AT&T, Verizon and T-Mobile bill about $11,000 every second between them and guide to roughly $59B of free cash flow this year. The market values all three together at $484B. This is a neutral read of where they stand: the money, the debt, the customers, the spectrum, the satellite threat, and what would actually grow their revenue.
Same industry, three different shapes. AT&T is the fiber-plus-wireless converger, Verizon the largest by revenue and the most indebted, T-Mobile the highest-margin and still the most richly valued, even after falling 34% in a year.
Revenue, EBITDA and free cash flow earned since January 1, ticking in real time. January to June uses each company's reported Q1 and Q2; from July 1 the meters run at each company's Q2 daily pace until Q3 is reported (Oct 21, 26 and 28). Nothing here runs past December 31, 2026.
Every month about 0.9% of each carrier's postpaid base leaves, so each one has to win those customers back just to stand still. The dots are real-time: one dot leaving is one line or account walking out the door, at the rate each company reports. Speed it up to see a day go by.
T-Mobile no longer reports phone net adds (it switched to accounts in 2026), so its number is an analyst estimate, shown hatched. The cable companies sell mobile on Verizon's network and keep adding hundreds of thousands of lines a quarter.
After the close on Thursday, Oct 8, SpaceX agreed to buy Grain Management's nationwide 800 MHz portfolio (up to 14 MHz paired; the WSJ reported about $8B). Low-band spectrum is what lets a network reach inside buildings and across rural miles. With EchoStar's AWS-4 and H-block already bought, investors read it as SpaceX assembling a mobile network of its own. The next day all three carriers fell hard.
T-Mobile closed at $148.58, a whisker above its 52-week low of $147.90. Verizon is roughly flat on the year; AT&T gave back its 2026 gains.
Q2 2026, all three combined. Roughly 40 cents of every revenue dollar becomes EBITDA; capital spending, interest, tax and working capital take most of the rest. What is left funds dividends, buybacks and spectrum.
Each company slices its revenue differently. AT&T moved to Advanced Connectivity, Legacy and Latin America in 2026; Verizon reports Consumer and Business; T-Mobile splits service from equipment. The bars use each company's own segments.
EBITDA minus capital spending is not free cash flow: interest, cash taxes, device financing and working capital sit in between. The grey step is that gap, as reported.
All three run their cash flow through the same three pipes: the network, the dividend and the buyback. Spectrum is the fourth pipe, and in 2026 it was the biggest for AT&T.
At about 7–8 times EBITDA and a free-cash-flow yield near 12%, the market is pricing the Big Three as businesses in slow decline. Their own guidance says free cash flow grows. One of those two is wrong. This tab lays out both cases without picking a winner.
AT&T uses its "$18B+" floor; Verizon and T-Mobile use the midpoint of their ranges.
Tower companies own the steel the carriers rent and trade at 18–21 times EBITDA. The carriers trade at 7–8. SpaceX trades at about 364 times: off this chart by a factor of 15.
If a stock is worth next year's free cash flow divided by (required return − growth), today's price tells you the growth the market is assuming. Pick a required return; the model solves for the perpetual growth rate built into each price and sets it against what management has guided.
Cheap is not the same as under-valued. Each case below uses only reported facts; the question at the bottom of each card is the one the next two quarters will answer.
Telecom is a borrowed business: networks and spectrum are bought up front and paid for over decades. All three sit at roughly 2.3–2.7 times EBITDA. AT&T is about to go higher because of the $23B EchoStar spectrum.
How much of each net-debt pile this year's EBITDA would cover, if every dollar went to debt (it doesn't: capex, interest, tax and dividends come first). It is a sense of scale, not a repayment plan.
The 2.68× AT&T reported for June 30 is before the EchoStar purchase closed on July 28. Management says leverage goes to about 3.2 times after the close (about $14.5B was drawn on term loans to fund it) and returns to about 2.5 times within roughly three years. Q3 on October 21 is the first quarter that shows it.
US postpaid churn runs 0.85–0.99% a month, which means the average postpaid customer stays around 100–118 months. That lifetime is why carriers can afford to pay hundreds of dollars to win a customer, and why a few basis points of churn are worth hundreds of millions.
Average lifetime = 1 ÷ monthly churn. At 0.86% that is 116 months. This is a long-run average for a stable base, not a promise for any one customer. Check what each figure measures: T-Mobile's 0.85% phone churn was said by its CFO on the call; its reported metric is account churn, 0.99%.
Pick a carrier to load its reported ARPU (or ARPA), churn and margin. Acquisition cost is not disclosed by any of the three: it is an assumption you set. Lifetime value here is undiscounted margin over the average lifetime; the discounted line uses a 9% annual rate.
The growth engine for all three is selling home internet to their phone customers. AT&T leads with fiber, T-Mobile with 5G fixed wireless, Verizon with both after buying Frontier.
Charter and Comcast sell mobile to their broadband customers and run it mostly on Verizon's network. In Q2 they added 854K lines between them, more than AT&T and Verizon's postpaid phone adds combined (616K). For Verizon that is wholesale revenue; for the market it is a fourth and fifth competitor that never had to build a tower.
In about a year EchoStar sold its spectrum to AT&T (≈$23B) and SpaceX (≈$19.6B), T-Mobile sold 800 MHz to Grain, and Grain agreed to sell an 800 MHz portfolio to SpaceX. Add an FCC re-auction and a new spectrum-pipeline law, and the question of who holds America's airwaves is open again.
SpaceX's total includes the Grain purchase at the price the WSJ reported (≈$8B, shown hatched; the companies did not disclose it). T-Mobile was a net seller.
Restored the FCC's auction authority through Sep 30, 2034 and set a pipeline of at least 800 MHz: 500 MHz from federal users, 300 MHz from the FCC, including at least 100 MHz of upper C-band within about two years.
160 MHz at 3.98–4.14 GHz in 3,248 licences. Order adopted Jul 22, 2026; the auction must close by July 2027; service starts December 2030 in the top 75 markets. Mid-band like this is the capacity layer of 5G.
Needs FCC approval. The SpaceX–EchoStar transaction was approved on May 12, 2026, with spectrum transferred May 22 and final closing targeted for Nov 30, 2027, when EchoStar is due to hold about 3% of SpaceX.
Starlink is already a broadband business doubling in a year. Its direct-to-cell constellation reaches ordinary phones. And in 2025–26 SpaceX bought licences on the ground. The open question for investors is whether that adds up to a fourth national carrier.
Each dot is a Starlink satellite at this moment, from public orbit data refreshed daily. Pick a place to count how many are above 25° in its sky right now.
starlink.json, written daily by the starlink.yml workflow in this site's repository. Run that workflow once (Actions → Starlink orbits → Run workflow), then reload.SpaceX's connectivity segment (mostly Starlink, worldwide) took in about $4.3B in Q2, up 66%. That is a seventh of T-Mobile's revenue, growing at a rate no carrier can match, at a falling price: ARPU is $66, down from $85 a year earlier, as cheaper plans and new countries pull the average down.
Between 2012 and 2015 all three sold or leased the bulk of their towers to Crown Castle and American Tower, and rent space back. The steel is now owned mostly by three tower REITs, which the market values at 18–21 times EBITDA. The carriers keep the spectrum, the radios and the customers.
Every structure the FCC requires to be registered (generally taller than 200 feet or near an airport), coloured by its owner of record. Shorter towers, rooftops and small cells are not in the registry, so this shows the big steel, not every site. Tap an owner to switch it off; hover a state for its counts.
towers-us.json, written by the towers-us.yml workflow from the FCC's weekly ASR file. Run it once (Actions → Tower data (FCC ASR) → Run workflow), then reload.Every lever below is sized with arithmetic on reported numbers: the base, the price per unit, the churn. The results are annual run-rate effects, not forecasts or dated projections. Assumptions are labelled where a company does not report the input.
At 110+ month lifetimes, every customer kept is worth years of revenue with no acquisition cost. A 0.05-point cut in monthly churn is worth $300–420M a year of run-rate revenue at any of the three (Lever Lab). Better CRM, proactive outage credits and bundle depth move it.
Fiber and fixed wireless are where all three add customers fastest. A household with both mobile and home internet from one carrier is harder to lose. AT&T has 38.6M fiber locations to fill; Verizon 30M after Frontier; T-Mobile is building through JVs.
1% on the postpaid base is worth $0.5–0.7B a year at each carrier. It only works if churn holds: the Lab's last slider shows how fast a price rise is eaten by customers leaving. Verizon's ARPA fell 1.4% in Q2; T-Mobile's rose 2%.
Watches, tablets, cars, trackers and business IoT add lines to existing accounts at little cost. It's the reason carriers moved to ARPA: the account is the customer, the line count is the upside.
AT&T's new 3.45 GHz and 600 MHz can carry fixed wireless (Internet Air) where fiber isn't built. Spectrum that sits idle earns nothing; T-Mobile sold 800 MHz it didn't need for $2.9B.
Even 1% of a carrier's base paying $10 a month is only $40–90M a year. Satellite earns its keep by removing a reason to switch, not by billing for it. The carriers' joint venture is designed to stop SpaceX owning that feature.
AT&T reports on October 21, Verizon on October 26, T-Mobile on October 28. These are the first results since SpaceX's spectrum moves and AT&T's EchoStar close. Every number on this site that runs "at Q2 pace" gets replaced by actuals then.