Cogent - Hot Topics

Introduction

Since I have published a few Insights on CCOI and appeared on Yet Another Value Podcast, I have received a lot of inbound interest in Cogent from other investors.  In the last few months, I have spoken to a couple dozen investors that have been looking at the company or are current investors.

I enjoy meeting and speaking to other investors, but it has been a pretty big use of time.  I thought I would consolidate my recent thoughts into a post.  

I will organize this into some of the hot topics discussed based on my conversations with other investors. This is a pretty specific post that assumes the reader is quite well versed on Cogent already.

What’s Going on with Waves?

This is the most popular and most important topic that consumes the most time by far - and for good reason. It is the primary swing factor in Cogent’s valuation.

High level, I think Cogent will build a big business in wavelengths.  Clearly I have been wrong about the timing of the ramp in this business.  When I diagnose why, I believe it boils down to one primary factor: I have been so focused on the supply-side dynamics, that I wasn’t appreciative enough of the demand-side complexities of meeting that supply, which creates time lags.  I’ll explain what I mean below.

Cogent has built a novel service layer in the fiber market.  It sliced a purpose-built layer of its fiber network specifically to address the wavelength services market.  This is was a novel approach to this market and was a large focus of my previous research on the company.  I was a bit obsessed with researching this angle because the telecom market hadn’t seen something like this before, and if it worked (hint: it does), it should be disruptive.  

It’s not a perfect analogy, but I’d compare Cogent’s wavelength network to a GPU for AI workloads.  CPUs can perform AI workloads too, but they are much less efficient because they are architected differently.  Cogent has brought GPU-like capability to the fiber connectivity market, while legacy providers offer services provisioned by their CPU-equivalent network and processes (general purpose, can do anything, but less efficient at specialized operations).  Lumen, Zayo and others can run wavelength services on their networks, but they are not architected to do them efficiently. The specialization of Cogent’s architecture opens new use cases and opportunities.

So, I spent a lot of time understanding the architectural elements of this effort.  I have spoken to dozens of customers since the acquisition was announced, mostly to understand how the market would react to this network coming to the market.  The feedback has been highly consistent: legacy wavelength providers install on inconsistent time frames, there are big lags in the provisioning cycles, and Cogent’s unique physical paths are attractive. That Cogent could offer all this at lower prices is a nice cherry on top, but I suspect most customers do not need aggressive pricing from Cogent to swing share its way. I’m not sure if we will ever discover if that’s true - Cogent is aggressive on pricing in all its businesses.  

In retrospect, I had assumed that customers would jump at the opportunity to work with Cogent once its network was done because this new network could solve so many known and verbalized pain points for them.  What I missed (and Dave Schaeffer did as well) was that customers were not ready to make that leap immediately (especially in Cogent’s reported wave connections) after Cogent completed its network for several reasons:

1. They probably didn’t believe Cogent could deliver on its installation SLA promises.  I have heard this repeatedly. Everyone is shocked that Cogent can install on time within 30 business days.

2. They had heard for the last several years (while Cogent was performing the network transformation) from legacy providers that this fiber is old, unreliable, and not as good.  The network performance data disproves this point, but it was an easy competitive counterpoint before Cogent was operationally ready to sell and install services in its efficient, scalable way.

3. Even if they could get past #1 and #2, customers do not (yet) work on immediate time frames for these services.  Looking back through all my notes from conversations with network engineers and architects, it was pretty clear that most are often planning their private fiber services quarters to years in advance.  There are immediate needs that come up here and there (mostly from hyperscalers), but the sweet spot for customer deliveries is likely between months 6-15 - although this is somewhat hard to pin down since the legacy providers often have difficulty delivering waves faster than 3 months anyway.  

4. Even when customers want waves, sometimes there are dependencies elsewhere in their networks that prevent them from completing the technical work required for them to receive service from Cogent, which causes lags between installation and delivery of service. There could be equipment delays, there could be delays procuring cross-connects in the endpoint facilities, etc. This is not a business where customers can simply flip a switch and go - it is a more technical service that requires engineering from Cogent and from customers. Cogent deployed network architecture that pre-engineered many of the technical steps and allows it to operate waves very similar to the way it operates IP services, but the ecosystem does not yet run like that.

Cogent has to prove itself to customers in market, and it’s doing that already in this early ramp-up phase.  It is delivering on its installation SLAs consistently.  It is surprising customers by being on time or early.  Its service quality is consistently excellent on the parameters defined by the contracts (latency, uptime, delivered capacity).  Customers that are working with Cogent are noticing its exceptionally fast and predictable installation times.  

Anecdotally, I hear customers that have seen Cogent’s service levels are already coming back for more and giving Cogent a wider view of their connectivity needs.  Before 2025, I estimate Cogent saw less than 15% of opportunities from customers, from which it generated a funnel of 10,000 opportunities over 18 months.  It would be transformational to see a larger view of opportunities in the market now that customers know Cogent can perform.

Based on Dave’s recent conference comments, Cogent has 2,000-2,500 contracted wavelength orders, plus another 2,000-2,500 opportunities in its sales funnel with defined paths and technical parameters, but not yet signed.  To hit Dave’s target of another 1,500 installations by year end, Cogent has to install and bill 60-70% of the firm orders it has on its books.  This appears aggressive given the pace we’ve seen YTD, but we are now entering the sweet spot for customer deliveries given the lags the company has already experienced with customers.  

Within a few more months, we should be approaching the cruising speed for this business - when Cogent has enough seasoned orders that are ready for installation/delivery/billing and also is originating more orders to refill and grow its backlog. 

The wavelength business should experience rapid growth over the coming years.  Dave mentioned at the August 2025 Oppenheimer conference that he expects wavelength connections to double or triple in 2026.  A doubling from 3,000 would be another 250/month in 2026 - a pretty conservative rate.  A tripling would be an incremental 6,000, or 500/month - what Dave believes to be the steady cruising speed for this business.

To summarize, Q2 2025’s 147 new billed connections and ~400 installations were disappointing, but understandable in the context of the demand-side dynamics.  

Cogent’s Capital Allocation Policy

This is another very hot topic right now. I will give my view on the context of how we got here, and then offer my views on the situation.

Cogent has used a Malone-like levered equity capital allocation policy for over a decade.  The strategy was to pay out over 100% of operating FCF to intentionally raise leverage, from 1x to 2x to 3x over several years.  Because of the recent acquisition integration-related outflows (DC conversion capex, Sprint losses, integration expenses, etc.) and the slower-than-expected ramp of wavelength revenue, for the last year Cogent this legacy policy has become a very aggressive dividend growth strategy.

I have heard a wide range of opinions on Cogent’s balance sheet and capital allocation strategy, many of which are sensible and valid.  My personal views are below.  

For over a year, I have thought the current strategy was inappropriate given the fact pattern.  When we knew a year ago that the “linear ramp” in wavelength revenues was way behind plan, the board should have altered its dividend strategy.  

In my opinion, it is inappropriate for management and/or the board to put common shareholders through the stress of an extremely highly leveraged equity story for an already complicated merger and corporate transformation.  In essence, the dividend strategy as implemented today is itself a levered play on wavelengths.  The only way they can maintain this policy is to materially accelerate revenue, EBITDA, and FCF growth.  Dave personally suffered tremendously from this capital allocation policy by losing all his pledged shares from a margin call.  There was no need to suffer through this level of elevated volatility.

Additionally, with the stock in the $30s, it makes much more sense to suspend or materially cut the dividend and repurchase shares, which would further reduce the cash burden of the dividend strategy in outer periods.  

I have advocated for a capital allocation policy that returns to responsible levels of leverage more rapidly and provides a clearer path forward.  This includes:

1. Cut the dividend to $0.25/share and do not grow it until net leverage falls below 4.5x. Suspending the dividend is another, more extreme option, but also sensible.

2. Once below 4.5x, pursue a dividend strategy with a payout ratio below 100% unless leverage falls below 2.5x.  

3. Do not increase gross debt in the near term to accomplish any capital returns.  All accelerated capital returns must come from asset sales and/or FCF.  

Cogent is not an insolvent company.  It can materially alter its leverage picture overnight by changing its dividend policy.  The pain of such a policy change was something to bemoan at $70/share, but there is nothing to fear at $37/share.  Clearly, the stock already anticipates it with an 11% dividend yield.

If wavelength revenue ramps as Dave expects from here, Cogent will deleverage rapidly and will have an opportunity to return to its prior dividend path.  Temporarily deviating from that path to implement a responsible capital allocation policy may only cost Cogent shareholders a few dollars of dividends per share (or less) before the company could return to its prior path.  However, the confidence the market would gain from responsible policy likely would be worth many multiples of those foregone dividends.

Asset Sales

Closely related to Cogent’s dividend policy are the asset sales.  I advocate urgently selling fallow assets, particularly IPv4 address space.  I know the counterargument Dave has given - that the market isn’t very deep for the space Cogent would like to move.  However, I don’t care - there is a clearing price.  Even if Cogent had to sell its Sprint space at $25/address, it would be worth it to get $250M of cash and execute an ASR.  Retiring 5-6M shares would materially enhance long-term shareholder value and reduce the current dividend burden by $20-25M per year.  

IPv4 addresses are relatively liquid assets, and so is Cogent’s stock. Even at “low” IPv4 prices, the math is too compelling - the board should take liquidity from the IPv4 market and use it to buy Cogent’s stock.  The lease-up rates in the IPv4 business are slow enough that the last ~10M addresses have minimal DCF value.  There is low opportunity cost vs. selling them today.

I expect something to transact on the data center facilities, but that market is far less liquid.  A year ago, Dave and the board made a bet on the “frenzy” in the data center market by pulling forward their capex plans to retrofit the data centers, but ultimately the bet will not pay off as well as they had hoped.  I don’t believe the $100M of capex will be lost permanently, but the market for smaller facilities is not participating in the frenzy we are seeing for massive AI farms.  Making a bet and getting it wrong is okay - but it’s better to reverse out expeditiously and move on.  I hope Cogent will exit its data center assets without being married to unreasonable price expectations.

Closing

The delays in Cogent’s wavelength revenue ramp have put the company in a more highly leveraged position with slower growth and lower EBITDA and FCF than previously thought at this moment in time.  However, I am optimistic that we are mostly experiencing a lag in timing, but not a significant deviation from the ultimate, multi-year outcome for the business.  Timing delays do not bother me; I would be much more concerned if the network didn’t perform well on one or several important dimensions of value that were contemplated at the outset of this transformation project.  

The stock’s significant underperformance gives the company an exciting opportunity to add value for long-term shareholders if it can liquidate fallow assets and execute buybacks at these depressed prices.  I hope the company will be simultaneously responsible and opportunistic with its capital allocation policy going forward, as discussed above.