The Last
Dial Tone

On 31 January 2027, Britain's Public Switched Telephone Network is scheduled to retire. The date has slipped once already, from December 2025, after failures during migration put vulnerable telecare users at risk and, in two reported cases on one provider's network, contributed to deaths. So the instinct to ask whether it will slip again is understandable. It is also, I believe, the wrong frame.

My hypothesis, stated plainly so it can be marked against events: the headline date will hold, but completion will be achieved partly on paper. A residual tail of lines, our modelling suggests somewhere between 150,000 and 450,000, will not be migrated in time. Instead they will be caught by transitional products that BT and Openreach have quietly pre-built for exactly this purpose: an emergency-calls-only line called EVAc, and analogue-style stopgaps known as PDPL and SOTAP that run to around 2030. The platform retires on schedule. A minority keeps a dial tone through the side door. That is a soft landing, not a clean finish, and not a delay.

Why lean this way? Three reasons run through this report. First, the network is failing in plain sight: it consumes around 1% of British energy usage, suffered more than 2,600 major incidents in a single year, and runs on parts nobody manufactures any more. Second, the date is a commercial decision, not a law, and the government has said it will not legislate a pause. Third, and decisively, the escape hatch already exists. The argument that forced the 2023 pause, that vulnerable people would be cut off, has been structurally defused by products designed to make sure nobody is.

We put the probability of this soft landing at 72%, a formal further delay at 23%, and a genuinely clean completion at just 5%. Part Three shows the working, and page 26 publishes the exact criteria by which we will score ourselves after the event. Authority is cheap to claim and expensive to demonstrate. We have chosen the expensive route.

If you read only one section, make it the briefing written for you: the regulator and industry reader starts on page 28, the business and care reader on page 34, and every UK household on page 40.

The Public Switched Telephone Network is the copper-and-exchange system that has carried Britain's calls since the Victorian era. Its defining trick is that the line itself is powered: the dial tone comes down the wire from the exchange, which is why an old-fashioned corded phone keeps working in a blackout. For a century that resilience was the gold standard. Today the platform underneath it is dying of old age.

The numbers are stark. The legacy network still accounts for around 1% of British energy usage, a figure given by Openreach's own migration lead in June 2026. It suffered more than 2,600 major incidents in 2024/25, each affecting 500 or more customers^[21], and Ofcom recorded a 45% rise in significant resilience incidents in 2024. The parts that keep it alive are no longer manufactured anywhere: repairs increasingly rely on cannibalising decommissioned equipment, maintained by a generation of engineers heading into retirement.

So the question was never whether the PSTN would retire, only when, how cleanly, and who is protected on the way. The government's own line counts below show a quarter-century of managed decline arriving at its final winter.

Here is the fact that reframes everything: the switch-off date is not a law, and it is not an Ofcom deadline. The House of Commons Library's April 2026 briefing puts it plainly: the withdrawal of the PSTN is industry-led, and decisions on migrating customers are made by the companies that operate and provide services on the network. BT Group owns the retirement of the PSTN platform itself; Openreach is separately withdrawing the wholesale products (WLR and the ISDN family) that ride on it; and the hundreds of communication providers who resell those products migrate their own customers.

Government has explicitly declined to legislate, stating it would not be appropriate to pass legislation to cease a migration, because the underlying network is failing regardless. Ofcom's power is over conduct during the migration, principally General Condition C5 on protecting vulnerable customers, and it has shown teeth: a £23.8 million fine for Virgin Media in December 2025. But neither the regulator nor the minister sets the calendar.

That cuts both ways for anyone hoping for, or fearing, another delay. Moving the date requires nothing more than a BT decision, exactly as happened in May 2024. But equally, no statutory lever exists to force one. In practice, the 2023 precedent shows the real mechanism: a serious incident, ministerial convening, and a voluntary industry pause within weeks. The trigger is harm, not arithmetic.

For the typical household, the replacement is Digital Voice: the same number, usually the same handset, but plugged into the back of your broadband router instead of the wall socket. In the year to July 2025 alone, Ofcom recorded 1.9 million residential lines making that move, two-thirds of them customer-led. BT passed three million Digital Voice households in March 2026, the majority of its base.

For business, the picture is sharper-edged. The highlighted row above is the single most under-reported fact of the entire programme: ISDN lines, the workhorses of older office phone systems, simply cease. There is no emergency fallback for them. Around 80% of BT's business voice customers had moved to all-IP by January 2026, which still leaves a long tail of those who have not, and Part Two shows that business lines are migrating at a fraction of the required pace.

Your broadband. The retirement is of the analogue voice platform. Copper-based broadband (FTTC, SOGEA) continues to work after January 2027, and full fibre is untouched entirely. Nobody's internet goes dark because the PSTN retires.

Fully unbundled lines. Around 3.38 million lines (as of February 2026)^[30] are run by providers such as Sky and TalkTalk using their own equipment inside BT exchanges, a setup called MPF. Because those lines do not ride on the WLR product being withdrawn, they continue past January 2027 and only fall away when their host exchange eventually closes, mostly after 2030. That one fact materially softens the January cliff for millions of homes.

The wires and exchanges themselves. Openreach's separate exchange exit programme will eventually shrink around 5,600 exchanges to roughly 1,000, but only 108 are scheduled by December 2030 and the bulk close after that. Even the pilots are instructive: the first closure, Deddington, took 26 months for about 1,800 lines^[40], and two further pilots slipped a full year because fewer than 300 lines were not ready. The copper estate winds down over a decade, not overnight.

Anyone caught short. As the next page explains, unmigrated voice lines are gated onto an emergency lifeboat rather than disconnected. Openreach's own framing, on the record in June 2026: there will be no big bang.

The Office of the Telecommunications Adjudicator stated it plainly in January 2026: roughly 2.7 million WLR lines remained, falling 44,000 a week, against a required pace of around 52,000^[30]. The rate has barely moved since: 45,000 in November and December 2025, 44,000 through the new year. By June, the base was 1.9 million with about 33 weeks left.

The arithmetic from there is unforgiving. At a constant 44,000 a week, around 1.45 million lines clear by the deadline, leaving roughly 400,000 stranded. Clearing everything would demand a sustained 31% acceleration, every single week, through the hardest cohort of all: the people and businesses who have ignored every letter so far, the niche equipment, the premises where voice and broadband sit with different providers.

That is why our forecast treats a residual tail as close to certain, and why the interesting question is its size and its safety, not its existence. Openreach's own migration lead concedes reaching zero is unlikely, calling the last stretch the hardest because it is made of customers who do not want to move or have niche products on the line.

Residential migration gets the headlines; business is where the residual will actually live. Openreach warned in February 2026 that around half a million business lines had still not moved, and the sector's weekly run rate was a rounding error against the requirement. B4BC's blunt assessment: at roughly 2,000 a week, clearing the backlog would take nearly five years^[4].

The awareness picture explains why. Survey after survey through 2025 and 2026 found large minorities of firms either unaware of the switch-off or treating it as low priority: roughly three in ten businesses unaware in Gamma's research, 42% of small firms unaware in Spitfire's, and 72% not treating it as a priority even when told. Vendor-commissioned surveys deserve a pinch of salt, and we flag that, but the direction is consistent across every source.

And for one group the consequence is absolute. ISDN2 and ISDN30, the digital lines behind countless older office phone systems, cease outright on 1 February 2027. Openreach confirmed in June 2026 there is no migration path and no lifeboat for them: no EVAc, no PDPL, nothing. A business that has not replaced ISDN by the deadline does not get a degraded service. It gets silence.

Around 1.8 million people in the UK rely on telecare: 1.3 million through alarms in their own homes and half a million in care homes^[10], supported housing and sheltered settings. These are the people the 2023 failures harmed, the reason the date slipped, and the cohort on which any future delay decision will turn.

The technical fix has matured. Prove Telecare, nationwide since October 2025 with more than 35,000 protected migrations completed by June 2026^[42], tests the alarm with its monitoring centre before a migration completes and reverts to copper if it fails. The data fix is murkier: over 96% of local authorities providing telecare had signed at least one data sharing agreement by May 2025^[21], but a council signing with one provider does not cover the others, and people who bought alarms privately sit outside every register. No national figure exists for how many of the 1.8 million have actually been identified, contacted and safely moved: a gap the government itself concedes.

Two facts deserve plain words. First, devices still fail over digital lines at measurable rates: 2.3 to 3.4% in TSA testing, around 5% in FarrPoint's estimate, and up to 11.5% in the worst cases reported by Which?^[44]. For a device whose one job is summoning help, single digits are not small. Second, a Which? investigation in August 2025 found three of the seven largest UK telecare providers still selling analogue-only devices to consumers^[44], a year after ministers asked the practice to stop. Our research found no confirmed migration-caused death since the two 2023 cases, and an Ofcom-run incident framework now reports quarterly to government, but the ingredients for an incident have not been eliminated.

The carrot sits alongside the stick: between 8 June and 31 October 2026, Openreach is funding free migrations for voice-only lines, the cohort least likely to move unprompted. And the modern alternative is simply cheaper: Thinkbroadband calculated a like-for-like SOGEA service undercuts the old line-plus-broadband combination by around £21 a year even before the rises bite.

For the forecast, October is the date that matters. The full doubling lands on 1 October 2026, the free window closes at the month's end, and the run rate response in the weeks either side is the single best leading indicator of how big the residual tail will be. Page 50 builds it into the watchlist.

One: run-rate extrapolation. The simplest honest model. Take the 1.9 million lines remaining on 11 June 2026, apply observed weekly migration rates across the roughly 33 weeks left, and see where the arithmetic lands. Its strength is that it cannot be argued with; its weakness is that rates change, which is why we run three rate scenarios rather than one.

Two: reference-class forecasting. Rather than reasoning from the inside of this programme, ask what actually happened across every comparable migration: the UK digital TV switchover, the 3G shutdown, and eight national PSTN retirements from Sweden to Australia. The base rate is unambiguous: headline dates broadly hold, clean completions essentially never happen, and managed tails are the norm. Page 22 holds the full table.

Three: scenario weighting. Combine the arithmetic with the base rates and the structural facts (the safety valves, the failing network, the absence of any statutory delay lever, and the one genuine wildcard of a telecare incident), and assign explicit probabilities. Vague forecasts can never be wrong; ours can, deliberately.

The result, held since our first analysis and re-tested against everything in Part Two: 72% the date holds with a managed tail, 23% a formal delay, 5% a clean completion, with a central residual of 150,000 to 450,000 lines on the morning of 1 February 2027.

What would move the numbers up. A weekly run rate sustained above 55,000 after the July price rise, a smooth EVAc pilot, and business migration breaking 5,000 a week by September would push scenario A toward 80% and shrink the residual toward the bottom of the band.

What would move them down. A rate stuck below 35,000 for two consecutive adjudicator reports, a stumbling EVAc pilot, or any serious telecare incident would lift the delay scenario sharply, toward 35 or 40%, and we would say so publicly within days.

Why publish this? Because it is the cheapest possible test of seriousness. A forecaster who defines success after the fact can claim anything; one who defines it in advance, in print, with numbers, is staking reputation on being checkable. We will publish our own scorecard in the first week of February 2027, whichever way it lands, and we will keep the model updated monthly on our site in the meantime.

There is a technical name for this discipline: making the forecast Brier-scoreable, after the scoring rule that rewards calibrated probabilities rather than lucky guesses. You do not need the mathematics to apply the test. You just need the table above and a calendar.

The case for a formal delay. The run rate has been flat at 44,000 for half a year against a stated requirement of 52,000, and the gap has never closed. Business lines, nearly half a million of them, are moving at a pace that would take years, and ISDN customers face a genuine cliff with no lifeboat. The telecare data problem is unsolved: no national completion figure exists, analogue devices are still on sale, and the very cohort that triggered the 2023 pause will be squeezed hardest in the final quarter. Meanwhile the 2025 outages at BT and Three put 999 access under Ofcom investigation, and the exchange exit pilots have already slipped a year over fewer than 300 laggard lines. If 300 lines can delay an exchange, the argument runs, what will 400,000 do to a nation?

The case that the tail will dwarf our band. If the hard-tail dynamic bites earlier than modelled, with the weekly rate decaying through autumn as the willing exhaust, the residual could exceed 600,000, making the lifeboats an embarrassment of scale rather than a footnote.

Why we still hold 72/23/5. Each point above is real, and none changes the structural facts. The exchange pilots slipped because closure there strands every line; the national programme does not strand anyone, by design, because EVAc exists. The CNI charter gives government an escalation path short of delay. And the steelman's strongest exhibits, the run-rate gap and the business lag, are precisely why we forecast a six-figure residual rather than a clean finish. They are arguments about the size of the tail, not the existence of a third date. We would rather publish the counter-case and be tested than hide it and be safe.

The Virgin Media decision defines the migration era. Ofcom's enforcement director called it unacceptable that vulnerable customers were put at direct risk of harm, and the remediation it accepted, manual record reviews, a continuous-engagement loop for non-responding telecare users instead of disconnection, and an agreed end-of-process with local authorities, now functions as the operating manual for every provider's final-quarter push.

Expect conduct enforcement at full volume through to February 2027 while the calendar stays untouched: General Condition C5 on vulnerable customers, A3.2 on emergency access, and the two open investigations into the BT and Three mobile outages of summer 2025 as the live resilience front.

Three official-looking line counts circulate, and conflating them has produced more bad reporting than any other single error. The Telecommunications Adjudicator's monthly WLR KPI tracks one Openreach product family and stood at 1.01 million in February 2026. The broader migration-tracking total covers every PSTN-dependent line across every provider and read roughly 2.7 million in the same season, falling to 1.9 million by 11 June. And the government's own series, published in a Lords written statement in March 2026, gives the long arc: 35.2 million lines at the 2000 peak, 6.5 million at the end of 2024, 3.6 million at the end of 2025^[43].

None of these contradicts the others. They measure different things on different dates, and the government series provides a rare independent cross-check on industry figures. The discipline that prevents error is simple: never publish a line count without its date, scope and source attached, and never let a product KPI masquerade as the national total.

One further measurement trap deserves industry attention. Connected Nations 2025 reported a fall in significant resilience incidents, but Ofcom's own text concedes part of the cause: as migration shrinks the PSTN, incidents increasingly affect too few customers to cross reporting thresholds. The regulator plans to consult on those thresholds in 2026. Until then, the falling incident count is partly a measurement artefact, and honest analysis says so.

The precedent is instructive. The two 2023 deaths only became public because a Financial Times freedom of information request to DSIT forced partial disclosure, the department releasing timing data while withholding causal findings. Public-interest grounds for the datasets above are at least as strong, and the requests are drafted to survive the standard exemptions.

For providers, the November 2024 non-voluntary migration checklist remains the operative compliance text for the final quarter: process re-assessment before resuming forced moves, no telecare user migrated without a confirmed working solution, battery back-up beyond the one-hour minimum where committed, a continuous-engagement loop rather than disconnection for non-responders, post-migration checks, and an agreed end-of-process with local authorities. The March 2026 charter adds a twelve-month minimum public notice before any network or service switch-off. Auditors and journalists alike should test conduct against that list, item by item.

If our forecast holds, the platform retires on schedule while a six-figure residual lands on transitional products. The wrong question for that morning is "did the switch-off happen?", because the answer will be a technically true yes that hides everything interesting. The right metrics are about the tail: how many lines landed on EVAc, what proportion of them are vulnerable or telecare households, how quickly the residual drains through 2027, and how many ISDN businesses went dark with no lifeboat at all.

We commit to tracking and publishing exactly those numbers, alongside the self-scoring of this report's forecast against the resolution criteria on page 26, in the first week of February 2027 and monthly thereafter. Residual composition, not the binary, is the story.

A closing note on standards. Ofcom's voluntary accreditation scheme for digital comparison tools audits whether information is accessible, accurate, transparent, comprehensive and up to date, with prices refreshed at least fortnightly and quarterly spot checks. Seven services hold accreditation. The sourcing discipline running through this report, dual streams, dated figures, published reconciliations, a marking scheme for our own predictions, is that standard applied to analysis rather than price tables, and it is the standard we believe switch-off coverage now needs.

The defence row deserves a plain statement. This report fabricates nothing, and nothing public exists on military or homeland-security exposure to the switch-off: no line counts, no readiness statistics, no exemption lists. That is probably as it should be for security, but it means the honest answer to "is the military ready?" is that the mechanism exists, the CNI escalation route, and the evidence does not. It joins the ISDN base and England's telecare completion rate on our published gap register, and our information requests, rather than in a confident paragraph nobody could check.

Schools carry one risk the audit alone does not catch: timing. Migrations and engineering work complete during the holidays precisely because buildings are empty, which is also exactly when a failed alarm line, lift phone or door-entry system fails silently. The fix costs nothing: put line testing on the first-day-back checklist for every term, alongside the fire drill. GP surgeries and pharmacies should treat the NHS fax and the alarm line the same way after any weekend works notice.

ISDN2 and ISDN30 were the workhorses of office telephony for thirty years, and on 1 February 2027 they simply stop. Openreach confirmed in June 2026 that no migration path and no lifeboat product exists for them: not EVAc, not PDPL, nothing. The replacement is voice over your data connection, in one of two shapes.

Hosted voice moves the phone system itself into the cloud: handsets or apps connect over broadband, the provider runs everything, and you pay per user per month. It suits most small and mid-sized organisations and removes the on-site equipment entirely. SIP trunking keeps your existing phone system and replaces only the ISDN lines beneath it with digital channels, which suits organisations with newer systems they are not ready to write off.

The clock is the real message. A single-site hosted migration can land in weeks, but multi-site estates typically need six to nine months end to end, and number porting alone runs ten to fifteen working days per batch. Count backwards from 31 January 2027 and the comfortable booking window is now, not autumn. Organisations starting in November will be negotiating from desperation, against engineer diaries that are already filling.

The carrot matches the stick. Between 8 June and 31 October 2026 Openreach is funding free migrations for voice-only lines, and the modern equivalent is already cheaper at standing rates: Thinkbroadband's like-for-like comparison puts a SOGEA line around £21 a year below the old line-plus-broadband combination^[23] before any of the rises bite.

Put bluntly: every quarter of delay now buys a worse price on the old line, a busier engineer market for the new one, and the same hard deadline at the end. The financially rational move and the operationally safe move have converged, which is exactly what the pricing was designed to achieve.

For care homes, sheltered housing and supported living, the switch-off is not an IT project. It is a duty-of-care question that belongs on the risk register and in front of the board, because the systems at stake are the ones residents press when something has gone wrong.

Three facts shape the work. First, grouped systems behave differently from dispersed pendants: warden call and communal alarm installations need coordinated, building-level upgrades, not a per-device swap. Second, the data safety net has holes: over 96% of councils providing telecare have signed at least one data sharing agreement, but one agreement does not cover every phone provider, and residents who bought alarms privately are invisible to all of them, so providers must assume identification is their own job. Third, the kit market is not yet clean: Which? found three of the seven largest telecare firms still selling analogue-only devices in August 2025, so insist on TEC Quality certified, digital-ready equipment in every procurement.

The migration itself now has a tested process. Prove Telecare puts a trained engineer on site who confirms the alarm connects to its monitoring centre before the switch completes, and reverts the line to copper if it fails; more than 35,000 such protected migrations had completed by June 2026. Demand it by name for every affected resident, test before and after, again at one to two weeks, again at three months, and document every test. That paper trail is precisely what a regulator or coroner will ask for, and Scotland, with 67% of its 142,000 telecare users moved to digital by April 2025 and 78% of connections digital-ready by May 2025^[20], proves the whole exercise is entirely achievable.

Lift emergency phones. More than 12,000 lift lines had still not migrated as of February 2026^[25], and the legal bar is unforgiving: EN 81-28 requires two-way emergency voice that keeps working in a power cut. The fastest compliant fix for most buildings is a GSM autodialler with battery back-up, prioritising high-rise, firefighting and care-setting lifts. An entrapment with a dead line is the scenario every building owner is one missed audit away from.

Alarm signalling. The Redcare closure of December 2025 already forced this market through its migration, around 120,000 systems crossing to successors, and it set the template: insurers expect a dual-path replacement certified to BS EN 50136, the so-called DP3 grade, combining IP and mobile signalling so no single failure silences the alarm. If your intruder or fire signalling predates that switch, confirm with both your alarm company and your insurer that the replacement path is in writing.

The quiet machines. Older card terminals that dial out over analogue lines, door entry and gate systems, fax machines still embedded in healthcare workflows, franking machines and building telemetry all share the same fate and the same fix: identify, replace with IP or mobile equivalents, and test. None of them is individually dramatic, which is exactly why they are the lines most often discovered in February rather than June.

For most households this is genuinely small. Your provider writes to you, typically about four weeks ahead. On the day, your phone moves from the wall socket to a port on the back of your broadband router. Your number stays the same, calls sound the same or clearer, and most modern corded and cordless handsets work without any fuss. Many people complete the whole thing themselves in two minutes; engineer visits are reserved for those who need them, including telecare users, where the engineer tests the alarm before leaving.

Two honest wrinkles. Phones plugged into extension sockets around the house stop working, because those sockets connect to the old line; the fixes are a plug-in adapter, a cordless base at the router with extra handsets around the home, or asking your provider about wiring options. And a digital line depends on your broadband and on mains power, which is why the next page deals with power cuts properly rather than waving them away.

One genuine upside deserves a mention: digital lines carry scam-call screening tools the copper network never could, which matters in a country where the landline is a fraudster's favourite door.

The power cut truth. An old corded phone drew its power down the line from the exchange and worked in a blackout. A digital phone does not, because the router needs mains electricity. The rules, set by Ofcom in 2018 and reaffirmed in 2023, require providers to give at-risk customers a free solution keeping emergency calls possible for at least one hour, and a cross-party group of MPs argued in March 2025 that one hour is not enough, recommending at least four hours in homes and six for the wider network. So if your landline is your lifeline, do three things: tell your provider you depend on it, ask for the free battery back-up unit, and register on their vulnerable customer or priority services list. BT customers can simply dial 150. A charged mobile in a drawer is a sensible second layer, signal permitting.

The telecare double call. Around 1.8 million people rely on pendant alarms and similar telecare. If that is your household, or your mum's, make two calls before anything changes: one to the phone provider to say a telecare device is on the line, one to the alarm provider to confirm the device is digital-ready. Never let the switch happen until both have said yes in plain words. The engineer-checked process then tests the alarm with its monitoring centre on the day and reverses the change if anything fails. Test the alarm yourself before the switch, straight after, a week or two later, and again at three months.

The switchover scam playbook. Fraudsters love a national deadline, and this one is a gift to them. The rules that protect you are short: your real provider will never ask for your bank details, card number or password to "complete your switch", will never ask for remote access to your computer, and will never demand an upfront fee with urgency attached. Anyone doing any of those things is a criminal, however convincing the caller display looks. Hang up, wait a few minutes, and call the number printed on your bill. The official campaign's own watchword applies: providers never ask for passwords or bank details.

If you do nothing at all. Here is the truthful version, without the scare and without the soothe. You will not wake up on 1 February 2027 cut off. Unmigrated lines are gated onto an emergency-only service that keeps calls and 999 working but carries nothing else, while the price of staying old-fashioned has been deliberately doubling. Doing nothing costs you features, money and eventually patience, but it is designed not to cost you safety. Engaging when your provider writes to you remains the better deal in every respect.

One date has dominated this report, and rightly so. But 31 January 2027 is not quite a national event. The Openreach PSTN and ISDN withdrawal covers the whole of the UK with two deliberate exceptions: the KCOM network in Hull and East Yorkshire, and the three Crown Dependencies of the Isle of Man, Jersey and Guernsey. Each runs its own copper retirement, on its own timetable, under its own rules. Island and Hull businesses that assume the mainland date applies to them are working to the wrong clock, in both directions.

Hull is the standout. KCOM, not Openreach, owns the access network there, so the WLR withdrawal and the wholesale price rises simply do not reach it. KCOM runs its own migration: a gradual, premises-by-premises move to full fibre with no single hard switch-off day, and an exchange-level interim that lets a home without a fibre connection keep its copper for now. KCOM reports around 60% of premises upgraded, a full-fibre footprint of 305,000 premises to its March 2025 results, roughly 200,000 of them in Hull, and battery back-up for vulnerable customers. Newer rivals, MS3, CityFibre and Grain, now pass most of the city too, so Hull's choice, timing and pricing all differ from the rest of Britain.

Isle of Man. Manx Telecom began retiring copper in 2025 and will finish no later than 2029, regulated by CURA. The island is split into seven zones switched in turn, each property gets at least 12 months from first contact, with Temporary Exclusions for those who need longer, and a dedicated guide sits at copperswitchoff.im. A £50m commercial build plus £10m of government funding targets 99% full fibre of roughly 41,000 properties, and no vulnerable customer is disconnected until a suitable solution, covering lift lines, personal alarms and care monitoring, is in place.

Jersey. JT declared Jersey effectively the world's first full-fibre jurisdiction in 2018, and free four-hour battery units were agreed for vulnerable and community-alarm customers. Its formal copper decommissioning programme began in September 2022. We flag honestly that an exact date for completion is not publicly confirmed, with sources differing, so we state only what is verified: fibre to every premises by 2018, decommissioning under way since 2022, regulated by the JCRA.

So the British Isles picture is really three stories at once. Jersey has already arrived, Guernsey and the mainland are close behind on very different clocks, and the Isle of Man has years to run. For a reader, the practical test is the same everywhere: find out who runs your network, then ask when your address is due. On the mainland that answer is a single date; in Hull and the islands it is a local programme you can plan around.

The honest forecast holds. The platform almost certainly retires on schedule while a six-figure residual lands on transitional products, and the only true cliff is ISDN. That is good news told plainly: this is a phone-platform retirement, not an internet switch-off, and almost everyone has somewhere better to land. The job between now and the final winter is simply to make sure nobody is left standing on the old platform by accident.

About BroadbandSwitch.uk. An independent UK broadband comparison service: postcode-level availability checking across 35+ live providers, a free speed test, plain-English guides, and a verified market directory tracking more than 425 UK ISPs. The service earns commission when readers switch through its links; analysis and rankings are never sold, and this report was produced without provider involvement or sight before publication.

The promise. The forecast in these pages is marked against its own published criteria in the first week of February 2027, whichever way events land. The monthly tracker, the corrections log and the scorecard all live at broadbandswitch.uk. Authority is not a claim. It is a habit, kept in public.