Cogeco CommunicationsTSX:CCACommunication Services
The US cable business is genuinely deteriorating and the market is right to worry about it. The question is whether that worry has been extrapolated into a dividend it does not yet threaten.
Price at publication
CAD 58.77
Closing price, Toronto Stock Exchange
Yield at publication
6.72%
Base sustainable DPS growth 5.0–6.0%
Yield at publication
6.72%
Base sustainable DPS growth
5.0–6.0%
FCF payout
29.9%
FY2025
EV/EBITDA
5.0x
FY2026E consensus
5Y dividend CAGR
9.1%
10Y dividend CAGR
9.7%
Cogeco Communications is a broadband operator with two very different halves. The Canadian half is competitive, growing modestly and highly profitable. The American half, Breezeline, is losing customers faster than its much larger peers, is under visible pricing pressure, and carries the scar of an acquisition that did not work. In fiscal 2026 the company wrote down roughly C$2.2 billion of that American asset base.
The share price has fallen from around C$110 in 2021 to under C$60. Over the same period the dividend rose every year. That divergence is the subject of this report.
The proposition is not that the market has invented a problem. It is that the company’s dividend economics currently look stronger than the narrative attached to its business, and that the gap between those two things may be wider than the evidence justifies.
Cogeco sells monthly broadband, and increasingly mobile, to about 1.6 million residential and business subscribers. Fiscal 2025 revenue was roughly C$2.9 billion. Revenue is recurring rather than transactional, which is what makes a dividend possible at all in a business this capital-hungry.
Canada, trading as Cogeco and oxio. Internet customers and segment EBITDA have continued to grow. Margins sit in the low-to-mid fifties. Mobile is being used as a convergence and retention tool and has been running ahead of management’s internal plan. This half of the company is not currently exhibiting the problems visible in the United States.
The United States, trading as Breezeline and welo. Margins remain high, near fifty per cent, but Internet customers declined materially year over year and US revenue fell substantially in the third quarter of fiscal 2026. The footprint spans thirteen states and includes Ohio, where systems acquired in 2021 have performed far below the original case.
Two structural features matter for anything that follows. The first is capital intensity: cable and fibre networks require sustained investment, and the relevant question is never whether capital expenditure can be cut for a year but what level is required to keep customer economics intact. The second is that competitive advantage here is real but narrower than it was. Sunk infrastructure, regional scale, existing customer relationships and high operating leverage still count. Fibre overbuild, fixed wireless access and easier switching have reduced what they are worth.
Broadband demand is durable. The economics of cable broadband are less durable than the demand for it, and those are different claims.
The pressure is industry-wide before it is company-specific. Charter reported 172,000 Internet customer losses in the second quarter of calendar 2026, with Internet revenue down 3.2% year over year and an adjusted EBITDA margin around 40.3%. Comcast reported 167,000 domestic residential broadband net losses in the same quarter. Two of the largest cable operators in the world are losing broadband subscribers.
That context matters, and it does not excuse Breezeline. On the evidence available, Breezeline’s recent customer and revenue trends have been worse than those large peers. The honest reading of the peer cross-check is four answers rather than one:
Reliable market-share series across all of Cogeco’s geographies are not available in a consistent form, so this report uses relative subscriber growth and operating performance as proxies rather than claiming a literal share figure. On those proxies, Canada is holding or gaining ground and the United States is not.
Competition in Ohio specifically is not hypothetical. Brightspeed reported in June 2026 that its Ohio fibre network had reached more than 350,000 homes and businesses, with roughly 200,000 further locations planned. That cuts both ways, and we return to it.
The dividend is the reason this company is being examined at all.
Dividend per share
1.56
FY2016
1.72
FY2017
1.90
FY2018
2.10
FY2019
2.32
FY2020
2.56
FY2021
2.82
FY2022
3.10
FY2023
3.42
FY2024
3.69
FY2025
3.95
FY2026
The current quarterly dividend is C$0.987, declared most recently on 15 July 2026, giving an annualised ordinary rate of C$3.948. Against the publication reference price of C$58.77 that is a yield of 6.72%.
The reconstructed record runs from roughly C$1.56 per share in fiscal 2016 to C$3.948 today. Growth has been consistent and is decelerating in an orderly way: about 7.05% in the most recent year, an 8.35% compound rate over three years, 9.05% over five and 9.73% over ten. The reconstructed fiscal 2010 to fiscal 2026 series contains sixteen consecutive annual increases and no cut or freeze. No special dividend is included in any of these figures.
The deceleration is not a warning sign in itself. Increases of roughly 10%, then 8%, then 7% are what a board does when it wants to keep raising a dividend while the operating environment gets harder and leverage still matters. It is a more rational pattern than holding the growth rate constant would have been.
| C$ m, except per share | FY2016 | FY2017 | FY2018 | FY2019 | FY2020 | FY2021 | FY2022 | FY2023 | FY2024 | FY2025 | FY2026 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Ordinary DPS | 1.56 | 1.72 | 1.90 | 2.10 | 2.32 | 2.56 | 2.82 | 3.10 | 3.42 | 3.69 | 3.95 |
| Revenue | 2,176 | 2,227 | 2,424 | 2,332 | 2,384 | 2,510 | 2,901 | 2,984 | 2,977 | 2,910 | — |
| Adjusted EBITDA | 983 | 1,005 | 1,086 | 1,108 | 1,149 | 1,206 | 1,393 | 1,421 | 1,442 | 1,443 | — |
| EBITDA margin | 45.2% | 45.1% | 44.8% | 47.5% | 48.2% | 48.0% | 48.0% | 47.6% | 48.5% | 49.6% | — |
| Free cash flow | 281 | 374 | 326 | 454 | 455 | 487 | 424 | 418 | 476 | 517 | — |
Two things stand out in that table. Free cash flow has grown faster than revenue over the period, at roughly 7% compound against 3.3%. And the dividend has been funded from a modest share of it throughout, rather than from an expanding payout ratio.
This is where the report has to be explicit about causation, because the answer is a number and the number is only as good as the chain behind it.
Dividend capacity is the last link, not the first:
Our base case is 5–6% annual dividend growth for the next several years, moving to roughly 3–4% thereafter. That second number matters as much as the first. Extrapolating the historical high-single-digit rate indefinitely would require believing that the last decade’s economics are the next decade’s, and this report does not believe that.
What has to happen in the business for 5–6% to hold. Canadian Internet customers grow at around 2–3% with segment revenue up 1.5–2.5% and margins around 53–54%. US subscriber losses narrow over two to three years toward flat, US revenue per customer stops deteriorating, and the US margin holds around 49%. Recurring capital intensity excluding expansion stays near 16–17% of revenue. Residual free cash flow goes to debt rather than to buybacks or acquisitions. On that path fiscal 2031 free cash flow per share lands somewhere around C$14.50–15.50 and net leverage falls to roughly 2.5–2.8 times.
Note what is not required. No consolidated margin expansion. No return to American growth. No buybacks. No rerating. The base case asks the American business to become less bad, not good.
What the bear case looks like. US subscriber decline persists at 2–3% a year, monetisation stays weak, US revenue falls 3.5–5% annually, margins drift to 47–49%, capital intensity runs at 18–19%, and Canada remains only modestly positive. Fiscal 2031 free cash flow per share is then around C$9.50–10.90 and sustainable dividend growth is 0–2%. The dividend is not cut in that world. It stops being a growth dividend.
What the bull case requires. Faster American stabilisation, continued Canadian strength, capital intensity in the mid-teens, and leverage below roughly 2.7 times unlocking buybacks that add to per-share growth. That supports 7–9% initially. It does not require a return to the monopoly-like cable economics of the past, and it does not require a 7x EBITDA multiple.
These three bands are the output of that chain, not an input to it. They were not chosen and then justified.
The derating began around 2021 and 2022 and has a clear list of causes: the Ohio acquisition raised US exposure and leverage, the expected US growth did not arrive, fibre and fixed wireless competition intensified, US customer and revenue trends deteriorated, and investor perception of cable shifted from stable compounder to declining infrastructure. The fiscal 2026 impairment of roughly C$2.2 billion put a number on part of that.
Every item on that list is a real change in the business. The question is what has changed only in expectations.
Canada is doing more of the work than the narrative allows.
The story attached to this company is an American one, but the Canadian segment is where most of the dividend capacity is generated and it is not currently showing American-style deterioration. Canadian Internet customers and segment EBITDA have continued to grow, and mobile is running ahead of plan. A price that assumes consolidated decline has to assume Canada follows, and there is not yet evidence that it is following.
The dividend is insulated by how little of the cash flow it consumes.
Fiscal 2025 free cash flow was about C$517 million against roughly C$155 million of cash dividends, a payout near 30%. At the current rate the annual cash cost is about C$167 million. Even if normalised free cash flow fell to C$350 million, the dividend would absorb under half of it. That buffer is not an argument that the dividend can grow; it is an argument that a growth disappointment and a dividend cut are separate events.
The American business may only need to stabilise, not recover.
Our base case does not assume Breezeline returns to growth. It assumes subscriber losses narrow toward flat over two to three years and revenue per customer stops falling. That is a materially lower bar than the one the share price appears to be failing, and Ohio has now produced four consecutive quarters of positive Internet subscriber growth against an active fibre overbuild.
Debt reduction is a return mechanism that does not require the market to change its mind.
With the dividend absorbing under a third of free cash flow, most of the residual cash can go to the balance sheet. Lower leverage reduces the interest burden and improves per-share dividend capacity mechanically, whatever the multiple does. If the shares stay cheap while leverage falls, later buybacks become more valuable rather than less.
Partly, and the strongest version of the bear case deserves to be stated in full rather than summarised into something easier to dismiss.
It runs like this. Cable broadband has permanently lost part of its moat. Breezeline will keep losing customers to fibre and fixed wireless. Stabilising subscribers requires lower prices, so US margins eventually follow revenue down. Defending the network requires materially more recurring capital. Canada eventually faces the same pressure with a lag. Free cash flow stagnates or falls, leverage stays elevated, dividend growth drops to zero, and the shares therefore deserve a persistent 6.5–8% required yield and a 4.5–5x EBITDA multiple.
That is not a caricature. Material parts of it are already visible in reported data: US customer losses that are worse than large peers, a US revenue decline greater than Charter’s or Comcast’s broadband decline, an active fibre build in Ohio, management’s own statement that US financial improvement is slower than expected, a C$2.2 billion impairment confirming that prior asset economics were overestimated, regulatory evidence that some fibre spending replaces existing plant rather than expanding it, a company-defined free-cash-flow measure whose definition has changed, and a history of large debt-funded acquisitions.
Our interpretation is narrower than a disagreement with any of that. It is that the price appears consistent with a No Recovery operating case rather than with our base case, while the cash cost of the dividend remains low relative to normalised free cash flow. The market does not have to be wrong that Breezeline is weak. The thesis only requires it to be too pessimistic about one or more of Canadian durability, American stabilisation, capital intensity, debt reduction, or long-term dividend capacity.
This is an inference drawn from evidence, not a measurement, and we hold it with medium confidence. A reader should be able to finish this report still believing that Cogeco deserves a 6.5–8% yield if US cable decline proves structural and capital intensity stays high. The report is stronger for showing that, not weaker.
A note on Ohio. Four consecutive quarters of positive Internet subscriber growth, achieved while Brightspeed builds fibre across the state, is meaningful evidence that stabilisation is achievable. It is not evidence that the acquisition thesis has been repaired, and this report does not claim it has. Ohio is a test we are watching, not a result we are relying on. Subscriber growth bought through promotional pricing is not economic stabilisation, which is why revenue per customer matters more than the subscriber line.
Near-term coverage is the strongest part of this case, and it should still be stated carefully.
Fiscal 2025 free cash flow was approximately C$517 million against approximately C$155 million of cash dividends, a payout of roughly 30%. At the current annualised rate the dividend costs about C$167 million a year. Run that against weaker outcomes: at C$350 million of normalised free cash flow the dividend absorbs about 48%; at C$250 million, about 67%; at C$220 million, about 76%.
The dividend survives our Delayed Recovery and No Recovery cases without requiring a cut. That is the finding, and the sequence it implies matters more than the finding itself:
A failure of the growth case is not the same event as a dividend cut. The realistic path is slower dividend growth, then a possible freeze, and only under materially worse conditions than our bear case does cut risk become real.
Why a low payout is not by itself sufficient. The denominator is the fragile part. Free cash flow is only meaningful after economically required network investment, and a company can flatter its payout ratio for several years by underinvesting. In a business migrating from hybrid fibre-coaxial plant to fibre, the line between maintenance and growth capital is genuinely blurred, and neither we nor any outside analyst can observe it precisely. A 2026 CRTC decision documents Cogeco replacing hybrid fibre-coaxial facilities with fibre-based EPON at specific Ontario locations, which is direct evidence that some fibre spending is replacement rather than optional expansion.
The arithmetic of that risk is unforgiving. On roughly C$2.85 billion of revenue, one percentage point of capital intensity is about C$28.5 million of annual cash flow. A shift from 16.5% to 20.5% recurring capital intensity would consume something like C$110–120 million a year, which is most of the current dividend. Low payout and structurally higher required capital expenditure are not compatible assumptions, and anyone relying on the first should be watching the second.
The research thresholds we will use later, and which readers can check against reported figures: dividend growth has to stop somewhere around C$300–350 million of normalised annual free cash flow, particularly with leverage above 3.5 times; the dividend itself becomes questionable around C$220–250 million combined with leverage above 4–4.5 times, declining EBITDA and elevated required capital expenditure. These are research judgments about when paying the dividend would start to compete with the durability of the business, not covenant limits.
The balance sheet. Net debt was approximately C$4.44 billion around the third quarter of fiscal 2026, net leverage roughly 3.2 times, weighted average cost of debt about 5.4% and average term about 3.8 years, with liquidity above C$0.9 billion and investment-grade senior secured ratings. The principal refinancing concentration sits in 2028, where two US facilities of roughly US$828 million and US$464 million mature. In August 2026 the company reopened its 5.299% senior secured notes due 2033 for a further C$200 million at a reopening yield of 4.565%, which tells us access to debt markets is intact and that the 2028 issue is one of cost and flexibility rather than of availability. Reducing debt before then remains economically valuable, and that is one reason the base case assumes residual cash goes to the balance sheet rather than to shareholders.
Leverage above 4 times while EBITDA and free cash flow are still falling is the combination that would change this assessment. Below 3 times is comfortable; 3 to 3.5 is a watching brief, which is where the company sits today.
This section carries more weight here than it usually would, because the valuation discount is partly a judgment about capital allocation rather than about operations.
The record is mixed and the worst item is recent. DERYtelecom in 2020, at C$405 million for roughly C$105 million of revenue and C$44 million of EBITDA in an adjacent Canadian footprint, was a reasonable strategic fit. MetroCast in 2017 and 2018, at US$1.4 billion and a tax-adjusted multiple around 9x EBITDA financed with substantial debt plus a US$315 million equity investment from CDPQ, was aggressive but coherent at the time. The Ohio systems acquired from WideOpenWest in 2021, at US$1.125 billion financed with roughly US$900 million of secured debt, were bought on a thesis of superior growth and attractive demographics that subsequent performance did not deliver. The fiscal 2026 impairment confirms material destruction of value against that thesis. We record it as a major capital-allocation failure, and nothing in this report treats it as a footnote.
The buyback record is better than the acquisition record. The 2022 normal course issuer bid repurchased 817,735 shares at a weighted average of C$82.46, which looks poor at today’s price though it should not be judged only with hindsight. The December 2023 repurchase and cancellation of 2,266,537 shares at C$51.40, a 10% discount to the prior close and roughly 5.1% of shares outstanding in a single transaction, was strongly accretive if the underlying business retains value. Share count has fallen roughly 14% over the long reconstruction, from about 49 million to about 42.3 million.
What the thesis assumes about the future is different from what the past shows. The current inferred capital-allocation hierarchy puts required network investment first, then the dividend, then debt reduction, then selective organic growth, then buybacks, and acquisitions last and only if highly disciplined. Recent behaviour is consistent with that. But this thesis depends in part on management not repeating the Ohio decision, and that is an assumption about future conduct rather than a conclusion from evidence.
The kill criterion is explicit: another large debt-funded acquisition before leverage is comfortably below roughly 3 times would materially damage this thesis, whatever the strategic rationale offered for it.
Valuation is secondary to the dividend case here, and the price still has to be attractive.
The starting point is what a business like this should yield now, not what it used to yield. Cogeco’s trailing yield ran between roughly 1.9% and 3.0% from 2016 to 2021, then 3.76% in 2022, 5.36% in 2023, 5.17% in 2024 and 5.65% in 2025, giving a five-year average near 4.3% against about 6.7% today.
A high yield relative to that history is not evidence of undervaluation. The argument that the yield should return to 4.3% would require believing that nothing economically relevant has changed, and a great deal has: the American business is weaker, the moat is narrower, leverage is meaningful, required returns are different, and capital intensity may be structurally higher. Historical yield is evidence about the past, not a law about the future.
Our normalized yield assumptions are therefore deliberately well above that average: 6.5–7.5% in the bear case, 5.5–6.0% in the base case, 5.0–5.5% in the bull case. Applied to the current dividend, a 6.0% yield implies roughly C$65.80, 5.75% implies about C$68.70 and 5.5% about C$71.80. An EV/EBITDA cross-check on roughly C$1.4 billion of normalised EBITDA and about C$4.4 billion of net debt gives roughly C$45 per share at 4.5x, C$60 at 5.0x and the high C$70s at 5.5x, which shows how sensitive the equity is to modest changes in the enterprise multiple when leverage is this size.
| Scenario | Fair value range |
|---|---|
| Structural damage | C$40–50 |
| Bear / no recovery | C$53–61 |
| Base | C$68–78 |
| Bull | C$78–90 |
These are ranges, not price targets, and no date is attached to any of them. There is deliberately no single fair value figure in this report and the base range should not be collapsed into its midpoint. Against the C$58.77 reference price, the market is currently paying roughly what our bear case is worth.
Capital appreciation is potential additional upside rather than the primary return mechanism. The dividend case does not require a rerating, and it should not be presented as though yield, dividend growth and multiple expansion can simply be added together. If the yield stays near 6–7% for years while the dividend grows, the income case can still work; and a share price that stays low improves the arithmetic of any later buyback.
Our confidence in this valuation is medium. It rests on normalised figures that the next three years will confirm or refute.
| Sustainable DPS growth | 5Y yield on cost | |
|---|---|---|
| Bear | 0.0–2.0% | 6.72–7.42% |
| Base | 5.0–6.0% | 8.57–8.99% |
| Bull | 7.0–9.0% | 9.42–10.34% |
The table shows only the five-year horizon because five years is the horizon the authored growth assumptions cover. Beyond that, our view is that base-case dividend growth normalises toward roughly 3–4% rather than continuing at 5–6%, and compounding the initial band over ten years would apply an assumption past the point where we are willing to defend it.
What the yield-on-cost figures are: the arithmetic result of applying an assumed growth rate to today’s dividend and dividing by today’s price. What they are not: a forecast, a measure of whether buying today is attractive, or evidence of anything. They answer one narrow question, which is what the current purchase price would yield if the dividend grew as assumed. A reader who disagrees with the growth assumption should discard the yield-on-cost figure with it.
A note for investors outside Canada. Cogeco reports, trades and pays its dividend in Canadian dollars, so this report states a gross Canadian-dollar yield. An investor whose base currency is not the Canadian dollar carries currency exposure on top of the business, and that is a different thing from the company’s own exposure to the US dollar through its American operations; the two should not be conflated. Canadian dividends paid to non-residents may be subject to withholding depending on residence, treaty position and account type. We do not publish a single after-tax yield, because there is no such thing as one that applies to every reader.
Ordered by how much each one would change the conclusion, with the chain each risk travels.
1. Breezeline subscriber and monetisation deterioration. If customers and revenue per customer both keep falling, revenue erosion becomes difficult to offset by cost. US revenue falls, margin follows, consolidated free cash flow falls, and dividend growth goes with it. This is the single most important risk in the report.
2. Structurally higher recurring capital expenditure. If competitive network investment settles above roughly 19–20% of revenue for prolonged periods without better customer economics, normalised free cash flow is materially lower than we assume, and the payout ratio that makes the dividend look comfortable was measured against the wrong denominator.
3. Canadian deterioration. Canada is the stabiliser. If Canadian Internet growth falls toward zero and segment revenue and EBITDA turn persistently negative, the thesis changes materially rather than incrementally, because the base case does not have a second source of resilience.
4. Another large debt-funded acquisition. This would conflict directly with the balance-sheet repair the base case assumes, and it is the risk with the clearest precedent.
5. The 2028 refinancing cluster. Not a question of access today. It becomes a cost and flexibility problem if EBITDA is declining at the same time, which is precisely when it would matter most.
6. US margin compression. High American margins are a buffer against revenue decline. Persistent deterioration below roughly 46–47% would indicate that the problem has moved from the top line into the economics.
7. Free cash flow quality. The company-defined measure has changed definition and includes items such as proceeds from disposals of property, plant and equipment. Over multi-year periods it needs to converge with economic cash generation, and if it does not, the payout ratio in this report overstates coverage.
Two further risks are real but do not by themselves break the thesis: mobile failing to improve retention economics, and Canadian wholesale or fibre regulation altering investment returns. A permanent valuation discount is the least important risk here, because the income case does not require rerating.
Concrete enough to check against reported figures, which is the point of writing them down before rather than after.
On the dividend:
On the mispricing:
An unexpected dividend freeze would require review rather than an automatic conclusion; the cause would matter. A cut would be direct thesis failure unless it came from a restructuring that clearly created better long-run economics.
Three questions, kept separate, because collapsing them is how this company gets misread in both directions.
Is Breezeline weak? Yes. The customer losses are real, worse than large peers, and accompanied by revenue pressure and a C$2.2 billion impairment.
Does that make Cogeco a poor business? Partly. Consolidated quality is genuinely mixed. Canada is good. The American half is not, and capital intensity is a permanent constraint on both. This is not a pristine compounder and nothing in this report should be read as claiming otherwise.
Does that make the current dividend unsafe? Not under reasonable stress today. It costs about 30% of free cash flow, it survives our delayed-recovery and no-recovery cases without a cut, and the balance sheet is being repaired rather than stretched.
What makes this worth publishing is the distance between the second answer and the third. The market is pricing the business, and pricing it with legitimate concerns. The dividend is a claim on a smaller and better-covered slice of that business than the price implies. If the operating bridge holds, the combination of a 6.72% starting yield, a low payout and moderate sustainable growth does not need a rerating to work. If it does not hold, the more likely outcome is a dividend that stops growing rather than one that gets cut.
We hold the dividend safety judgment with high confidence, the dividend growth judgment with medium-high confidence, and the business quality, mispricing and valuation judgments with medium confidence. Those are not the same number and should not be averaged into one.
What we will be watching, in the order the evidence usually arrives: US Internet subscriber growth, US revenue per customer, the Ohio subscriber trend, Canadian Internet customer growth, and capital intensity split between expansion and recurring spending. Those five move first. Segment revenue and margins, free cash flow per share, leverage and the dividend declaration confirm later what they implied. Future updates to this thesis will be written against those measures, and will ask whether the evidence is moving toward the bear, base, bull or structural-damage case rather than restating the original argument.
No updates since publication.
This is general research, not personalised investment advice. It does not take account of any individual reader’s objectives, financial situation or needs. Every price shown is the price observed at the time of the analysis and is never updated afterwards.
Unpriced receives no payment from the companies it writes about. The author’s dealing policy and the timing rules that apply to it are published in the repository as TRADING_POLICY.md.
Draft — pending professional legal review.