anticipated, leading them to require additional components like cross-connects or optics on their end to utilize these wavelengths, which they are not used to due to delays from other providers. We are building trust with our customers and enhancing our installation speed. We can scale up to 500 wavelengths per month and will work towards this goal. Over the coming quarters, we believe our customers will adapt to our quick provisioning, which differs from the industry's standard practice. Consequently, we expect the number of wavelengths we install but do not bill to decrease. We share detailed performance metrics, which are vital until we reach a substantial base of billable revenue from our wavelength offerings. The 27% sequential revenue increase suggests we are gaining market share, though starting from a small base. Moving to your data center inquiry, we are still negotiating with the four initial parties that made offers and have received two additional offers, totaling six letters of intent for facilities ranging from our entire portfolio to single locations. We are cautious about projecting the sale proceeds into our recurring revenue, and we've refrained from doing so in the past. We've noted concerns about some counterparties being unable to provide significant nonrefundable deposits as we transition from intent to contract. Adjusting our pricing seems premature; we will allow the market to determine this. We do not have an initial cost basis for the data centers, only the investment made to upgrade these facilities. This gives us considerable flexibility in selling these noncore assets. We are conducting tours and engaging third-party consultants to evaluate the assets on behalf of financial sponsors. To address your question directly, we have received offers that vary from our full asking price to much lower amounts, making it difficult to predict what we will ultimately accept until a binding contract is in place.

Gregory Bradford Williams, Analyst

Just to be clear on your waves comment then. So the 147 additional waves, that's what you've billed, but you've installed far more than that, just not billed yet. Is that the right way to think about it?

David Schaeffer, CEO

That is absolutely correct, Greg. And as I stated previously, at some point in the future, it will only make sense for Cogent to report not on a funnel, not on orders installed but not billed, but actually installed and billing. But until we build a larger base, I think giving these incremental KPIs is helpful for investors. Next question, please.

Operator, Operator

Your next question comes from the line of Chris Schoell with UBS.

Christopher Joseph Schoell, Analyst

Last quarter, Dave, I think you talked about the business returning to top line growth by mid-Q3. Can you just provide your latest thoughts here and to the extent expectations have changed, what has shifted versus several months ago? And I also believe you slightly raised the margin expansion target long term. Can you just walk us through what gave you confidence to do that today and the main drivers of upside there?

David Schaeffer, CEO

Yes, sure. Thanks for both questions, Chris. So as we stated on our last earnings call, we expected our rate of revenue decline to materially decelerate; it actually did. It went from a sequential decline of $5.2 million quarter-over-quarter to $800,000. We knew that in July, we had a significant resale agreement that was actually terminated in June, but we had a tail that we had to support until July. That is behind us. And with that, one remaining large, noncore contract now terminated, we have clear visibility to monthly growth in revenue. Whether that is sufficient to get to aggregate positive for the quarter, it's very close, and I'm not prepared to say; there's FX and some noise around customers. But the rate of decline for the quarter may be lower than the $800,000. It could actually be a positive number. And then from that point forward, we expect to see positive revenue growth each and every quarter sequentially. It's important also and it ties in your margin point that the revenue growth we are experiencing is almost exclusively on-net services, whether they be IP-based services or wavelength services. The revenue declines are coming from much lower margin, in some cases negative margin off-net services. The fact that we delivered 200 basis points of margin expansion sequentially last quarter, looking back at the 8 quarters since the acquisition of the Sprint assets, we have outperformed 200 basis points annually and now with a return to growth, we feel very comfortable that we will replicate the type of margin expansion that Cogent historically had prior to acquiring Sprint. Just to remind investors, from the period in 2005 through 2023 when we acquired Sprint, over that 18-year period, Cogent organically, without an acquisition, had experienced an average rate of margin expansion of 220 basis points. So this is something that is not theoretical, but it is our actual historical results. We had a margin reset that occurred from acquiring a declining and negative margin business. We've taken more than $220 million of costs out of that business, and we have experienced better than 200 basis points since the initial acquisition. While there were puts and takes for severance and other reimbursables by T-Mobile as part of the transaction, with all of that extraordinary payment behind us, we now have a high degree of confidence that the 200 basis points, as I just outlined, is sustainable going forward on a year-over-year basis. Again, to remind everyone, this is meant to be a multiyear average. Some years we'll beat it. Some years, we may miss it. But on average, over the next decade, we will deliver more than 200 basis points a year of margin expansion on average.

Operator, Operator

Your next question comes from the line of Walt Piecyk with LightShed.

Walter Paul Piecyk, Analyst

Dave, can you just remind us in terms of sources of capital or where you could borrow against? Because I think unless your CapEx falls off a cliff, you're probably going to need some incremental capital to fund the dividend growth by early 2026. So can you just remind us what you can tap to bring funds in to fund that dividend?

David Schaeffer, CEO

Yes. Sure, Walt. So first of all, I'm going to disagree with your premise with over $300 million of cash on the balance sheet. Secondly, we have indicated that our capital spending was elevated in the second half of '24, in the first half of '25, primarily due to the upgrading of the data centers from DC power to AC. That capital expense is behind us now and was approximately a $100 million spread over four quarters. So as a result of that, we anticipate our annual rate, and this will be the rate in the second half of '25 and the full year '26, to be approximately $100 million in capital expenditures. On top of that $100 million of CapEx, we do make principal payments on capital leases. Due to GAAP accounting, it is shown at a different point in the cash flow statement, but it should be treated like CapEx. That number for the first half of this year was $16.5 million. We have indicated that the annual run rate for those principal payments should be about $40 million. So we were actually slightly below. It was elevated materially in '24 due to the buyout of the Verizon IRU as a one-time event, which we called out with its associating discount. As a result, the total amount of cash expenditures for principal payments and on capital leases and CapEx should be around $140 million a year. In terms of additional borrowing capabilities, we have borrowing capabilities at three levels in Cogent. We have additional capacity available in the group entity, where our leverage today is substantially below the covenant thresholds, and our debt service coverage is substantially above. So at Cogent Group, there is several hundred million dollars more of borrowing capacity, either secured or unsecured allowed in our current indentures, and it will likely be that we will look at the 2027 unsecured debt and probably look to refinance at some time in the latter part of 2026, possibly raising incremental capital but not necessarily. We also have incremental borrowing at Cogent infrastructure. That includes both the IPv4 ABS, which will have additional capacity based on the growth and cash flow in that as well as the ability to borrow against the other assets that reside in that entity. And then finally, there is always borrowing capability at the holding company level, which has no debt associated with it. However, we do not anticipate needing material incremental borrowings to either fund the dividend or operations as on an LQA basis. Our leverage peaked last year and has been declining and on an LTM basis, just arithmetically is going to continue to decline. While we are at 6.6x net leverage on an LTM basis at the end of this quarter, we anticipate that over the next six quarters, that number will fall below 5x and continue to delever.

Walter Paul Piecyk, Analyst

It's not 6 based on the math. In your calculation, you're essentially taking a net present value of TSA payments and also including those payments in the denominator. If you consider your reported adjusted EBITDA and actual net debt, the leverage is 7.5 times. Regarding the statement that gross debt isn't increasing, we'll have to see if that holds true in future quarters since you're paying $50 million in dividends each quarter, your operating cash burn is $30 million, and you have $300 million left in cash. If you claim that borrowings won't increase, we'll just wait and see how that unfolds in future quarters without a significant cut in capital expenditures. I'm not sure how the math can be represented this way when these are the reported numbers. The leverage stands at 7.5 times, and it’s 1 time if you exclude the TSA payments, which you're treating as an NPV. Excluding TSA payments, the leverage is 12 times based on trailing EBITDA.

David Schaeffer, CEO

So Walt, we've had this discussion on multiple earnings calls.

Walter Paul Piecyk, Analyst

It's math. Everyone has the numbers. They can do their own math.

David Schaeffer, CEO

I totally agree with that, and we absorbed a number of losses from T-Mobile, which have depressed our EBITDA, and we received a stream of payments over 54 months equaling $700 million. The net present value of those payments goes down each and every quarter as we receive those payments.

Walter Paul Piecyk, Analyst

You're trying to include it in the numerator and the denominator. You can do that and present it that way, but investors obviously have to make their own decision. Dave, can you just update us on in terms of the pledged stock. Has the Board put any limitation on that? And can you just walk us through the mechanics, if there's any further drop in the stock, how does that impact, if at all, the pledged shares?

David Schaeffer, CEO

So I as an individual have received my compensation from Cogent almost exclusively in stock for 25 years. I paid taxes on that stock as it vested. I have a basis in my stock as of the beginning of this year of $155 million. I borrowed against a portion of that stock in order to fund those tax payments so I did not have to sell any. I was fortunate enough to have income from other sources, primarily my real estate portfolio. As the D.C. real estate market deteriorated, I had additional pressure to reduce leverage on my real estate portfolio, which forced me to begin selling Cogent stock. Some of that stock is pledged. As a result, I have to reduce the pledge amount as well as receive cash to fund equity injections into my real estate. I've tried to be extremely transparent with investors, probably more than most people in my situation would be. I am committed to making sure that as an individual, not as Cogent, my lenders are made whole even though many of my peers in my industry have walked away from their assets. The policy that the Board has has not changed in terms of my ability to pledge, my ability to not hedge in any way, which if I had that ability would negate some of the pressure that forces me to sell.

Walter Paul Piecyk, Analyst

But is there a cap on what you can pledge? Because if you just search GPT, I think this has occurred with Elon and other companies and boards have actually implemented caps on the percentage of shares that can be pledged.

David Schaeffer, CEO

There is not at Cogent.

Walter Paul Piecyk, Analyst

Who in the Board makes that decision?

David Schaeffer, CEO

The Audit Committee.

Operator, Operator

Your next question comes from the line of Nick Del Deo with MoffettNathanson.

Nicholas Ralph Del Deo, Analyst

Dave, you noted that you had provisioned but not yet billed for several hundred waves. Did that metric go up quarter-over-quarter?

David Schaeffer, CEO

Yes.

Nicholas Ralph Del Deo, Analyst

Okay, meaningfully?

David Schaeffer, CEO

Meaningfully. And we commented on this when we reported Q1 numbers and the majority of the wavelengths were installed near the very end of the quarter, and it was the reason for the disconnect between 2% revenue growth and 18% unit growth. We also explained that we wanted to be careful not to alienate significant customers by being too aggressive and pressuring them to accept wavelengths. We have several large customers that have been truly shocked by our ability to provision in the windows that we have outlined. As a result, they were not prepared to take the wavelength services. They typically order their cross-connects and order their pluggable optics and accept them in a 3- to 4-month window from placing new order. But in Cogent's case, we had one very large wave order that we were actually able to provision nearly 100 waves in 7 days. I mean they were truly amazed at that, but they came back and said we can't take them. I understand investors' frustrations that they want to see an installed number. We only report installations based on billing revenue. That's been our policy and practice since inception. We have given incremental KPIs to help investors understand that there's demand here. But as I've said on previous earnings calls, and I want to repeat today, the ultimate goal is to be measured only by GAAP revenue growth. Listen, on that metric, it looked really good, but it was because a lot of waves installed at the very end of the previous quarter, which got us to 27% sequential and 150% revenue growth. With a small base and with these lags, it is going to be lumpier than I would like. Our piece of it is much smoother than the lumpiness suggests. I think there will be a point in time when we can implement a forced billing discipline, but we are unwilling to do that at this point.

Nicholas Ralph Del Deo, Analyst

Okay. Got it. And then maybe just turning to the data centers. What do you think is holding those bidders back from putting down deposits? Because it seemed like you expressed a bit more caution on that front than you have historically.

David Schaeffer, CEO

Well, it's because another 10 weeks has passed since we have publicly commented on this, and there are no firm deposits in hand. I would say it's kind of twofold. I think for the operators that have LOIs in, they are struggling to get capital committed, and we've had offers for what I would consider de minimis deposits relative to the size of the portfolio, and I view those as just unacceptable to go to contract, letting someone tie up the portfolio for, say, 1% of the proposed purchase price, that is just inappropriate. You wouldn't do that if you were buying a hotel or a shopping center or any real property asset. For the private equity sponsors, they are trying to get comfort around the revenue stream that their management team is going to generate while they have continued to spend money and do condition reports, do tours, quality assessments, and they are being cautious because what they would like to do, which I would do if I was in their shoes, is derisk it by going to the management team you're backing and saying, show me an actual end-user contract that I can underwrite. We've been very clear that the assets we are looking to divest of have no recurring revenue associated with them. I think these two constraints have slowed down the process. We've tried to caution investors not to place a lot of value on these. I still think they will be monetized, but I'm not in a position as I am with wavelengths to give you any clarity on the when and how much until we actually conclude a binding transaction with a meaningful deposit.

Operator, Operator

Your next question comes from the line of Mike Funk with Bank of America.

Michael J. Funk, Analyst

Good to hear from you again. So given that I'm newer back to the story, let me ask some basic questions here. On the provisioning of circuits, I heard your comments; some customers are not ready to take delivery as soon as you're available. But from my perspective, it signals a disconnect between your sales team, your provisioning team, and the customer; I would expect better coordination. So I guess, where is the disconnect if customers aren't ready to take, and what are you doing to alleviate that or improve coordination with customers? And do you expect them to shorten their delivery acceptance time?

David Schaeffer, CEO

Yes. So for IT services, which are the bulk of Cogent's revenues, they are 87% of our revenues, we have a 25-year track record. We install services on-net at an average of about 9 days. We also allow customers two times to push out the delivery, and then we have a kind of forced billing discipline that's in place. This policy for transit, DIA and VPN services has been in place for over 20 years. Customers are accustomed to it. There is very little float in terms of IP orders installed but not billed. It's not zero, but it's a couple of percent. In the wavelength market, we are a new entrant. Secondly, we made representations that were much more aggressive than any of the other vendors in the market in terms of our speed to deliver, the breadth of locations that we could deliver, and the quality of the service we would deliver. I think it's totally legitimate for customers to say, "show me." Secondly, for the period between the deal closing in May of '23 and the end of '24, so in that roughly 18-month period, we were doing one-off provisionings, but they did not go smoothly. We did not have all of the automated systems and processes in place, and we did not have the ubiquity of coverage. We have sold about 1,000 wavelengths in that kind of semi-manual but lengthy process. At the beginning of this year, we began provisioning in a streamlined automated way. We have surprised our customers both in terms of where and how quickly we could deliver. Those customers are starting to adjust their behavior. But they've been accustomed to going to the two other major vendors and getting a 3- to 4-month delay with a failure rate of, in some cases, up to 50%. We have not had any situations where we have committed to provision and could not provision. There are a few cases where the provisioning windows were as long as 90 days and certain vectors at certain speeds, but a very small percent, about 6% of the total footprint has those limitations. So the customers are now starting to understand the first part of our value proposition, which is we're not misleading them, and we really are able to provision the way we say we can.

Michael J. Funk, Analyst

I think I don't want to take too much of your time, but I mean, I think one other potential view concern could be maybe customers have over-provisioned or over-purchased wavelengths and are now maybe dragging their feet, delaying acceptance of delivery because if you look at other parts of the ecosystem, whether it's data center capacity or other pieces, customers can't take that fast enough, right? And so from my seat, I'm hearing this and I'm thinking the customers over-provisioned just to make sure they had enough potential inventory capacity and now they're pushing out or delaying acceptance. I mean that's the concern from my standpoint, but I definitely hear what you're saying you’ve improved provision. Can I ask one more quick one, Dave? I think in the past, you talked about a 4Q exit run rate for waves, I think $20 million was the last time you spoke. Can you update us there, please?

David Schaeffer, CEO

Yes, we still feel confident that we will hit that quarterly run rate in the fourth quarter. And I do want to go back to the over-purchasing and delay comment. That is not what we've heard from customers. It's really, we're just surprised you did it when you said you would do it. I do think it goes back to the other comment I started to make earlier, which is we still have to prove to people that not only can we provision faster, the quality is going to be higher than that of our competitors. And every vendor can make claims. The only way you validate them is by delivering and monitoring. I feel comfortable that we are building credibility, and the size of the funnel that we are accelerating and building is giving us that confidence.

Operator, Operator

Your next question comes from the line of Frank Louthan with Raymond James.

Frank Garrett Louthan, Analyst

A couple of questions. First, where are you getting most of your wavelength customers? Are they new to Cogent? Or are they from your existing base? And then as far as going forward with the data centers, do you need to hire more experienced dedicated salespeople for that space? Or do you have any sales channel relationships that can help with converting that space?

David Schaeffer, CEO

Okay. Two very different questions. First of all, on the wavelength customers. To date about 3/4 of them have been existing Cogent transit customers and about 25% of them are brand new to Cogent. Whether that mix will continue to hold as we build the pipeline and install, I'm not sure, but to date, of the 4,600-plus in the funnel and the 1,500-plus installed, the mix has been 3/4, 1/4. Regarding the data centers, Frank, what we are doing now is trying to do two very different things. One, we continue to do one- and two-rack retail deals into our retail footprint. We have 187 facilities where we have retail space available. We're at about 14.5% utilization in that footprint, and the entire 628-person Cogent sales force has been the ones who have been focused on filling that footprint up. I think that will continue. The wholesale disposition is a very different process. There, we actually have one of our real estate professionals focused on that disposition process.

Operator, Operator

Your next question comes from the line of Michael Rollins with Citigroup.

Michael Ian Rollins, Analyst

Two topics, if I could, please. So first, when you look at the opportunities to improve revenue in the future, you talked about waves a bunch on this call. Can you talk more specifically about the customer verticals and how those are each progressing in terms of the corporate, the NetCentric, and the enterprise? And then secondly, you mentioned earlier in the call your target to reduce net debt leverage. Can you give us just a little bit more of an explanation of how you see both the numerator and denominator evolving over these next couple of years and what are the critical points of execution to deliver on each of those?

David Schaeffer, CEO

Yes. Sure, Mike. Very good questions. So first of all, as I stated in the prepared remarks, we remain focused on selling on-net services. Just to remind everyone on this call, every dollar of revenue gets classified by on-net, off-net customer type, geography, and by product type. In our investor presentation, we give you a great deal of granularity and breakdown of the various mixes of customer type, product and on-net and off-net. Our primary business is selling on-net services. We get much higher contribution margins from those on-net services. For our corporate customers, roughly half of their purchases are on-net, half are in locations that we concluded we cannot economically get a return on invested capital to bring on-net and buy off-net services for. For the enterprise base, based on their very disparate geography requirements, we are only about 10% to 15% on-net and almost exclusively off-net. That is business that came to us from the acquisition of Sprint. In the NetCentric segment, about 90% of our revenues are on-net, we are focused on on-net corporate and on-net NetCentric. Within that, wavelength as a product is almost exclusively on-net. The customer verticals for wavelengths are typically hyperscalers who are using them for AI or for content distribution, other content distribution customers, regional access networks who connect their networks together, international carriers who extend their networks, and finally, some enterprises should build their own private closed networks. Each of these represents drivers for wavelengths. For transit services, the market is typically divided between access networks which we have about 8,050 or so that pull down content around the world, and about 5,000 content-generating customers that push applications out. They could be either hyperscalers for their core business; they can be CDNs; publishing companies; or application service providers. This is a good pivot into the second part of your question, which is how do we deleverage. We deleverage through three mechanisms. The first being growth in aggregate revenue. Since acquiring Sprint, we've deleveraged and improved EBITDA solely through the margin improvement that has occurred faster than the decline in the payment streams from T-Mobile. Over time, those payment subsidies will go away, and we need to grow EBITDA out of top line growth with high contribution margin products. It is why we are confident in saying we can return to a combined basis of 200 basis points of margin expansion year-over-year. That is a significant delevering in and of itself and then returning from what is effectively negative 1% growth we reported this quarter to a year-over-year growth rate of between 6% and 8% coupled with the deleveraging gets you to EBITDA growth rate in the low to mid-teens, which is comparable to where Cogent had been historically prior to acquiring the Sprint business. In terms of the aggregate leverage, it has been Cogent's policy since 2010 to return more than 100% of free cash flow. We did that successfully between 2010 and 2020, maintaining a net leverage range of around 3x levered. With the pandemic and the slowdown in our corporate business, our leverage crept up to 4.2x net levered. We acquired the assets from T-Mobile, all with the subsidy payments, we initially deleveraged due to the asymmetry of those payments down to below 3x levered. As those payments step down from $87 million a quarter to $25 million a quarter, our leverage has ticked up. We peaked in net leverage this quarter at 6.6x net leverage on a fully consolidated basis. That number will come down, but our gross leverage will probably not materially come down. It will come down by improving the aggregate amount of EBITDA.

Thaddeus G. Weed, CFO

I would like to briefly address the numerator and the denominator. The 6.6 figure is consistent. When it comes to debt, the net debt is calculated by deducting the $244.8 million from T-Mobile, which includes cash and cash equivalents for the short-term and long-term investments for the long-term portion. This gives us the net debt for the numerator. For the denominator, the EBITDA is based on historical data, looking back over the last 12 months that have been paid in cash. The numerator reflects the balance sheet date, while the denominator reflects the past year. There is no double counting, and we consistently apply the treatment of payments from T-Mobile, both those received in the past and those expected in the future.

Operator, Operator

Your next question comes from the line of Tim Horan with Oppenheimer.

Timothy Kelly Horan, Analyst

Dave, can you just reiterate your wavelength kind of longer-term $500 million target and timing and confidence there? Secondly, can you just give us some your best guess on timing of the data center resolution? And then third, can you give us some sense of what the actual EBITDA numbers will be for the second half of the year, either third or fourth quarter or full year? Any kind of sense would be helpful.

David Schaeffer, CEO

Yes. Sure. Always good to hear from you, Tim, and thanks for the questions. So first of all, on the confidence on wavelengths, we are actually more confident today than we were on last quarter's call or the quarter before or since we acquired Sprint. The reception that we have received from the customer base, the orders that we have in our funnel, and the customer feedback that we've gotten from the orders that we have installed all give us confidence that we will reach our $500 million run rate on wavelength revenue by mid-year 2028, which is identical to what we laid out in our justification in September of '22 when we announced the potential transaction that ultimately closed in May of '23. Regarding the data centers, I am not going to put a date. Of course, we have never done it before. We have interested parties. We could tell they're spending money, they're hiring professionals, they are doing analysis, they have put in offers. But until we have an actual monetized deal with a meaningful at-risk deposit, I'm unprepared to put a stake in the ground of saying when we're going to close because we have no history. We have had a great deal of interest. We've had some parties who say, it's not for them and they've moved away, but more have been interested than not and many parties are continuing to do work. With regard to EBITDA, I'm going to qualify this by saying we don't give quarterly or even annual guidance. What I will say is we expect meaningful sequential growth in EBITDA each and every quarter going forward at or greater than the pacing that we delivered in the last several quarters. I think you can model that out, but we feel comfortable that we are both deleveraging due to the growth in EBITDA and growing our cash flow. The 6% to 8% top line growth, I think, is now much more realistic than it was when we announced the deal at 5% to 7% in '22. And again, to remind investors, from 2005 through 2020, Cogent with no acquisitions organically grew at 10.2% a year. Our growth rate went negative when we acquired the Sprint business as was planned. We initially thought we could only return to 5% to 7%. We've become comfortable that we've groomed out the undesirable revenue. The rate of revenue decline sequentially improved from $5.2 million negative to $800,000 and should be flat to slightly positive in Q3 and positive from that point going forward. And because we have demonstrated post-closing more than 200 basis points a year of margin expansion, and in fact, we delivered 200 basis points sequentially in the last quarter, we feel comfortable that's the right goalpost going forward. Hopefully, that was helpful, Tim.

Operator, Operator

We have no further questions for today. That concludes the Q&A session. And I would now like to turn the call back over to Dave Schaeffer for closing remarks.

David Schaeffer, CEO

I'd like to thank everyone for being on today's call. Hopefully, we were clear in answering your questions. We look forward to seeing investors at some upcoming conferences and remain extremely encouraged about our growth prospects and our ability to expand our free cash flow, and maybe most importantly, our commitment to return capital to shareholders. Take care, all, we'll talk soon. Bye-bye.

Operator, Operator

This concludes the meeting. You may now disconnect your lines. Have a pleasant day, everyone.