Comcast Corporation

Comcast's main business is a physical network: coaxial and fibre cable running past about 59 million American homes and businesses, through which it sells Xfinity broadband internet, pay-TV, home phone and — using Verizon's mobile network — Xfinity Mobile phone plans. It also owns NBCUniversal (the NBC and Telemundo networks, the Peacock streaming service, Universal Pictures and DreamWorks, and the Universal theme parks in Orlando, Hollywood, Osaka and Beijing) and Sky, Europe's largest pay-TV company. A ten-year-old's version: Comcast owns the wire that brings the internet into your house, and on the side it makes films, shows the Olympics and runs the Harry Potter rides.

The wire is almost all of the profit. Of about $34.5bn of adjusted operating profit before depreciation (EBITDA) in 2025, excluding the cable channels spun off in January, broadband, mobile and business connectivity produced about $32bn — some 93%. Content and theme parks, a third of the revenue, contributed most of the rest, after Peacock's losses. That is why the shares rise and fall on one number each quarter: how many broadband customers Comcast gained or lost.

For two years the answer has been "lost". Comcast's residential broadband customers fell by 654,000 in 2025 and 232,000 more in the first half of 2026, and the share of homes it passes that buy its internet fell from 52.1% to 47.9% in nine quarters. To slow the losses it cut prices — everyday pricing, a five-year price guarantee, a free mobile line — and broadband revenue started to fall. The shares, at $21.90 on 19 October 2026, are a third below February's high and back to levels last seen in 2011–12, while the company still generates about $17bn of free cash flow a year.

Two lessons for an owner. The first is that Comcast's best decisions were always about the wire: buying cable systems next to cable systems, adding broadband, and now adding mobile lines to the same bill. Its media decisions have been more expensive — Sky most of all — and in 2026 it decided to give the media business to its shareholders altogether. The second is that the last decade has been a lost one for owners: with dividends reinvested, Comcast has returned −6% over ten years while the S&P 500 quadrupled. Much of the cash that could have been returned at today's prices was spent on shares at ~$39. Buffett's 2011 line applies: "what is smart at one price is dumb at another."

The economics of a cable plant. Most of the cost of a cable network is fixed: the plant in the street, the hubs, the technicians. Each extra customer on an existing street adds revenue at very high margin, and each one lost removes it. That is why penetration — the share of homes passed that subscribe — matters more than any other number. At 52% in early 2024, Comcast's plant was very profitable; at 47.9% it still is, but each point lost costs roughly $0.6bn of revenue a year at today's prices.

Why revenue is falling now. Until 2024 Comcast raised broadband prices most years and lost few customers. In 2025 it reversed course: national "everyday" pricing, all-in plans, a five-year price guarantee and a free mobile line for a year, to stop customers leaving for fibre and 5G home internet. The result in 2026 is fewer customers lost but less revenue per customer — broadband average revenue per user fell 3.1% in the first quarter and 3.8% in the second, and residential connectivity EBITDA fell 6–8%. Management calls it "an investment period".

What grows. Mobile (448,000 lines added in the second quarter, a record) and business services (+5% EBITDA) are growing; mobile is still small — 7% of the addressable lines in Comcast's footprint — and runs on Verizon's network under a wholesale deal. Theme parks were the growth engine of 2025 (a first $1bn EBITDA quarter after Epic Universe opened) and turned negative in mid-2026.

For thirty years cable broadband was close to Buffett's ideal of an economic franchise: needed, without a close substitute in most streets, and lightly regulated on price. Two of those conditions have weakened at once. Fibre from AT&T, Verizon-Frontier and others is a close — often better — substitute where it is built, and 5G home internet is a cheaper one; Comcast's own response, cutting price to keep customers, is the evidence. His 1990 warning about commodity businesses — "it's impossible to be a lot smarter than your dumbest competitor" — reads uncomfortably next to Comcast's CFO complaining in September about rivals' "irrational pricing" of gigabit fibre at $30–40 a month. The plant remains a valuable, hard-to-replicate asset, and in much of the footprint it is still the fastest connection available. We score the moat 5: a real network that has lost its monopoly and, for now, its pricing power.

The customer count. Comcast lost 375,000 residential broadband customers in 2024, 654,000 in 2025 and 232,000 in the first half of 2026 — about 2% of its base a year. The first quarter of 2026 (−65,000) looked like a turn, but the CFO said more than half the improvement came from one-off marketing around the Olympics and the Super Bowl; the shares fell 12.9% the next day. In September he told investors the third quarter would not improve on the year before. Meanwhile Comcast keeps adding mobile lines — 448,000 in the second quarter — which ties customers to the bundle and is the strongest evidence that its strategy is working at least partly.

The price. The bigger change is per customer. Broadband revenue per user fell 3.8% in the second quarter, and domestic broadband revenue fell 5.5%, its first sustained declines. Management expects the price reset to stabilise customers first and revenue later. If customers fall ~2% a year and price ~3%, broadband revenue falls ~5% a year; with fixed costs, EBITDA falls faster unless costs are cut.

Our reading: a melting ice cube, but a big one, melting slowly, and priced for a faster thaw. Nothing in Comcast's numbers suggests its customers are about to disappear: it still has 28.5 million broadband homes, it still adds mobile lines at a record pace, and its business unit is growing. What has disappeared is the ability to raise prices every year. At under six times trailing adjusted earnings and a free-cash-flow yield above 20%, the market is pricing a decline much steeper than the 2–5% a year the evidence supports. Buffett's 1991 letter is the right framework: a media business whose earnings "bob around" is worth around ten times earnings, not twenty-five — and Comcast trades at six. The risks are that the decline accelerates as fibre and wireless spread, and that the break-up leaves the remaining cable company with too much debt. The third-quarter results on 22 October are the next data point.

We would not change our view on one quarter. What matters is the direction over the next three or four: if broadband losses stay around 2% a year and revenue per customer stabilises in 2027 as management promises, Comcast is worth well above today's price. If losses double while prices keep falling, our pessimistic case — roughly today's price — becomes the central one, and the dividend becomes the whole return.

The pattern. For twenty years the competitor in a Comcast street was the phone company's slow DSL line. Today it may be a fibre line as fast as Comcast's, a 5G box at half the price, or a satellite dish — and in some streets all three. The cable plant remains the most widely available fast connection in its footprint, but the competition is now national and well-financed, and two of the competitors (AT&T and Verizon) sell mobile on their own networks, the bundle Comcast can only rent. The comparison on our board is Verizon — at once Comcast's mobile landlord and its fibre rival.

Integrity: no veto. Comcast's disclosures are thorough and its new segment reporting makes the economics clearer. Capital allocation: poor for a decade. The Sky purchase and the 2021–25 buybacks consumed some $85bn at prices that have not been justified; the board is now doing the sensible things — paying down debt, suspending buybacks, separating the businesses — after the price has fallen. Owner mentality: weak. Executive pay is high (say-on-pay won only ~58% support in June 2026), and no insider has bought shares at prices the company itself was paying $39 for. We score management and capital 4.

Buffett and cable. Berkshire never made Comcast a large holding, but it watched cable closely. Buffett described in his 2006 letter how cable and the internet were taking advertising and readers from the newspapers Berkshire owned — the same disruption that is now reaching cable itself. Confidential 13F amendments show Berkshire held between 5 million and 12 million Comcast Class A Special shares from 2004 to 2010, a holding worth a few hundred million dollars at most, and sold it. Which of Berkshire's managers chose it is not disclosed. The lesson he drew about media in 1991 — that businesses which lose their franchise status deserve much lower multiples — is the lens for this report.

Which earnings to use. Comcast's GAAP earnings for 2025 ($5.39 a share) include a $9.4bn gain on selling its Hulu stake to Disney; its adjusted earnings ($4.31) exclude that gain and the amortisation of acquired intangibles from the NBCU and Sky deals. In 2026 the comparison is further muddied by Versant, which was in last year's numbers and not this year's — it contributed about $3.2bn of EBITDA a year. On a like-for-like basis, Comcast's EBITDA fell about 7% in the first half of 2026.

The cash is the case. Whatever earnings measure is used, Comcast turns a large share into cash: free cash flow on Comcast's own definition (after capital spending and software) was $17.8bn over the last twelve months, about $5 a share, against a share price of $21.90. Even allowing for higher broadband capex in 2026 and the loss of Versant, owner earnings — free cash flow less stock pay, normalised — are around $3.60 a share.

The verdict on the dividend: safe in cash terms, uncertain in shape. A dividend that takes 27% of free cash flow is not at risk from the business; it is at risk from corporate decisions. The board froze it in January, suspended buybacks in July and will decide in the coming months how the dividend is divided between NBCUniversal and the remaining Comcast. Our expectation is that a shareholder holding both after mid-2027 receives at least what they receive today, because the cash supports it and both companies want investment-grade ratings and income investors. But the 17-year growth streak has ended. The comparison on our board is Verizon, whose higher yield takes about 55% of its free cash flow.

What is it worth? We discounted owner earnings of about $3.60 a share at 10% for ten years. If they decline ~2% a year for five years as broadband customers and prices erode, then hold flat, ending at 8 times, Comcast is worth about $30. If the decline is much faster — ~6% a year, then ~4%, ending at 6 times, as fibre, wireless and satellite spread — about $20, close to today's price. If mobile, business services and the parks offset broadband and earnings grow ~2% a year, ending at 10 times, about $43.

At $21.90 the price sits about 27% below our central value and close to our pessimistic case — the market is pricing the fast decline. For a business we score 5 on its moat and 4 on management, a margin of safety is required, and this is one. Our advice: accumulate in thirds. Buy the first third now; the second after the third-quarter results on 22 October if broadband losses are no worse than guided; the third below about $20 (a 6.6% yield) or once the NBCU-and-Sky capital structure is published. This is a bet on a slow decline, paid at 6% a year while you wait — not on a revival.

A second way to value Comcast is to price the two companies it will become. These are rough figures — the debt split and Sky's allocation between the segments have not been published — and they are our estimates.

The connectivity company keeps domestic broadband, mobile and business services: roughly $29–30bn of EBITDA after Sky's broadband and a share of corporate costs move to NBCU. Charter, the closest pure cable company, has traded at about 5–6× EBITDA in 2026; at 5× the connectivity company is worth roughly $147bn before debt. NBCUniversal and Sky — the parks, studios, NBC, Peacock and Sky — produce about $5–6bn of EBITDA in a normal year, with the parks the most valuable part; at 8×, roughly $44bn. Together about $190bn, less ~$83bn of net debt and other claims, leaves about $107bn, or ~$30 a share — the same as our central value. At 4.5× and 6× instead, the answer is about $24, still above today's price.

The cross-check says the same thing as the cash-flow method: today's price assumes either a much faster decline in the cable business or a much worse outcome from the separation than the evidence supports. It also says where the risk lies. Almost all the value is in the connectivity company, so an owner of Comcast after 2027 is mainly an owner of a cable network — the NBCU shares are a bonus, not the thesis.