Cogent Communications - The Wavelength Opportunity

This post is about optical transport services, a specific and technical part of wireline telecom that is not well known to most investors, but is critical network infrastructure for cloud infrastructure providers (AMZN, GOOG, MSFT), AI companies (OpenAI, Anthropic, etc.), content companies, and other cable/telco providers (CMCSA, CHTR, etc.). 

It is an important end market to Cogent (NASDAQ: CCOI), one of Recurve's portfolio companies, but also to Lumen (NYSE: LUMN) and Zayo (owned by Digital Bridge, NYSE: DBRG).  In this post, we will give some background on Cogent's acquisition of Sprint Wireline, discuss the optical transport market, and finish with some thoughts on why it should create a step-change in Cogent's free cash flow profile over the coming years.

Background

Cogent recently acquired Sprint's wireline business from T-Mobile on May 1, 2023, and we think it will be transformative for the company.  This acquisition brings three areas of value to Cogent:

1. Lightly utilized assets consisting of:

• A nearly empty physical fiber network with over 19,000 route-miles of long-haul, inter-city fiber (interesting story on the network here) and over 1,200 miles of metro (intra-city) fiber.

• A portfolio of technical facilities, some of which are used to run the fiber network, and 45 of which will be converted into co-location data center space with about 400,000 square feet of raised-floor space and ~50 megawatts of accessible power.

• 9.75m IPv4 addresses. These are currently transacting at $40-60 per address (see here for more detail).

2. A legacy wireline B2B business which, at closing, was generating about -$220m of free cash flow (-$190m of EBITDA, $30m of capex), but will eventually generate about $50m of free cash flow after Cogent restructures the business and extracts visible and low-risk network synergies. 

3. $700m of cash payments from T-Mobile, comprised of $350m paid monthly in the first 12 months, and $350m paid monthly over the following 42 months. 

There are many pockets of value from this acquisition, but the topic of this post is to dive into the empty long-haul fiber network.  Cogent is in the process of converting the fallow, unconnected network into a national network upon which it can sell optical transport services, a large and growing >$4 billion end market driven by demand from cloud service providers, other hypserscale tech companies, content companies, and cable/telecom companies. 

Now, let's walk through what optical transport is and de-mystify part of the black box within wireline B2B services. 

Optical Transport - What is It?

Optical Transport (also called Wavelengths), is dedicated, point-to-point fiber-based communication between two specific locations.  It is purchased as a service and operated by companies that own and operate physical fiber networks. 

Buyers of optical transport need these large, dedicated pipes to connect their servers and networks at much higher service levels than could be provided by the broader internet.  Sending data over wavelengths sidesteps the internet and provides dedicated, agreed-upon capacity, service levels (e.g. 100 gbps with <5 ms latency) and paths, as well as service level agreement (SLA) protocols for what happens in the event of service disruptions (this is important since the physical path is specific).  It's a premium data pipe for targeted, customer-specific use cases - not for general purpose data transfers which the internet fulfills well and at much lower cost.  Internet protocols look for the shortest logical hop from the current position on the way toward the end destination, but do not have SLAs on speed, latency, path, or otherwise. Cogent is one of the world’s leading internet transit companies, carrying about 25% of global internet traffic. It knows the wholesale connectivity markets well.

Wavelengths are provisioned by buying cross-connects within a carrier-neutral data center from a customer's servers (e.g. AWS) to the transport provider's (e.g. Cogent) racks in each location.  Once those connections are made, the customer can move data between those two CNDC endpoints through that provider's dedicated wavelength service using Optical Transport Network (OTN) protocols. 

Now, let's talk about who needs and buys wavelengths in the market:

1. Hyperscalers and cloud service providers buy wavelengths to connect their data centers and availability zones for large data transfers.  They use a combination of metro (intra-city) and long haul (inter-city) fiber-based solutions.  Sometimes they build physical fiber, sometimes they lease fiber via long-term dark fiber IRUs, and sometimes they buy optical transport services from other fiber providers. 

2. Regional access cable/telco providers buy wavelengths to connect their "islands" of service to each other and/or to build/augment their core networks.  For example, if Comcast operates network assets in the Bay Area and Los Angeles, but not the Central Valley, it would may not have contiguous network assets that connect northern and southern California.  It would need to connect those "islands" of service with long-haul fiber service via owned and operated physical fiber, dark fiber leases, or wavelengths. 

3. Other content companies (CDNs, video game companies, media companies) use wavelengths to replicate their content across different zones for faster/lower latency local delivery to end customers.  Lately, Cogent has seen an influx of new demand from companies in the artificial intelligence industry which ingest enormous volumes of data to train their large language models (LLMs).

4. Federal, state, and local government agencies buy transport for secure file transfers outside the internet.

5. Some (but few) large enterprises, universities, and other major institutions use wavelengths for large file transfers between and among their campuses.

Optical Transport Suppliers and TAM

Now, let's talk about the suppliers. 

The national long-haul fiber networks in the US are owned and operated by five entities: Lumen, Zayo, AT&T, Verizon, and Cogent/Sprint.  Additionally, there are regional providers like Crown Castle, Cox Business, Windstream, Arelion, Frontier, and others.  Vertical Systems Group shows the wavelength rankings in 2022 in the chart below (note: Cogent did not sell wavelengths in 2022):

Source: Vertical Systems Group

Cogent estimates that the long-haul optical transport market is about $2 billion of annual spend and that Lumen and Zayo control 90% of the market, with Lumen being "far and away the largest."  These companies don't break out their wavelength contributions specifically.  In addition to long-haul, there is another >$2 billion of metro, intra-city wavelength spend to connect facilities within the same metro area.  We believe Cogent will disintermediate some of that metro wavelength spend.  Let's explain how and why through a series of maps. 

Below is a map of Cogent's city footprint in the US, overlayed with the Sprint fiber network map (not a perfect juxtaposition of two different maps, but you get the idea).  These are the markets in which Cogent has a presence in carrier-neutral data centers (CNDCs) and a map of the physical long-haul routes used to connect them. Cogent connects to about 800 facilities around the country.

Source: Google Maps, Cogent Communications, Recurve Capital

Several wavelength providers can offer connectivity between Chicago and Seattle, but they may not terminate in the exact endpoints desired by the customer.  In those cases, that customer will purchase a metro wavelength to connect the long-haul endpoint to its local endpoint in that market. Using an air travel analogy, it’s like having to take a connecting flight to get to downtown Chicago from O’Hare.

Let's zoom in on Chicago to get a sense for how diverse the endpoints are within a single market. Using our travel analogy, Cogent will be able to provide direct air travel to nearly all destinations - not just to the hub. This will disintermediate a portion of the short-haul, intra-city wavelength market which should provide a competitive advantage for Cogent competing against a more expensive, more technically complex, and slower provisioning solution.

Source: Google Maps, Cogent Communications, Recurve Capital

We can see that Cogent offers connectivity in dozens of data centers in Chicago. The same is true for all major markets in the US. Cogent populates the most CNDCs in the country and is the most inter-connected network in the world.

Even though the total spend in the wavelength market is over $4 billion per year, we think Cogent will disrupt a portion of the metro wavelength market.  It will be good for Cogent and potentially disruptive for incumbent providers.

Finally, let’s talk about typical wavelength contract terms. Wavelengths are priced based on the parameters of their SLAs - capacity, distance and path of the route, latency, and more.  Prices for 100 gbps waves tend to be about $2,000-3,000/month on 1- to 3-year contracts (3 year is most popular) and 400 gbps waves tend to cost $5,000-6,000/month.  Like-for-like prices decline about 10-15% per year, but customers also upgrade to higher capacity wavelengths over time as their needs evolve.  The market is expected to grow in the mid- to high-single digits from a combination of volume growth (more routes) and capacity upgrades by customers (more ARPU), offset by like-for-like price reductions. Theoretically, about 1/3 of the market should be up for grabs every year, but it does seem that some growth-oriented customers (like the hyperscalers) are reluctant to switch providers, preferring instead to add capacity at every stage. It too early to know with certainty if market share will shift on contract expirations, or if share will shift primarily through share of volume growth. At this stage, our best guess would be that less than half of annual contract renewals would seriously consider switching.

Cogent's Opportunity

Cogent is repurposing the nearly-empty network it acquired from Sprint into an optical transport network.  It will take some time and investment (much of which is spent already) to integrate the Cogent and Sprint assets so that all 800+ CNDCs are available for transport services, but the company expects to get there by year-end 2024.  Along the way, it is already taking orders and booking wavelength contracts in the limited footprint it has today, although it will take some time to convert them into revenue given elongated installation times during the network integration phase.  We believe Cogent has an opportunity to take significant share of new wavelength orders coming into the market, and also has a good chance of winning some existing routes from customers as well. 

Additionally, the Sprint network has 90% physically unique paths which is important for customers that need route diversity for resilience and redundancy - especially the cloud service providers.  For example, many wavelength buyers I have spoken to are excited to get access to the Chicago - Seattle route specifically because Cogent's path through North Dakota and Montana is physically unique.  This means line cuts that impact one provider likely will not impact Cogent’s services. Cogent will be able to market a full nationwide network that has been built and managed almost exclusively for Sprint's first-party traffic up until this acquisition.  It has routes and paths that customers are eager to access to add capacity and resilience to their networks.

For Cogent, the optical transport market alone should contribute north of $10/share of incremental free cash flow over the next 5-7 years.  Cogent has guided investors to targets of linear growth to $100 million of run-rate Wavelength revenue in mid-2024 (from $8 million at closing) and $700 million of revenue in 7 years, all of which comes at 95% incremental EBITDA margins. Early demand trends have been healthier than those targets.  The company already has the largest wholesale B2B sales force in the industry from its IP transit business and it has relationships with all the major customers that buy optical transport services.  This is a rare and unique opportunity for a company to create a step-change in growth from a completely new but adjacent, non-cannibalistic business - all at a negative cost basis due to T-Mobile's cash payments.

Cogent's run-rate revenue targets imply about 3,000-3,500 wavelength contracts for $100m of revenue (at ~$2,500 average MRR per wave) and 20,000-25,000 wavelength contracts for $700m of revenue.  It sounds like a lot, but it will be able to sell n(n-1)/2 unique routes, where n is equal to the number of connected CNDCs.  Cogent will have about 320,000 unique point-to-point routes available for sale. 

Looking at it from a selling perspective, Cogent has about 250 reps that sell to wholesale buyers of connectivity.  If each rep sells one net new wavelength contract per month (again, at $2,500/month of MRR), Cogent would be booking $7.5 million of annual recurring revenue (ARR) per month, or about $90 million or ARR per year.  Lately, reps have been selling and installing 4-5 services per month (across the product portfolio), but there have been periods when they averaged 6-7 services per rep per month.  This assumes no growth in the sales force, but it tends to grow MSD to HSD annually. In short, Cogent has the sales rep capacity to handle another 1-2 wavelengths per rep per month which would allow the company to reach its targets. 

Of course, it's easy to look at spreadsheet math and see how much cash flow can be generated from this market.  It's important to remember that Cogent must first cross some key milestones in its network integration with Sprint's fiber network before it can access the whole market.  Significant progress will be made over the course of the next 12-18 months, at which point Cogent's wavelength offering should be at the peak of its disruptive powers.  Once fully integrated, Cogent should be able to win on various dimensions. Below we outline the company’s competitive advantages:

1. It will be able to provision wavelengths faster (2 weeks vs. 3-9 months for competitors).

2. It will have a more ubiquitous network with full distribution to 800 CNDCs (vs. 350 for Zayo and Lumen)

3. It provides network resilience from its 90% unique physical paths

4. It entered this business at a negative cost basis and will never be undersold

5. It has the largest and most aggressive sales force in the wholesale connectivity market

Closing

We love to find the structurally faster/better/cheaper, disruptive companies in their end markets because if our understanding of their advantages are correct, there is significant market share to gain.  Cogent already is faster/better/cheaper and disruptive in its classic Cogent markets, but this acquisition gives it access to another avenue of growth that should be easier to penetrate given its existing business, brand, and customer relationships.  We hope this post provided some helpful insights into the exciting opportunity in front of Cogent over the coming years. 

Insights