100 days until the UK's copper phone network goes dark

What happened

Openreach, the BT-owned infrastructure arm that runs most of the UK's copper access network, has set a hard stop for the Public Switched Telephone Network: February 10, 2026 — roughly 100 days from now. After that date, PSTN and ISDN services terminate across England, Scotland, Wales and Northern Ireland. There is no extension, no grace period, and no exception process for stragglers. The Register's reporting makes the point blunt: the industry has had years of warning, and the regulator is done moving the goalposts.

The migration target is All-IP: voice calls carried over broadband using VoIP, delivered to homes and businesses as Single Order Generic Ethernet Access (SOGEA) or FTTP where available. Openreach's rolling regional stop-sells — the ones that quietly killed new PSTN orders region by region since 2023 — were the warning track. February 10 is the wall.

The number everyone is quietly worried about is not the residential one; it's the long tail of embedded devices still dialling out over copper that nobody has an inventory for. Openreach's own estimates have put the remaining PSTN line count in the low millions as recently as this summer, with a disproportionate share belonging not to grannies with corded handsets but to lift emergency phones, fire panels, burglar alarms, telecare pendants, payment terminals, remote SCADA endpoints, and fax machines in GP surgeries and solicitors' offices. BT previously paused the migration for vulnerable telecare users after a series of incidents in 2023 and 2024 where migrated lines failed during power cuts, leaving pendant users unable to call for help. That pause has now expired.

Why it matters

The PSTN is a 1980s-era circuit-switched network built on top of infrastructure that in some places predates the moon landings. Keeping it alive costs Openreach more every year as spares vanish and the people who understand D-side cabling retire. The economic case for shutdown is uncontroversial inside telco. The operational case — can everything that depends on dialtone actually be migrated in time? — is where this gets interesting for anyone who ships software or hardware into the UK.

The failure mode of PSTN death is not that calls get worse. It is that specific devices, in specific installations, stop working entirely and silently on a Tuesday in February. PSTN provides two things VoIP does not: line power (the exchange energises the copper pair, so a basic phone works through a blackout) and a deterministic analog signalling path that cheap microcontrollers from the 1990s can speak natively. Lift phones are the canonical example. UK building regulations require a working emergency phone in every passenger lift. Many of those phones are hardwired POTS units installed when the lift was commissioned, with no onboard radio, no battery, and firmware that assumes it can seize a line, dial DTMF, and talk to a monitoring centre over an analog bearer. On February 11, a non-trivial fraction of them will pick up the handset and hear nothing.

Alarm panels are the second-worst case. The grade-2 and grade-3 intruder alarms wired into most UK commercial premises were specified around a PSTN keep-alive: the panel pings the Alarm Receiving Centre over the line every 25 hours, and ARC pages the response service when the ping goes missing. Over VoIP, that keep-alive either stops working or becomes unreliable in ways that generate false alarms and burn goodwill with the police. The industry's answer — IP communicators and dual-path 4G backup — has been shipping for a decade, but installation lags, and the clock is 100 days.

Then there's the long tail that nobody wrote down: the diesel generator at the back of a datacentre with a modem for remote start, the Zebra printer in a warehouse that faxes delivery notes, the lone PDQ terminal in a pub that falls back to dial-up when the broadband router sulks. These are the devices that will not appear on anyone's migration spreadsheet because nobody filed a change ticket when they were installed in 2004.

The comparable precedent is Germany's Deutsche Telekom All-IP migration, which finished in 2022 after a seven-year run. Germany's rollout was slower, less dramatic, and still produced a long tail of broken fax machines, alarm systems and medical devices that nobody discovered until the dialtone went. The UK is compressing a similar transition into a shorter window with a sharper edge.

What this means for your stack

If you ship a UK-facing product or service, the audit question is small and specific: does any part of your stack, or any part of your customers' stack that touches your product, depend on a POTS line to work? The non-obvious places to look:

- Fax-based onboarding or compliance flows. Still common in UK healthcare, legal, and insurance. If your integration partner receives faxes over a BT line, that line goes dark in February and your webhook starts timing out. - PCI terminals with dial-up fallback. Many Verifone and Ingenico terminals in UK retail have a modem fallback for when IP is down. The fallback becomes a brick, and the terminal's "resilience" story becomes a single point of failure. - Out-of-band management. Older server rooms, PBXs, and industrial controllers sometimes have an analog modem on a POTS line for emergency console access when the network is down. If that was your break-glass for a UK colo, you need a 4G OOB appliance before February. - Telecare, lone-worker, and lift monitoring SaaS. If you sell into facilities management or social care, your product probably talks to devices that talk to copper. You should already have a migration playbook; if you don't, this week is a good time. - Automated phone-based 2FA and voice OTP. VoIP-delivered voice calls to UK numbers already have higher failure rates than SMS. That gap will widen as the migration completes and more of the terminating leg is VoIP-to-VoIP with inconsistent CLI handling.

The practical move is to build a one-page inventory: every outbound and inbound phone number your systems depend on, who owns the line, and whether the terminating device has been confirmed working on All-IP. If you can't answer that for a line, assume it dies in February and plan the fallback.

Looking ahead

The UK is not alone — Germany is done, the Netherlands finished in 2019, France is pacing toward 2030, and the US is doing it piecemeal under FCC order 19-72A1 with no national deadline. What makes the UK notable is the hard date and the regulator's willingness to let the deadline bite. The lesson for the rest of the world is less about telephony and more about legacy-protocol sunsets generally: the last 5% of migrations are not users dragging their feet, they're embedded systems nobody has the source code for. February 10 is going to teach a lot of UK facilities managers that lesson in one go.