Cogent Communications Holdings, Inc. The KeyBanc Technology Leadership Forumis 2026

Brandon, always thank you for hosting me. I'd like to thank investors for taking time out of their day, and maybe most importantly, KeyBank for yet another beautiful venue.

Probably a B, Brandon Nispel. On the cost cutting, I think we have done well. We actually increased our target and are running ahead of schedule. On the integration of the customers into our systems, processes, and naming conventions, as well as the migration onto our network, I think we have done well. On the conversion of the TDM voice network to a Wave network, we did well. Where I think we have underperformed is on the wavelength growth on that network. It took us almost two years to convert that network to Wave enable 1,139 locations. We were pretty much on schedule for that component. Where we have been behind is on the generation of revenue on that Wave network. We have provisioned Waves to approximately 2,500 unique wavelengths. That is to 608 of the 1,137 locations, and as of quarter end, 548 unique customers.

Those are all impressive metrics. The fact that the Wave business grew 9.2% sequentially and 68% on a year-over-year basis is impressive, but what is disappointing is the aggregate size of that base. We anticipated being able to get to 25% of the North American inter-city wavelength market within a three-year period of selling. Based on the pacing we have now, it will probably be longer than that. I think we are still confident in our ability to gain share, and we actually are encouraged by the fact that I think the wavelength market turned out to be a bigger market than we expected. The other area of disappointment has been in the enterprise customer base that we acquired. Just to remind investors, there was $485 million of revenue at acquisition. It was previously declining at 10.9% year over year.

It was about 10% non-core products, and it was 93% off-net. We went through a very deliberate process of converting as many customers to on-net, to squashing those non-core products, and to improve margin. We have been able to get the margins on that acquired customer base from negative 60% margins to breakeven. I think that's an impressive task. The fault, however, is that the revenue stream from those acquired customers has declined to about $150 million run rate.

I think there's three parts to that answer. One, our general philosophy is to give long-term, multi-year guidance to remind investors for 18 years, Cogent was a public company with no quarterly or annual guidance, and we grew an average of 10.2%, and we delivered 220 basis points a year of average margin expansion. We could do that because we had operating experience in the products we were selling and the customers we were selling to. The second point is, when we acquired Sprint, we needed to put some markers in the ground because I think investors rightly said, "Something is wrong if a smart company like T-Mobile is paying you $700 million to take an asset." And we needed to explain what we were going to do with that asset and how we were going to repurpose it.

To that end, we laid out a market share gain in wavelengths, which is a relatively concentrated market where customers are frustrated with the current suppliers and said we could replicate what we did in the IP market, and we thought we could do it quicker. I still do think we will do it quicker, but I also think we were too aggressive in the pace at which we would get there. Again, going from zero to a run rate of $65 million is impressive in a year and a half, but it is still not $500 million. I think we got wrong the pacing of that growth in Wave businesses. The third point is the guidance we gave around cost-cutting. There, I think we were very accurate. We actually exceeded our objectives and have been able to grow margins better than we and investors had anticipated.

I think it's important to look at our business through three discrete lenses. The lens of the legacy Cogent business. That is corporate on-net, corporate off-net, selling both VPNs and DIA services, on-net transit services, and then the sale or leasing of IPv4 addresses. The totality of that business, which had been the business that was growing at 10.2% a year, was negatively impacted by the pandemic. The pandemic slowed the growth in the corporate segment of that business, and it is now closer to a 5% growing business. With that slower growth rate, the rate of margin expansion in that business has slowed from about 200 basis points to 100 basis points. Still a good business, but not as good as it was pre-pandemic. We had hoped that the impact on office occupancy and employee number of days in the office was going to revert to pre-pandemic levels.

That has not happened. It is not appropriate for us to keep talking about the pandemic. This is the new reality that we face, and that legacy business, unless something changes, is probably a 5% growing business. The second lens to look at Cogent is to look through and see the acquired enterprise customers. We actually accelerated the rate of revenue decline due to our termination of non-fiber delivered off-net services, the decision to focus on higher bandwidth, the decision to terminate non-core services. So what was going into the acquisition declining at about 11%, accelerated to nearly a 25% annual rate of decline for the past several years. That business has gotten relatively small, but even in its current state, it is sufficiently large at that rate of decline to mask the growth in the other segments of the business.

The third piece is new businesses, something that Cogent had never done before. The most important of those is wavelengths. Sprint was not in the wavelength business. Cogent was not in the wavelength business. We took that asset, enabled it to sell wavelengths across the entire footprint at any of three speeds, and do it with rapid provisioning. We are encouraged by the market receptivity, as witnessed by the number of customers, but we have a ways to go. The legacy market for wavelengths is relatively static. The growing portion of that market are both neoclouds and hyperscalers. We have been frustrated by the fact that Wave that have been installed are not always accepted due to supply chain constraints, equipment availability, space availability, power availability, or even things as simple as the business model not yet being fully mature with our customers.

With that said, we think that these added applications of agentic AI inference and training will allow the totality of the wavelength market to grow. Many of our competitors have been at conferences such as this, touting anecdotally their growth, but have been very reluctant to provide the granularity that we provide. We provide granularity in unit number of Wave, ARPU, total revenue, and most of our competitors bury wavelengths into a larger set of products with anecdotal statements rather than specifics. So it is a mixed bag on Wave. We wish it was better. In absolute terms, it is doing well, but we need to see some of these constraints be alleviated so we can accelerate the quarter-over-quarter growth rate, year-over-year growth rate, and hit our multi-year target.