AT&T had its best operating year in a decade and the shares are down 24%. On 8 October, SpaceX agreed to buy nationwide low-band spectrum. AT&T fell 10% in a day.
Nothing happened to the network. No customer left. The company did not report anything. What changed is that somebody announced they were coming, by a route the moat does not cover.
Four companies in our library carry yields between 5.0% and 6.9% and our four lowest scores. Three of them have a named newcomer at the door, and the fourth has something stranger. They are worth looking at together, because they are all the same problem.
What the moat was, and who arrived
| The moat was | Who arrived | Our score | |
|---|---|---|---|
| AT&T | Spectrum scarcity. You cannot build a national network without licences nobody is selling | SpaceX — buying nationwide low-band spectrum to make Starlink a fourth carrier | 4.8 |
| Comcast | Cable in the ground. Nobody digs up a city twice | Fibre, 5G home internet and Starlink — none of which dig | 5.4 |
| UPS | Density. Cost per package falls with volume, so the biggest network wins | Amazon, which built its own density and has just opened it to every shipper | 5.4 |
| Kraft Heinz | Brands and shelf space. The ketchup everybody asks for by name | The shopper — thirteen points of volume gone in three years | 4.4 |
Read the middle column and then the third. In every case the moat is still there. AT&T's licences have not been revoked. Comcast's cable is still in the ground and still works. UPS still runs the densest delivery network in America.
The entrant simply does not need to come that way.
A moat only works against someone who has to cross it. Satellites, antennas and a rival's own warehouses do not cross anything — they arrive from a direction the defence was never built to face.
Beating a moat, and going round it
This is the test, and it is the only one that matters at the moment a newcomer is announced.
A rival who has to beat the moat must spend what you spent, wait as long as you waited, and arrive at parity with no advantage. That almost never happens, which is why real moats last.
A rival who can route around it pays none of that. They are not building a second cable network; they are reaching the same house from orbit, or from a tower, or from a warehouse they already own for another reason. The moat is not breached. It is bypassed, and bypassing is cheap.
So the question to ask of any newcomer is not "can they match the incumbent?" It is "do they have to?"
Being bypassed does not automatically mean being destroyed. The second question is whether the entrant takes the cash flow or only the growth.
A newcomer who takes the growth leaves a large, profitable, shrinking business that can pay you for years — sometimes for decades. A newcomer who takes the cash flow leaves a business that cannot fund its own defence.
Those two outcomes look identical in the first eighteen months, and the share price treats them as the same thing. They are not remotely the same thing.
What has actually happened so far
| Cash flow today | What the numbers show | Dividend | |
|---|---|---|---|
| AT&T | $16–18bn a year | Best operating year in a decade. 110m wireless customers, 13m fibre homes | Frozen since a 47% cut in 2022 |
| Comcast | ~$17bn a year | Two years of broadband customer losses; penetration 52% → 48%; cutting prices | Frozen after 17 raises |
| UPS | Enough, and no more | Cut Amazon by more than half on purpose; each remaining package earns more | Takes all the free cash flow; frozen for 2026 |
| Kraft Heinz | $3.8bn a year | Thirteen points of volume gone in three years; $44.6bn written off | 6.8% on $1.60 |
The last column is the one to read twice. All four boards have already acted. Two froze after long raising records, one froze after cutting 47%, and the fourth is paying 6.8% out of a business that has lost thirteen points of volume.
A frozen dividend is a board saying, in the only language it has, that it wants the cash for something else. Sometimes that is prudence. At UPS, where the payment consumes all of the free cash flow, it is closer to arithmetic.
Four points that are not four points
Broadband penetration of the homes Comcast passes fell from 52% to 48%. That reads like a four-point slip.
It is 7.7% of the customer base, gone. On roughly 28.5 million broadband homes, a loss of that proportion is measured in millions of relationships, each of which was paying every month and now is not.
Percentages of percentages are the most reliably misread numbers in company reporting, and they are usually presented in the flattering direction.
What the market is paying for all this
| Price | Multiple | Yield | Against a 5.07% Treasury | |
|---|---|---|---|---|
| Comcast | $21.90 | 5.8× trailing | 6.0% | +0.93 pts |
| AT&T | $22.18 | 9.6× 2026 | 5.0% | −0.07 pts |
| Kraft Heinz | $23.68 | 11.5× 2026 | 6.8% | +1.73 pts |
| UPS | $94.57 | 13.1× 2026 | 6.9% | +1.83 pts |
Two things are worth saying plainly about that table.
Comcast at 5.8 times trailing earnings is not a cheap stock. It is a forecast. Nobody pays six times earnings for a business they expect to be intact in ten years. The multiple is the market writing down the terminal value.
And AT&T, at a 5.0% yield, pays you marginally less than a Treasury bill for owning a leveraged network under attack from a company that lands rockets. Whatever the case for the shares is, it cannot be the income.
What the defence costs while you wait
Defending a bypassed moat is expensive, and the bill lands on the shareholder before the outcome does.
AT&T has just borrowed to pay $23 billion for spectrum — buying more of exactly the asset whose scarcity is being challenged. Kraft Heinz's new chief executive has paused the planned split to spend $700 million reviving brands Americans are buying less of. UPS closed dozens of buildings and paid tens of thousands of people to leave.
None of those are foolish. All of them consume the cash that would otherwise have been yours, in a period when the return on the spending is unknowable.
Comcast spent $45.6 billion on buybacks at around $39 a share. The shares are $21.90.
That is roughly 1,169 million shares bought, worth about $25.6 billion today — approximately $20 billion of shareholder money destroyed, a 44% loss on the largest capital allocation decision the company made.
It is the single clearest illustration we have of what happens when a company buys its own stock believing a reprice is temporary. The cash is gone, the competitive position is unchanged, and the shares are at half the price.
Announcements are not customers
It would be lazy to end there, because the bear case has a real weakness: almost none of this has shown up in the cash yet.
AT&T had its best operating year in a decade. Comcast still produces about $17 billion a year. UPS is making more money on each package it keeps. The market has repriced four large businesses on what is, so far, mostly an announcement and a direction of travel.
Starlink has not signed a mass-market mobile customer. Amazon's network is newly open to third parties and unproven at it. Price cuts to defend broadband are visible; a collapse is not.
If the entrants take the growth and leave the cash, these are four very profitable businesses on single-digit and low-teens multiples paying 5–7%. That is a genuinely attractive proposition, and it is the bull case stated honestly.
Four questions, in order
Write down the path the newcomer takes to the customer, then check whether the incumbent's defence sits anywhere on it. If it does not, the moat is intact and irrelevant at the same time — which is the condition every company in this article is in.
Separate the two explicitly. A business losing new customers while keeping its existing ones can pay a large dividend for a very long time. A business losing existing customers cannot. Comcast's two years of absolute declines are a different fact from slowing growth, and only one of them is survivable at a 6% payout.
$23bn of borrowed spectrum. $700m to revive brands. Dozens of closed buildings. Add it up before you value the cash flow, because the shareholder funds the defence out of the same money the dividend comes from.
All four of these dividends are frozen or were cut. Boards see the internal numbers a year before shareholders see the consequences, and a freeze after seventeen consecutive raises is not a neutral act. It is the most honest disclosure in the whole report.
We scored these 4.4 to 5.4 and we would not own them for the yield alone — the fourth step above is why. But a low score is not a short thesis, and three of the four are priced for an outcome that has not happened yet.
The thing to watch is not the next announcement. It is the first quarter in which one of these entrants reports customers rather than intentions, and the incumbent's cash flow moves in response. Until then the share prices are trading on a forecast, and forecasts about bypass technology have a long history of being both directionally right and years early.
A network under attack from orbit, a cable business losing customers, and a parcel giant that shrank on purpose.
Each moat tested claim by claim, with the six dividend tests and a valuation by three methods.
Frequently asked
Is AT&T's dividend safe after the SpaceX news?
The payment is not the immediate issue; the price of defending it is. AT&T generates $16–18 billion of free cash flow a year against a 5.0% yield, but the dividend has been frozen since a 47% cut in 2022, it has just borrowed to pay $23 billion for spectrum, and a ten-year Treasury now yields slightly more than the shares do. You are being paid less than a government bond to own a network under attack.
Why did Comcast's shares fall so far?
Because it has lost broadband customers for two years and is cutting prices to keep the rest. Penetration of its homes passed fell from 52% to 48% — which sounds like four points and is actually 7.7% of the customer base gone. The market has repriced the shares to 5.8 times trailing earnings, which is what it pays for a business it believes is shrinking rather than pausing.
Can UPS keep paying a 6.9% dividend?
Only if the plan works. The dividend currently consumes all of the free cash flow and was frozen for 2026, which is the clearest signal a board can send short of cutting. UPS deliberately halved its Amazon volume to make each remaining package more profitable — and Amazon has now opened its own delivery network to every shipper in America, which is the part that decides it.
How do you tell if a moat is actually broken?
Ask whether the newcomer has to beat the moat or can simply route around it. A rival building a second cable network has to beat it. A satellite, a fixed-wireless antenna or a private label on the same shelf does not — it reaches the customer by a path the defence does not cover. Then check whether the new entrant takes the cash flow or only the growth, because those are very different outcomes.