Cogent Communications

Cogent delivered a disappointing second quarter. Revenue declined 4.3% year over year and 1.5% sequentially, with weakness across customer types. EBITDA declined from a year ago, and free cash flow was negative, even with the payments from T-Mobile. Data center sales helped the balance sheet.

Why it matters: We could look past weak growth given the ongoing decline of noncore legacy Sprint revenue, but key growth areas, notably wavelength services, were very disappointing. The number of net new wavelength circuits in service should be increasing each quarter, but it is actually falling. Management blamed wavelength performance on customers' inability to get needed equipment. The excuses are wearing thin. The firm claimed it had 6,000 circuits installed or in the pipeline a year ago, but it ended the quarter with only 2,445 installed. Cogent also claims it is selling wavelengths to all major AI-related firms, but that makes the lack of growth acceleration all the more concerning. The pace of net corporate customer connection losses also accelerated to the worst level in two years.

The bottom line: We now expect revenue to decline 4% in 2026 rather than 2% and cut our expectation for wavelength revenue in 2030 by nearly 30% to $205 million. As a result, our fair value estimate goes to $17 from $25. We assign Cogent a Very High Uncertainty Rating, but we view the shares as overly punished. Margin weakness during the second quarter is somewhat concerning, but there's a lot of noise in Cogent's cost base. The firm reduced headcount by 6% during the quarter, likely reflecting a weaker growth outlook, but SG&A costs were roughly flat. This likely reflects severance costs that should benefit future periods. We've increased assumed net proceeds from asset sales to $350 million from $300 million with the successful closing of the sale of 10 data centers during the quarter. The firm has used the proceeds from this sale to repurchase its debt at a discount.

Our $17 fair value estimate attempts to balance the risks Cogent faces with the opportunities it is pursuing to unlock value. Our valuation implies an enterprise value/EBITDA multiple of about 11 times our expectations for 2028 EBITDA, the first year after the payments from T-Mobile cease. We also assume Cogent is able to realize another $100 million in net asset sales over the next year or so, above the proceeds it received in June 2026 from the sale of 10 data centers. Continuing to sell assets should alleviate some of the balance sheet risk investors now face.

We think Cogent will drive revenue growth from new customer connections and opportunities, such as wavelengths, as it moves beyond the loss of acquired Sprint business. In total, we forecast revenue to grow at about 4% annual rate through 2035, including a 4% decline in 2026 as unprofitable Sprint business runs off.

In Cogent’s corporate business, where it provides internet and private networks to companies located within its over 1,800 connected multitenant office buildings, we expect revenue to decline another 9% in 2026 and then grow at an average of 4% per year, as increasing customer connections are partially offset by deflationary pricing.

Cogent’s netcentric business is positioned to capitalize on the growing demand for connectivity around AI. We expect Cogent’s core IP transit service to benefit from growing demand that more than offsets the deflationary nature of the business. Additionally, the most significant opportunity for Cogent is its wavelength business, which it can now offer through its acquired long-haul fiber network from Sprint. Cogent estimates that the intercity US wavelength market produces $2 billion of revenue annually, and management has announced its target of capturing 25% of this market. This target now looks even more ambitious with failure to see accelerated adoption in early 2026, which management has blamed on equipment delays.

We think wavelength services will be a material driver for Cogent, but we have materially reduced our expectations for this service. We assume total wavelength revenue of about $205 million in 2030, which would leave Cogent far short of its market share goals. We now expect annual total netcentric revenue growth of less than 8% through 2035, down from our 9% expectation earlier this year.

Cogent’s margins are still depressed—the 30% adjusted EBITDA margin in 2025, which includes $100 million in payments from T-Mobile that end in 2027 and excludes stock-based compensation, is well below the 38% margins Cogent generated between 2020 and 2022. Cogent recorded only a 17% EBITDA margin in 2025 when excluding these payments and including stock-based compensation. We expect the focus on wavelengths and other on-net services will structurally alter the firm’s margin profile. We model EBITDA margins of 26% in 2030 and 32% in 2034.

We project capital spending to be low over our forecast. When including the value of new IRU leases, Cogent's primary means of obtaining new fiber assets, we assume capital expenditures to decline below 10% of sales by 2035, from more than 35% over the past three years. However, we have increased our near-term capital spending expectations due to higher equipment prices.

We assign Cogent a narrow moat rating based on cost advantages stemming from its savvily acquired fiber network that spans 95,000 intercity and more than 35,000 metro-area route miles. Cogent’s competitive success can be attributed to this network, which is composed mainly of smaller fiber assets that it acquired or leased at steep discounts following the telecom bubble in the early 2000s. Cogent’s network operates at a significantly lower capital cost than rivals that have built networks or acquired assets at higher prices.

Cogent’s 2023 acquisition of Sprint’s legacy fixed-line business has altered its trajectory. T-Mobile is paying Cogent $700 million in structured payments through 2027 to take this money-losing business off its hands. The new Sprint business has hurt Cogent’s returns on invested capital, which averaged about 20% from 2018-22 but has been barely above breakeven recently.

However, we do not expect the deal to be value-destructive over the long term—we expect Cogent to continue canceling the most uneconomic Sprint-related contracts while integrating the network assets it has acquired. We forecast returns on invested capital to reach the firm's cost of capital by 2031 and to continue improving into the midteens beyond that.

Cogent’s most significant advantage stems from its low-cost network. Rather than directly investing in its network by laying thousands of miles of fiber, a very capital-intensive process, Cogent has focused on acquiring inexpensive fiber assets and leasing indefeasible rights of use (IRUs) from businesses that have already laid dark fiber. Cogent was founded in 1999 at the peak of the telecom bubble but made its name during the telecom bust from 2001 to 2004, when it acquired 13 distressed companies at steep discounts. Management notes it spent $60 million and assumed about $800 million in debt and other obligations to acquire $4 billion in property, plant, and equipment (PP&E).

Additionally, Cogent has continuously pursued capital leases, or IRUs, to expand its fiber network, rather than build it itself. Its IRU portfolio has nearly $600 million in net assets and is contracted with over 350 vendors, with an average remaining term of nearly 17 years. Cogent has developed a network spanning more than 130,000 route miles, with an average cost of $25,000 per mile based on gross PP&E. Comparatively, Lumen, Cogent’s closest competitor, has $128,000 per fiber route mile in gross PP&E and is likely around $50,000 per mile if we include its legacy copper network. Cogent's lower-cost network has allowed it to undercut rivals on price while still generating solid returns on capital.

Beyond its more efficient capital base, Cogent’s network leads to an operating cost advantage. Its competitors' networks are built to capitalize on different services, such as voice, and include legacy technologies, such as MPLS-based VPNs (multiprotocol label switching-based virtual private networks). Alternatively, Cogent’s network is structured to solely provide internet connectivity. As internet-based technologies continue to replace legacy services, Cogent’s simpler cost structure positions it to price accordingly.

The legacy Sprint business burns cash, but Cogent should be able to continue driving efficiencies by sunsetting legacy services and upselling Sprint customers to modern, high-capacity services. Furthermore, Cogent can capitalize on new revenue streams, including high-capacity services like wavelengths, dark fiber leasing, and IPv4 address leasing. Lastly, the Sprint acquisition includes an extensive real estate portfolio that Cogent is in the process of selling.

With the Sprint network, Cogent is well-positioned to capitalize on growing AI-related demand for high-capacity connections between data centers.

Cogent's low-cost structure, a byproduct of how it built its network, prevents other firms from competing profitably on price for comparable services.

Cogent is sitting on a trove of hidden assets acquired with Sprint, including IPv4 internet addresses and a huge portfolio of excess real estate.

The Sprint business is a far bigger disaster than Cogent anticipated. Three years after closing the acquisition, EBITDA remains far below pre-deal levels, absent T-Mobile support payments.

Cogent has failed to deliver on the promised growth from wavelength services. It has also failed to find buyers for most of Sprint's real estate, despite a hot market for data center space.

The recent dividend cut doesn't bode well for Cogent's business or for future asset sales.