The FCC has cleared the last federal barriers to shutting off the old copper telephone lines. AT&T is pulling loop-powered lifelines, built and paid for by the public over three generations, out of the ground across 500 wire centers, and selling the metal into the same AI-and-grid buildout that made copper scarce. A field note on what the $6 billion maintenance bill hides, who already paid for the network three times over, and who is left in the dark when the last line that works without power is gone.
AT&T is not switching off the copper because it is too expensive to keep. It is switching it off because the metal is finally worth more pulled out of the ground than left in it.
On March 26, 2026 the Federal Communications Commission (FCC) adopted Report & Order 26-19 and removed the two federal steps that once let regulators and competitors slow a copper retirement: the Section 214 discontinuance application and the Section 251(c)(5) network-change disclosure.01 What remains is a 90-day notice to the customer. The timeline belongs to the carrier now.
The gradient below is real. It shades each AT&T state by the number of wire centers already filed for retirement, from Texas at 458 down to Nevada at 5, out of roughly 4,600 copper wire centers the company still runs.03 Territory outside AT&T's legacy footprint sits with other incumbent local exchange carriers (ILECs) like Verizon, Lumen and Frontier, on slower and separate tracks.
On July 6, 2026 the FCC cleared AT&T to cut copper landline service for roughly 184,000 Californians, authorizing shut-off across portions of 360 wire centers on or after June 1, 2027. AT&T sued the California Public Utilities Commission (CPUC) and the state Attorney General; the CPUC is challenging the federal rules in the Ninth Circuit.02 Its argument is the one this whole note circles: loop-powered copper is the line that still works during an earthquake, a wildfire, or a grid outage.
The case for retirement rests on a single figure. AT&T spends roughly $6 billion a year maintaining a copper network fewer than 3% of its customers still use.03 Read against the rest of the income statement, that figure is smaller than the framing wants it to sound.
In 2025 AT&T booked $125.6 billion in revenue, $24.2 billion in operating income, and $21.9 billion in net income, the last more than doubling year over year.04 The copper upkeep is about 5% of revenue. It is also about 27% of the profit the company keeps. The network is not bankrupting anyone. It is a standing claim on a quarter of net income that earns almost nothing back, and no public company carries that forever.
The maintenance framing leaves out a second number. As telcos migrate to fiber, they can recover on the order of 800,000 to 960,000 tonnes of copper this decade, worth more than $7 billion at the prices that held when the estimate was made, and copper has climbed since.05 Once cables are pulled, stripped and cleaned, the profit can top 30% after processing.
AT&T states it plainly. The same executive who calls the copper network "very inefficient" also runs its copper recovery and resale, and told Bloomberg the company is scaling "significantly" because of where copper prices are.06 AT&T runs at least four reclamation centers and plans more. It has already turned the buildings into cash too, an $850 million sale-leaseback of 74 copper central offices to a real-estate developer.07
The wire in a fire-alarm circuit and the wire in a scrap bale are the same metal. Retirement is the paperwork that moves it from the first ledger to the second.
Anyone who senses something wrong in tearing out working line has the direction right and the reason backward. Nothing is being discarded. A loop-powered, blackout-surviving asset is being reclassified as inventory, at the moment that inventory became one of the most contested materials on earth.
The copper being harvested was never the carrier's to find in the ground for free. The Public Switched Telephone Network (PSTN) was built across the twentieth century as a regulated monopoly under rate-of-return rules. Ratepayers funded the plant, pole by pole and splice by splice, and the company was guaranteed a return for building it. AT&T California and AT&T Kansas were, by name and by charter, utilities the public financed.
Then the public paid to modernize it, again and again, and mostly did not receive what it bought. Since the 1990s, telecom analyst Bruce Kushnick and the New Networks Institute have documented an estimated $400 billion collected through rate increases and tax perks for a fiber-to-the-home future that never fully arrived, a figure the carriers dispute but have not cleanly rebutted.13 The recent numbers are harder. Under the FCC's Connect America Fund (CAF), AT&T, Frontier and Lumen took a combined $7.3 billion-plus from 2015 to 2020 to wire nearly three million rural homes and businesses, on top of the $438 million incumbents accepted under CAF I.1214 Price-cap carriers were handed roughly $9 billion against almost no build requirement beyond a 10 megabits-per-second (Mbps) floor that was obsolete on arrival.
A UC Santa Barbara and Berkeley study, with Ookla, found that once the subsidy money stopped, service in many of those funded locations faded.12 One trade outlet called the program what a former FCC official called it: graft. Then the government ran it again, another $6 billion-plus through the Rural Digital Opportunity Fund (RDOF), aimed at many of the same addresses. By one accounting, roughly $100 billion in federal broadband funding across 2010 to 2020 moved the digital divide by less than one percent.14
AT&T's case is not empty. The copper is inefficient, fiber is better where it reaches, and the company says 90% of the population in its copper footprint will eventually get it. None of that answers the sequence. The public paid to build the network, paid again for an upgrade it mostly did not receive, subsidized the rural stretch a third time, and is now watching the physical metal come out of the ground and go to private profit, while the bill for the replacement lands back on the same customers at three to ten times the old rate.
Build it with ratepayer money. Charge again for a fiber future that never ships. Then bale the copper and book the scrap. Three invoices, one network, and the customer is on all three.
Copper was worth reclaiming in 2024. In 2026 it is worth fighting over. Prices have climbed more than 60% since April 2025, cleared $13,000 a tonne for the first time in January 2026, and are being called toward $15,000.08 Early in 2026 S&P Global warned of a shortfall grave enough to pose a systemic risk to global industry, projecting a deficit as large as 10 million tonnes by 2040 as demand rises roughly 50%.09
The driver is the one straining everything else at once: artificial intelligence (AI). A single large AI data center can lock in up to 50,000 tonnes of copper, and the grid infrastructure to power it needs three to four times more copper than the building itself.10 US data centers drew about 4% of the country's electricity in 2024; the projection is 14% by 2030, with global data-center electricity roughly doubling to 945 terawatt-hours (TWh).11 Copper was named a US critical mineral, and a 50% import tariff landed in August 2025.
So trace the wire. It is pulled from an elevator phone in a retired wire center, baled, sold, and drawn back out as a busbar in a data center whose power demand is part of why the grid is under strain, in a market where that same demand is why the copper was worth pulling. The lifeline network is being fed, tonne by tonne, into the machine that is stressing the grid it used to survive.
Copper carried its own power. The central office sent a small current down the line, so a Plain Old Telephone Service (POTS) phone, a fire panel, an elevator call box kept working when the building lost power. Every option AT&T steers customers toward gives that up. The line now depends on power at the premises, a battery measured in hours, and a network that itself depends on the grid.
The pricing is not incidental to the migration. It is the migration. POTS lines that once ran $30 to $60 a month now routinely appear at $150 to $400, and some grandfathered lines have been billed above $2,700 a month, pressure to move before the shutoff notice even arrives.15
The lines most exposed are not desk phones. They are the ones nobody thinks about until they fail inspection: National Fire Protection Association (NFPA) 72 fire-alarm panels, American Society of Mechanical Engineers (ASME) A17.1 elevator emergency phones, Enhanced 911 (E911) interfaces, blue-light campus stations, and alarm and access-control panels.16 An elevator phone that goes silent fails its annual inspection, and a failed inspection can pull the elevator out of service. A fire panel that loses its communication path can void the insurance condition that required monitoring in the first place.
Federal notice is 90 days for residential customers and 180 days for non-residential and interconnecting entities.17 When a wire center retires, every building it serves needs replacement at once, and installers get overwhelmed. The buildings that audited early get scheduled first. The rest go to the back of the line.
The customers who remain on copper are not a random 3%. They are older, poorer, more rural, more likely to depend on a monitored medical device, and more likely to live exactly where the promised replacement never came. This is the population the retirement reaches first.
Now set that profile against the map in Section 01. AT&T's heaviest copper footprint runs through Texas, the rural South, and the lower Midwest, the same counties that rank high on age and poverty and low on fiber availability, and the same counties CAF and RDOF were paid to wire and largely did not. The people about to lose the line are, disproportionately, the people the upgrade money was supposed to reach first.
AT&T says no one will be left without voice or 911. It offers Phone-Advanced where fiber or mobile reach, satellite where they do not, and copper "as long as needed" where nothing else works. Weigh that against its own limits. The FCC has not yet blessed Phone-Advanced as a POTS replacement.20 The satellite and fixed-wireless options depend on power the copper never needed. And "as long as needed" is a promise with no meter on it, made by the party that profits when the copper comes out.
Strip away the spectrum of the fight and California's objection is simple and physical. In an earthquake, a wildfire, or a public-safety power shutoff, the copper landline is often the last thing in the building with a dial tone, because it draws power from a central-office battery miles away rather than from a wall socket that just went dark.02 The replacements invert that. Fixed wireless needs a powered tower. Fiber voice needs a charged battery on site. Both fail in the exact conditions a landline was trusted to survive.
The circuit closes here. The metal reclaimed from the resilient network is sold into the AI-and-grid buildout that is making the grid less able to keep those replacements powered. The copper leaves the one system designed to work when the power is out, and joins the system whose growth is a reason the power goes out. In fire and outage country, that trade is not abstract. It is the difference between a call box that works during the evacuation and one that does not.
A network is being retired for being old. What is being removed is its independence from the grid, at the moment the grid can least be trusted, from the people least able to route around its failure. Field Notes 05 · the throughline
Map shading uses per-state counts of AT&T wire centers filed for retirement, a public tally of §214/§251 filings; it is not a claim that every address in a state retires at once, and timing varies within each state. Connecticut is classed as other-ILEC because AT&T sold its Connecticut wireline to Frontier in 2014. California is broken out because its retirement is under active litigation. The price chart mixes IMF/World Bank annual averages (2020–2025) with an August 2026 LME spot mark; the two are labeled because a spot print and an annual average are not the same measure. The line-decline curve plots two FCC-reported points; the dashed leg is arithmetic at the reported 17.9%/yr rate extended to 2034, not a forecast. The $400 billion "broken promises" figure is an estimate from Bruce Kushnick / New Networks Institute, an advocacy source the carriers dispute; it is presented as contested, while the CAF/RDOF figures are FCC and peer-reviewed. The $7B copper-recovery figure is a 2024 estimate at then-current prices, and copper has since risen more than 60%, so treat it as a floor. AT&T's ~$6B/yr maintenance and <3% copper-customer share are the company's own figures (some sources cite <5%). Demographic reliance figures are national CDC survey data; the county-level overlap with the AT&T footprint is argued from footprint geography, not a joined dataset.
AT&T assured us no one would be left without a line. The copper had no comment. It was already on a truck.