The British state has always failed at projects. The Concorde was a financial disaster. The Channel Tunnel ran 80% over budget and was a year late. The Millennium Dome cost £789 million and was ridiculed within months of opening. Failure is not new.

What is new is the scale, the consistency, and the fiscal context in which it now happens.

The UK tax burden is around 36% of GDP and is forecast by the Office for Budget Responsibility to rise to around 38% by 2030-31, a post-war high.1 Annual debt interest is now around £110 billion, more than the country spends on defence and comparable to the entire schools budget.2 The NHS waiting list remains above 7 million after peaking near 7.7 million in 2023.3 Social care has a structural funding gap of £8 to £10 billion a year.4 The Office for Budget Responsibility has formally judged the long-term path of the public finances to be unsustainable on current policy.5 Citizens are paying historic levels of tax and receiving visibly degraded services.

And alongside all of this, for the last twenty-five years, the British state has been losing money on a scale and with a regularity that should have produced a public reckoning many times over. It has not. The losses have been absorbed, normalised, forgotten, and replaced with new losses. Each individual failure has been treated as an isolated episode, attributed to particular ministers or contractors or unforeseen events. The pattern has not been named.

This essay is an attempt to name it.

The numbers below are taken from National Audit Office reports, Public Accounts Committee findings, parliamentary written answers, departmental accounts, and major investigative journalism based on Freedom of Information requests. They are conservative wherever conservative figures exist. They are organised not by department, era, or political party, but by the type of failure each project represents. Because what becomes clear once you arrange the failures by failure mode is that the pattern is structural. The same things go wrong, in the same ways, across decades, across departments, and across changes of government.

Consider, first, the projects where the state spent enormous sums and delivered nothing.

The Garden Bridge in London consumed £43 million of public money and was never built.6 Transport for London authorised the funding. The Department for Transport added more. After years of legal disputes, planning challenges, and cost escalation, the project was abandoned in 2017 with not a single foundation laid. The bridge exists only as a paper trail.

The Stonehenge tunnel was cancelled in 2024 by the new Labour government after £179 million had been sunk into pre-construction works and consultation.7 The Development Consent Order was formally revoked in March 2026. The tunnel will not be built. The £179 million is not recoverable.

The first round of the UK’s Carbon Capture and Storage competition consumed £68 million between 2007 and 2011 before being cancelled. The second round, launched in 2012, consumed a further £100 million before being cancelled by the Treasury at the 2015 Spending Review. Total spend across two rounds: £168 million. Plants delivered: zero.8 The technology was not commercialised. The UK launched a third programme in 2021.

The Building Schools for the Future programme was cancelled in July 2010 by Michael Gove. Estimated sunk cost on schemes that had reached design stage: between £250 million and £500 million. The programme was scrapped before most of the planned works were even contracted. Capital education spending fell from £10 billion in 2010-11 to £4.6 billion in 2013-14, and never recovered the lost ground.9 The RAAC concrete crisis that emerged in 2023 was not caused by that decision alone, but it exposed the consequences of a long-running failure to maintain and renew the school estate.

The e-Borders programme, intended to track everyone entering and leaving the UK, was cancelled in 2010 after Raytheon was dismissed for poor performance. Programme spend reached £342 million. A subsequent legal settlement awarded Raytheon £150 million plus £35 million in legal costs. The Public Accounts Committee concluded in 2016 that the total cost of the system, when finally delivered eight years late and as a different programme, would exceed £1 billion.10

The BBC Digital Media Initiative was abandoned in 2013. £125.9 million had been spent. The Public Accounts Committee found that almost all of it had been wasted. The system was intended to digitise the BBC’s archive and production workflow. It produced almost nothing of operational value before being scrapped.11

The NHS National Programme for IT was officially dismantled in 2011 after consuming around £10 billion over a decade. It had been intended to produce a unified national electronic patient record. It produced no such system. Some component pieces survived, including NHSmail and parts of the Spine. The central vision did not.12

This category alone, projects where the state spent and delivered nothing, accounts for billions of pounds across two decades, every penny of it gone with no asset to show for it.

Now consider the second category. The projects that did deliver something, but where what was delivered was useless, dangerous, or vastly less than promised.

The Ajax armoured vehicle programme was contracted in September 2014 at £5.522 billion for 589 vehicles.13 The original in-service date was 2017. By 2021, soldiers were being medically discharged because the vehicles vibrated so severely that crews experienced hearing loss and joint damage. The MoD had paid the contractor more than £3.1 billion by late 2021, with only 26 vehicles delivered. The Initial Operating Capability declaration of November 2025 was withdrawn in January 2026 after soldiers fell ill during training. More than a decade after the 2014 contract, the programme was still facing serious safety and operability questions. The senior responsible owner was removed from the role after safety concerns were underplayed.

The Watchkeeper drone programme was contracted in 2005 with a budget of £700 million for 54 vehicles. By 2023, the cost had reached £1.35 billion. Eight of the 54 drones were lost in crashes. The vehicles flew 1,191 of an expected 18,000 hours between 2018 and 2022. The cost per flight hour, originally projected at £8,000, rose to over £88,000.14 The programme was cancelled in November 2024 with no like-for-like replacement until at least late 2026.

The Nimrod MRA4 maritime patrol aircraft programme was contracted in 1996 with an original budget of £2 billion for 21 aircraft. The number was reduced repeatedly. By the time of cancellation in 2010, only nine aircraft were under construction and the cost had reached between £3.4 billion and £4 billion. None entered service. The completed airframes were physically destroyed by excavator at the Woodford site in 2011.15 The UK was without a maritime patrol aircraft capability for the next decade and eventually purchased nine P-8 Poseidon aircraft from Boeing for £3 billion.

The F-35 fighter programme has cost the UK £11 billion to date for 37 aircraft of a planned 138. The National Audit Office concluded in July 2025 that the programme had delivered “disappointing” returns. Only around one-third of the fleet is fully mission capable at any given time. Roughly half is mission capable in any sense. The whole-life cost of the UK’s F-35 capability is now projected at £71 billion to 2069.16

The Queen Elizabeth class aircraft carriers were approved at £4.1 billion for two ships in 2007. By 2019, the cost had risen to £7.6 billion.17 HMS Queen Elizabeth withdrew from a NATO exercise in February 2024 with mechanical problems. HMS Prince of Wales spent nine months in dock in 2022 after a propeller shaft failure. The carriers are operational but with persistent reliability issues that limit their actual deployable capability.

The smart motorways programme spent approximately £6 billion converting hard shoulders to active running lanes on stretches of motorway across England. The safety controversy centred on stopped vehicles in live lanes, where the risks were shown to be far higher than on motorways with a permanent hard shoulder. The programme was paused and then scrapped in April 2023 after years of public campaigning by bereaved families. National Highways was then funded to spend an additional £900 million retrofitting emergency stopping areas onto sections that should never have been built without them.18

The Green Homes Grant was launched in 2020 with a £1.5 billion budget to insulate 600,000 homes and create 82,500 jobs. The programme was closed in March 2021 after eight months. £314 million had been spent. 47,500 homes had been upgraded. 5,600 jobs had been supported. The administrative cost per home insulated exceeded £1,000.19

The smart meters programme has a whole-life cost of around £13.5 billion for the rollout of around 32 million meters by 2023. The original 2019 deadline was missed three times. Around 9% of installed smart meters were not operating in smart mode. The annual saving per household, originally projected at substantially higher levels, was found by the National Audit Office in 2023 to be approximately £56 a year. Five pounds a month.20

This category, projects that delivered something demonstrably useless, dangerous, or vastly inferior to what was promised, runs into tens of billions of pounds.

Now the third category. Projects that did deliver something more or less recognisably aligned with the original specification, but at a cost so far above the original budget that the comparison becomes almost meaningless.

HS2 was approved in 2010 at £37.5 billion in 2009 prices for the full Y-shaped network connecting London to Birmingham, Manchester and Leeds. By 2024, the eastern leg to Leeds had been cancelled, the northern leg to Manchester had been cancelled, and the remaining London-to-Birmingham stub was estimated to cost between £45 billion and £66 billion in 2019 prices. Adjusted for inflation, the upper end approaches £80 billion. The cost per kilometre, around £200 million, is six to ten times the equivalent figure for high-speed rail in France, Germany, Spain, or Italy.21 The project will deliver a partial railway, between two cities that are already connected, at multiples of the original cost.

Crossrail, now the Elizabeth Line, was approved at £14.8 billion and delivered at £18.9 billion. The central section opened in May 2022, three and a half years late. Whitechapel station alone, originally budgeted at £110 million, cost approximately £831 million by completion. Bond Street rose from £110 million to £660 million. Paddington from £147 million to £647 million. The combined overrun on station works ran into billions.22

The Astute class submarine programme was approved at £4.3 billion for six boats. By March 2024, the Infrastructure and Projects Authority recorded the cost at £11.256 billion. The average delay across the boats has been 27 months.23

The Hinkley Point C nuclear power station was approved in 2013 at £18 billion in 2015 prices. By 2024, EDF was forecasting between £41.6 billion and £47.9 billion in 2024 prices. The first reactor is now expected in 2030 or 2031, more than thirteen years after the original target. The cost overrun is borne contractually by EDF, but UK consumers will pay the locked-in strike price of £92.50 per megawatt-hour in 2012 prices, equivalent to roughly £129 per megawatt-hour in 2024 prices, for thirty-five years. If market reference prices fall below the indexed strike price, consumers make up the difference through the Contract for Difference mechanism.24

The Sellafield decommissioning programme has seen its lifetime cost projection rise from £72 billion in 2006 to £136 billion in March 2024. The Nuclear Decommissioning Authority’s range for the final cost extends to £253 billion. The National Audit Office concluded in October 2024 that it could not conclude that Sellafield was achieving value for money. Each decade of delay in the construction of the Geological Disposal Facility for the country’s nuclear waste adds £500 million to £760 million in storage costs. One project alone, the Replacement Analytical Project, was paused in 2024 with a £1 billion cost overrun and a five-year delay.25

The Restoration and Renewal of the Palace of Westminster has produced cost estimates ranging from £15.6 billion for full decant over 20 to 25 years, to £39.2 billion for partial decant over 60 years, as set out in costed proposals published in February 2026. The cost of delay has been quantified at approximately £70 million per year in additional spend, plus £250 million to £350 million per year in construction inflation. A final decision on the long-term option is now scheduled for no later than 2030.26

This category runs to many tens, perhaps hundreds, of billions of pounds, depending on how the figures are aggregated.

Now consider the pandemic spend. A category of its own, in which the British state acted under extreme pressure and time constraint, but produced a scale of waste that has no historical comparator.

NHS Test and Trace was allocated £37 billion across 2020 and 2021. Lifetime spend reached £29.3 billion. At one point in April 2021, 2,239 of the central staff were consultants, representing 45% of the workforce. The National Audit Office concluded that the programme could not be shown to have produced a reduction in transmission proportionate to the spend. Bills for individual consultants exceeded £6,000 per day.27

The NHS Personal Protective Equipment procurement programme spent around £12.6 billion against an expected total of £13.1 billion. Subsequent accounts recognised very large impairments and write-downs, reported at up to £14.9 billion, reflecting unusable, unsuitable, expired or overvalued stock and contract losses. £10.5 billion of contracts had been awarded directly without competition. £1.5 billion had been awarded without due diligence. 1.5 billion items were past expiry by 2022.28

The Bounce Back Loan Scheme issued £47 billion across 1.5 million loans during the early months of the pandemic. The expected loss to the taxpayer is £17 billion. Estimated fraud is £4.9 billion, later revised in some analyses to £3.5 billion. The National Investigation Service was given a target to recover just £6 million across three years, equivalent to recovering 0.013% of estimated fraud.29 Lord Agnew, the Treasury minister with responsibility for fraud, resigned at the despatch box in January 2022, describing the government’s attitude as “woeful”.

The Eat Out to Help Out scheme paid out £849 million during August 2020. HMRC subsequently estimated that £71 million of this, around 8.4%, had been claimed fraudulently or in error.30 Independent academic analysis attributed a measurable surge in COVID transmission to the period the scheme was active.

The Birmingham Nightingale hospital, which cost £66.4 million to build, treated zero patients across its operational life. The Nightingale programme as a whole cost £532 million by the end of 2022. Some sites were repurposed for other clinical use. The original purpose, the surge capacity for which the programme was designed, was never required at the scale anticipated.31

The Coronavirus Job Retention Scheme, the Self-Employed Income Support Scheme, and Eat Out to Help Out together produced £4.5 billion of fraud and error, of which only an estimated £1.1 billion was recoverable. The chief executive of HMRC told the Treasury Committee in November 2022 that the remainder was lost.32

Pandemic spending, treated as a single category, includes tens of billions of pounds in waste that even on a charitable interpretation cannot be defended as the price of necessary speed. The procurement systems failed. The fraud controls failed. The accountability mechanisms failed. The money flowed out and substantial portions of it did not come back.

Now the local government category, which deserves attention because it shows the same patterns operating at the level of councils that are nominally responsible to local electorates.

Birmingham City Council issued a Section 114 notice in September 2023, formally declaring itself unable to balance its budget. The proximate cause cited was an equal pay liability then estimated at between £650 million and £760 million, on top of approximately £1.1 billion in earlier equal pay settlements. The implementation of an Oracle ERP system, originally budgeted at £40 million, had reached an estimated £100 million in remediation cost. The largest local authority in Europe was effectively bankrupt. The picture has since become more complicated. The October 2025 settlement reached with the trade unions was reported as costing in the region of £250 million, raising serious questions about whether the original liability was substantially overstated and whether the Section 114 notice was justified on the figures the council had provided. The episode still exposed serious governance and financial-control failures, but the financial story is no longer the simple one originally told.33

Croydon Council issued three Section 114 notices between November 2020 and November 2022. Total toxic debt reached £1.6 billion, much of it accumulated through the council’s commercial property company Brick by Brick. The government issued capitalisation directions of £150 million in 2020 and a further £224 million in 2024. Council tax in Croydon rose by 15% in February 2023 under special permission from central government.34

Thurrock Council issued a Section 114 notice in December 2022 with a £470 million in-year funding gap. The cause was £655 million invested through a single financial intermediary, primarily into solar farms. £275 million of investment losses were declared, alongside £129 million required to repay investment debt. A civil claim against the intermediary alleged that £150 million of the council’s money had been used to fund a “lavish lifestyle”.35

Woking Borough Council issued a Section 114 notice in June 2023 declaring a £1.2 billion deficit. The council had borrowed £1.3 billion from the Public Works Loan Board to fund subsidiary commercial property companies, including the Victoria Square development on which £490 million was subsequently written down. The council’s annual core funding was approximately £16 million. The debt-to-budget ratio exceeded 100 to 1.36

Nottingham City Council established Robin Hood Energy in 2015 as a not-for-profit municipal energy company intended to challenge the Big Six. Cumulative losses reached £34.4 million by March 2019. The company collapsed in 2020. The total cost to taxpayers was £38 million.37 The council issued a Section 114 notice in November 2023.

These are not isolated cases. They are the visible tip of a pattern of local authority financial collapse driven by a combination of austerity, commercial overreach, governance failure, and the collapse of in-house financial expertise. The Section 114 notices are now issuing at a rate that has no precedent in modern British local government.

Now the compensation for state failure. This category is distinct because it represents not waste during a project but the fiscal cost of past misgovernment, recognised after the fact.

The Post Office Horizon scandal has produced compensation provisions of £1.8 billion to date, with final costs likely to exceed this figure given the scale of the affected population. Around 3,500 sub-postmasters joined the Group Litigation Order. More than 100 wrongful convictions have been overturned. The Horizon IT system, supplied by Fujitsu, cost in excess of £1 billion across its operational lifetime. The Post Office Horizon Inquiry has cost more than £60 million.38 The new IT system intended to replace Horizon is itself running into difficulty.

The Infected Blood scandal has produced a compensation provision of £11.8 billion, set aside in the Spring Budget 2024 following the Inquiry’s final report. Around 30,000 people were infected with HIV or hepatitis through contaminated blood products supplied by the NHS between the 1970s and the 1990s. More than 3,000 have died.39 The state knew about the risks. The state did not act. The compensation, paid forty years late, represents the price of that failure.

The Windrush compensation scheme has paid £128 million across 3,842 successful claims by January 2026. The administration of the scheme has itself been the subject of repeated criticism for delays, hostility to claimants, and a low payment rate.40 Many of those affected died before receiving compensation.

The Grenfell Tower fire produced an inquiry that cost approximately £170 million, alongside cladding remediation programmes that have cost the public purse in excess of £5 billion. The fire itself killed 72 people.41 The regulatory failure that led to it was a product of decades of deregulation, lobbying, and institutional negligence.

These compensations represent the recognition, finally, of failures the state had concealed, denied, or ignored for years or decades. They are, in fiscal terms, the bill for past governance.

Now the consultancy spend that runs through all of the above.

Estimates vary, but all point in the same direction. Tussell analysis put UK public-sector consultancy spending at around £3.4 billion in 2023-24, a near-fivefold increase from approximately £700 million in 2016. HM Treasury’s central-government estimate for 2022-23 was around £1.36 billion. The National Audit Office has warned that government lacks a clear picture of the true total.42 The Big Four firms, Deloitte, KPMG, EY and PwC, dominate the market alongside Accenture, PA Consulting and a handful of others. NHS Test and Trace, at its peak in April 2021, employed 2,239 consultants representing 45% of its central workforce. The Treasury and Cabinet Office pledged in November 2024 to halve government consultancy spend by 2026, a target whose achievement remains uncertain.

The pattern across all of this is consistent.

Projects are approved on the basis of optimistic business cases that systematically understate cost and overstate benefit. The Treasury’s own Green Book guidance has required optimism bias adjustments since 2003.43 The National Audit Office has documented case after case in which the guidance was either not applied or applied inadequately.

Projects are then delivered through procurement systems that transfer risk to suppliers nominally but bear the cost when those suppliers fail or renegotiate. The Infrastructure and Projects Authority’s most recent assessment rates 41% of UK government major projects as Amber, indicating significant issues, and 12% as Red, indicating that successful delivery no longer appears possible. The 12% Red rating is the highest proportion since the Government Major Projects Portfolio was created.44

Senior responsibility rotates faster than the projects can be delivered. The UK has had eight Transport Secretaries since 2019, eight Defence Secretaries since 2010, and five Education Secretaries since 2019. The HS2 Ltd CEO position has changed hands multiple times. The senior responsible owner of the Ajax programme was removed from the role in 2025 after safety concerns were underplayed. Permanent Secretaries change every three to four years on average. The institutional memory required to deliver projects of fifteen or twenty year duration does not exist within the people responsible.

The civil service capacity to deliver complex projects has been hollowed out. Successive rounds of headcount reduction have removed the in-house expertise that would once have sat alongside contractors. The space has been filled by consultants. The consultants charge multiples of civil service rates. They produce reports. They advise on programmes that subsequently fail. The cycle continues.

Direct personal accountability has been rare. Ministers and senior officials are more often reshuffled, reassigned, or allowed to move on than visibly sanctioned in proportion to the scale of failure. No major contractor has been struck from the government supplier list as a consequence of cost overrun or non-delivery on a single project, even repeatedly. The financial penalties built into contracts are typically a fraction of the renegotiated profit margins. The Public Accounts Committee produces critical reports. The reports are noted. The pattern continues.

This is the part of the essay where it becomes necessary to step back from the specific projects and ask what they reveal in aggregate.

What they reveal is a state that has lost the capacity to deliver. Not in any one area, but across every area in which it operates. Defence, health, transport, energy, welfare, IT, buildings, local government. The same pattern. The same failure modes. The same absence of consequence. The same political and institutional incentives that direct effort toward announcing projects rather than completing them. The same hollowing out of internal expertise. The same dependence on consultants who have, themselves, no operational stake in the outcome.

A country that loses the capacity to deliver projects loses, eventually, the capacity to deliver anything else. Hospitals do not get built. Schools do not get rebuilt. Reservoirs do not get constructed. Power stations do not get commissioned. Railways do not reach the cities they were promised to. Justice systems do not adjudicate cases within reasonable timeframes. Police do not attend crimes. Care does not reach the people who need it. The state continues to exist, but it stops doing the things the state is for.

This is what the failed projects reveal, and why they matter beyond their individual price tags.

The connection to the broader sense of decline is direct. The British public has noticed, increasingly, that the country no longer works. Roads are pitted. Trains are late. Hospitals do not have appointments. Courts are scheduling trials into 2028. Police forces have been widely criticised for weak responses to lower-value shoplifting, creating the perception that some everyday crime is no longer meaningfully enforced. Care workers remain among the lowest-paid essential workers in the economy, with persistent concerns about insecure hours, unpaid travel time, and effective minimum-wage compliance. Schools have crumbling concrete. The Royal Navy cannot reliably deploy its aircraft carriers. The army cannot reliably operate its newest armoured vehicles. The drone fleet has been retired. The maritime patrol capability was lost for a decade.

These are not separate failures. They are the visible surface of a deeper failure of capacity. And the cost of that deeper failure, in cumulative terms across the projects that constitute it, runs into hundreds of billions of pounds.

The fiscal context makes this more devastating, not less. The British public is paying historic levels of tax. The fiscal headroom for further investment has effectively gone. Debt interest at around £110 billion a year now consumes more than is spent on defence and is comparable to what is spent on schools. The Office for Budget Responsibility has formally concluded that the long-term path is unsustainable.

Inside this context, the failed projects represent something specific and unforgivable. They represent the choice the British state has made about what to do with the resources its citizens have provided. The choice has been to spend extraordinary sums on programmes that did not work, that delivered late or never, that produced equipment soldiers cannot operate or vehicles that vibrated their crews into the medical discharge queue, while the basic services on which the country depends have been allowed to degrade.

Money has not been the problem. The British state has spent prodigious sums during this period. The problem has been what the money was spent on, and how, and who oversaw it, and what came out the other end.

Citizens are paying historically high taxes for visibly degraded services, and at the same time funding a procession of failed, late, or wasteful programmes that would themselves have funded the services that have been degraded. The two facts are not coincidence. They are connected. The capacity that has been lost was lost in part because of how the money was spent. The civil service that should have been delivering the country’s basic services has been gutted while the consultancy market has expanded to fill the space. The infrastructure that should have been built has been replaced with consultancy reports about why infrastructure cannot be built. The basic functions of the state have been hollowed out by the failure of its grand projects.

This is what the receipts show. Not that Britain has had bad luck with a few projects. Not that costs sometimes go up. Not that civil servants are imperfect or ministers occasionally rotate. Not that public sector procurement is intrinsically difficult. All of these are true. None of them explain the pattern.

The pattern is a state that has substantially lost the ability to complete what it starts, in a country whose citizens are paying more for the privilege than at any time in living memory, while the things they actually need become harder and harder to access.

If this is not the definition of a country in decline, then there is no definition. The receipts are public. The numbers are sourced. The names of the projects are known. The departments responsible are identified. The ministers and the senior officials and the contractors who delivered each of these failures have, in the overwhelming majority of cases, faced no professional consequence. Many have moved on to similar roles, often advising on the next round of projects.

There is no indication, in any of the data, that the pattern is breaking. The Equipment Plan affordability gap has been reasserted by the National Audit Office in every recent year. The Infrastructure and Projects Authority’s red ratings have not improved. The Public Accounts Committee continues to produce critical reports at a steady cadence. The consultancy spend, despite explicit pledges, continues to rise. The next generation of failures is being commissioned now, with the same business cases, the same procurement structures, the same accountability vacuum, and the same political incentives that produced the previous generation.

The receipts show what was paid. They do not show what was learned. On the available evidence, very little.

A serviceable country pays for what it gets. It expects projects to be delivered. It expects ministers to face consequences when they fail. It expects civil servants to be held to account. It expects contractors to be penalised for non-delivery. It expects the Auditor General’s reports to be acted upon, not merely filed. It expects, in short, that the receipts and the goods correspond.

Britain does not have a serviceable country in this sense. It has, on the available evidence of the last twenty-five years, an unserviceable one. A country whose citizens pay, and whose state takes the money, and whose state does not, with sufficient regularity to be a structural feature, deliver what was paid for.

The receipts are above. The bill has already been settled. The goods, in case after case, never arrived.

That is the country we have become. And until the failures begin to produce consequence, for the people who oversaw them and the institutions that enabled them, there is no reason to believe the pattern will end.

The next set of receipts is being printed now.

There are two numbers that should be sitting on the desk of every policymaker, bank executive, and pension trustee in Britain. They are not.

The first number is £9.18 trillion. That is the total value of residential property in the UK, as measured by Savills at the end of 2025. It is an all-time record. It is more than three and a half times the annual gross domestic product of the country. It is over 3.8 times the market capitalisation of the entire FTSE 100.1

The second number is 70%. That is the share of UK workers in occupations the International Monetary Fund classifies as exposed to artificial intelligence, according to the UK government’s own January 2026 assessment of AI and the labour market. It is higher than the United States, which sits at around 60%. It is higher than the advanced-economy average, which also sits at around 60%. On this measure, the UK is more AI-exposed than the US and more AI-exposed than the advanced-economy average.2

Both numbers are real. Both come from serious published research. They describe, between them, the largest pool of wealth in the country and the labour force that underwrites it. Almost nobody is connecting them.

This essay is an attempt to connect them. It will argue that Britain has built a £9 trillion asset class on top of a labour force that is being targeted, more directly and more comprehensively than almost any peer economy, by the first general-purpose technology in a generation that can perform cognitive work at scale. It will argue that the trade was not chosen deliberately. It was made one mortgage at a time, over the course of a generation. It will argue that the early signals of the compression are already visible in the UK labour market data, though the data does not yet prove AI is the cause. And it will argue that the trade is not yet visible in the housing market because every institution with a stake in the status quo has a structural reason to stay quiet.

To understand the trade, it helps to start with what UK housing actually is.

It is not, in any meaningful economic sense, bricks. It is not land. It is not planning permission. Those are all inputs. What UK housing is, at the margin, is a function of what the next buyer can borrow. And what the next buyer can borrow is a function of the next buyer’s wages. Mortgage regulation in the UK treats loans above 4.5 times income as high loan-to-income lending, with a cap on the share of a lender’s book that can sit above that threshold. In practice, a large share of UK mortgage capacity is shaped by income multiples in the broad range of four to four and a half times borrower income. Which means that UK housing, as an asset class, is a leveraged derivative of British pay.

The composition of British pay is decisive. According to the most recent House of Commons Library data, 81% of UK gross domestic product is generated by services, and 83% of UK employment is in services. Law. Finance. Consulting. Accounting. Administration. Media. Technology. Middle management. Professional and technical services. These are not marginal categories in the British economy. They are the British economy. Manufacturing accounts for roughly 9% of output. Agriculture, under 1%. Construction, around 6%. The rest is service work of one kind or another, concentrated disproportionately in London and the South East, which together hold over 40% of UK housing value on just 26% of the housing stock.3

These services occupations are also, according to the IMF’s detailed exposure modelling, the occupations most exposed to artificial intelligence. The IMF’s Labour Market Exposure index, which maps AI capabilities onto the tasks performed in specific occupations, consistently ranks legal work, financial analysis, business operations, clerical work, and content production among the highest-exposure categories in any developed economy.4 Goldman Sachs, in its 2023 research on AI and labour, reached broadly the same conclusion, estimating that around two-thirds of jobs in the United States and Europe have some exposure to AI automation, with the highest-exposure categories concentrated in knowledge-intensive white-collar work.5

The UK is more exposed than the US, and more exposed than the advanced-economy average, because the UK is more services-heavy than either. This is not a contested point. The UK government’s own January 2026 labour market assessment acknowledges it explicitly. It cites the IMF’s estimate that around 70% of UK workers are in AI-exposed occupations, compared with around 60% for the US and for advanced economies in general, and it attributes the difference directly to the UK’s “service-sector-intensive economy.”6 The UK government paper also notes that exposure indices of this kind are indicative rather than precisely validated against real-world outcomes. The 70% figure measures the share of UK workers in occupations containing tasks that AI could perform or enhance. It does not measure how many jobs AI will actually replace.

The reason this matters for housing is that the UK has financed its residential property market, aggressively, on the back of those same services wages.

Total UK household debt now stands at around £2 trillion. Of that, over £1.6 trillion is mortgage debt. The UK household debt-to-income ratio, as of the fourth quarter of 2025, stood at 117.5%. Lloyds Banking Group alone carries a mortgage book of around £317 billion. Around 41% of first-time buyers now receive parental help with their deposits. Legal & General has described the Bank of Mum and Dad as comparable in scale to a top-ten UK mortgage lender, though the comparison is a metaphor for family contributions rather than a regulated lending institution. The average first-time buyer is 34 years old.7

Note what this means in practice. The Bank of Mum and Dad is not a separate pot of money sitting outside the housing market. It is housing equity being recycled. Parents gift deposits from the value of their own homes, either by remortgaging, by downsizing, or by passing on wealth that is itself predominantly tied up in property. Which means the deposit being put down by the younger generation and the mortgage being taken out against it are, in a very real sense, the same asset counted twice. The supposed cushion under the UK housing market is, on inspection, a mirror. It reflects the thing it is supposedly cushioning. It does not provide independent support.

The banking system sits on top of all of this. For the major UK retail banks and building societies, mortgage lending is one of the largest and most systemically important asset classes. The pension system sits, in a broad systems sense, on top of that. UK defined benefit pension funds hold roughly a third of all UK gilts. Defined contribution pension funds hold substantial allocations to UK bank equity, UK commercial real estate, and increasingly to UK residential property through build-to-rent. In a broad macro sense, much of Britain’s retirement system is indirectly long UK services wages: through tax receipts, gilt sustainability, bank equity, property-linked wealth, and household balance sheets. The exposures are not all direct, but they do all trace back, through different paths, to the earning power of the same labour force.

This is the edifice. A £9.18 trillion asset class, leveraged four to four and a half times against a labour force that the IMF identifies as unusually AI-exposed, underpinning a banking system that is itself underpinning a pension system that is itself underpinning the retirement of a generation that expects the whole structure to hold.

Now run the scenario.

The following numbers are not forecasts and are not drawn from any official model. They are simple scenario arithmetic designed to show how income compression could transmit through a leveraged housing market. Treat them as illustrations, not predictions.

Imagine, conservatively, that AI compresses UK services wages by 10% in real terms over the next decade. Not a collapse. A compression. Roughly in line with what the IMF’s own recent research on AI adoption in labour markets suggests is plausible for middle-skilled workers in AI-exposed economies. The gains from AI, on the IMF’s analysis, concentrate at the high-skill and low-skill ends of the labour market. Middle-skilled workers, who happen to be the exact cohort that clears the UK housing market, benefit least.8

What does a 10% compression in services wages do to UK house prices?

The intuitive answer, that prices fall 10%, is wrong. The actual answer is that prices fall more, because the leverage amplifies the move. The mortgage multiple works in both directions. A 10% fall in income reduces maximum borrowing by 10%. A 10% reduction in borrowing capacity, applied across the marginal buyer pool, removes a cohort of buyers from the market entirely. Comparables reset downward. Transactions slow. Forced sales begin, initially at the margin, among the most recently leveraged. Each forced sale prints a new, lower comparable, which feeds the next round of revaluation.

The historical record gives us a sense of how far this can go. The 1989 to 1995 UK housing downturn produced a nominal house price fall of around 20%, and a real-terms fall of 37% over six years, with the national average not recovering its 1989 peak in nominal terms until 1998.9 The 2007 to 2009 fall was shallower but faster, with nominal prices falling by around 16% between late 2007 and early 2009. Both of these were shocks triggered primarily by interest rate moves and financial-sector stress, not by fundamental wage compression. Neither attacked the underlying income base of the market. A sustained services wage compression would come closer to attacking the underlying income base itself.

The Bank of England’s own 2025 Bank Capital Stress Test, the regulatory framework that calibrates bank capital requirements, models a scenario in which UK residential property prices fall by 28% from their starting point, alongside an unemployment rate peaking at 8.5% and a base rate of 8%.10 The banks passed the test. The Common Equity Tier 1 ratio of the aggregate UK banking system fell from 14.5% to 11% at the scenario’s low point, and no participating bank required remedial action. The headline finding was that the UK banking system is well capitalised and can absorb a 28% house price fall without systemic distress.

Read that again carefully. The Bank of England has already modelled a 28% house price fall as a severe but plausible tail risk scenario. The banks have stress-tested their books against it. The regulator describes this shock as within the range of what the financial system can absorb.

The shock the Bank of England modelled was a traditional macroeconomic recession. High interest rates, high unemployment, temporary asset price dislocation. The recovery in their model follows a recession-shaped path. Prices fall, then stabilise, then recover as the economic cycle turns.

That is not the scenario this essay is describing.

The scenario this essay is describing is a structural, secular compression of services wages driven by the substitution of human cognitive labour by machine cognitive labour. There is no cycle to recover from. There is no interest rate cut that restores the underlying income. The labour market that historically generated the wages that underwrote the mortgages does not come back in its previous form. It compresses, holds at the new level, and may then compress further as the technology improves.

Japan is the closest historical comparator we have, and the comparison is sobering. The Japanese property market peaked in 1991. Residential land prices in major Japanese cities fell continuously for roughly 15 years. Some urban land-price indices, particularly in the largest Japanese cities, saw peak-to-trough falls of extraordinary scale, in some cases exceeding 70%.11 National prices began to rise again only in 2018, roughly 27 years after the peak, and only in a few prefectures, with the rest of the country continuing on a long downward drift. The Japanese shock was not primarily about AI. It was about a demographic reversal combined with a banking crisis and the end of a speculative bubble. But the structural characteristic was the same: an income base that could no longer support the property valuations that had been built on top of it.

The Japanese comparison is imperfect, and I do not want to overstate it. The UK is not Japan. The UK has population growth, the UK has a different banking structure, the UK has a different political economy. But the Japanese experience demonstrates one thing with particular clarity. When the underlying income base of a housing market changes structurally rather than cyclically, the asset revaluation can take decades rather than years, and the recovery can be partial rather than complete. It is possible for a country to spend a generation living inside a revalued housing market.

With that in mind, here are three illustrative scenarios. Again, these are not forecasts.

If a 10% services wage compression over a decade produced a 20% to 25% fall in UK house prices, broadly in line with the 1989 experience, £9.18 trillion of housing value becomes £6.9 to £7.3 trillion. Between £1.8 trillion and £2.3 trillion of wealth is erased. That is more than the entire FTSE 100 market capitalisation.

If the compression is 15% rather than 10%, and the amplification factor is 2.5 rather than 2, the fall approaches 40%. Housing wealth becomes £5.5 trillion. Over £3.6 trillion of wealth is erased. That sits above the Bank of England’s current stress test scenario but within the range of historical comparables.

If the compression is closer to 20%, sustained over 15 to 20 years, and combined with the demographic and fiscal pressures already shaping the UK’s long-term trajectory, the UK approaches the territory Japan occupied in the 1990s and 2000s. Nominal prices could fall by 30% to 50% from peak, with a recovery measured in decades rather than years.

None of these are predictions. They are scenarios. They describe what happens if a particular set of assumptions about AI, wages, and leverage holds. The assumptions may not hold. The technology may plateau. The productivity gains may flow through to wages more evenly than the IMF’s research suggests. Policy may intervene. Migration may reshape the labour force in ways that restore the wage base. Any of these things could change the outcome.

But here is the point worth sitting with. The Bank of England has modelled a 28% fall as a severe but plausible tail risk. The Japanese experience shows that structural income shocks can produce 50% falls over a generation. The historical UK record shows that 20% nominal falls are possible even from cyclical shocks, and that real-terms falls of 37% have been produced by interest rate and financial stress alone. The scenarios I am describing are not extreme. They are within the range of what regulators, historians, and comparable economies tell us is possible.

And the technology that would trigger the scenario is no longer speculative. It is in production, deployed across every major services firm in the country, performing tasks that were, until recently, done by salaried human workers.

Here is where it gets uncomfortable.

The early signals of the kind of labour market shift the scenario section describes are already visible in the UK data.

The UK unemployment rate rose from 4.4% to 4.9% over the year to February 2026. Some 206,000 more people were unemployed at the end of that period than at the start. Payrolled employees, measured directly from HMRC’s real-time PAYE data, fell by 74,000 year-on-year as of February 2026, and by 87,000 over the comparable Labour Force Survey period. This is not a softening of the rate of growth. It is an absolute contraction of the taxable employment base, at a moment when the UK’s fiscal projections require that base to be growing.12

Vacancies tell the same story. The ONS recorded 711,000 vacancies in the three months to March 2026, a fall of 29,000 on the previous quarter, and the lowest level of vacancies in the UK economy since the final quarter of the pandemic-affected labour market in early 2021.13

The redundancy data is more striking still. According to Insolvency Service HR1 data, published by the ONS as official statistics in development and compiled in Freedom of Information analysis by the insolvency firm Liquidation Centre, 315,605 UK jobs were flagged for potential redundancy in 2025. This was the highest annual total since 2020, and it represented a 45% increase on 2021. In the first two months of 2026 alone, 736 UK employers filed HR1 notifications covering 56,396 jobs at risk, an 8.65% increase on the equivalent period in 2025. February 2026 saw 430 HR1 submissions filed with the Insolvency Service. In February 2009, at the gathering peak of the global financial crisis, the figure was 433.14

HR1 forms measure potential redundancies, not actual ones, and the data is labelled by the ONS as statistics in development. The comparison with February 2009 does not mean actual redundancies are running at financial crisis levels. It does mean that the rate of redundancy planning, as measured by employer filings with the Insolvency Service, is running at a level not seen since the financial crisis, and that this is happening outside the context of a 2008-style financial shock. AI may be part of that story. So may weak demand, higher labour costs following increases to employer National Insurance and the minimum wage, regulatory uncertainty around the Employment Rights Act, and wider business caution.

The composition of the hiring weakness points more directly toward the AI channel. Entry-level job postings in the UK have fallen by roughly 30% since the launch of ChatGPT in late 2022, according to Adzuna data. UK tech graduate roles fell by 46% in 2024, with industry forecasters projecting a further decline through 2026. Graduate-targeted hiring in accounting and professional services, two of the UK’s most AI-exposed sectors by IMF taxonomy, has fallen by over 30%. A CIPD survey of UK businesses found that in over a quarter of larger firms, headcount is expected to be reduced because of AI, with junior roles most likely to be affected. A Boston Consulting Group survey found that 41% of employers say AI is already allowing them to cut the number of employees. A World Economic Forum survey found that 40% of employers globally expect to reduce their workforce where AI can automate tasks.15

The pattern is not random. It is concentrated in exactly the occupations the IMF identifies as most AI-exposed. Junior legal work. Financial analysis. Consulting analysis. Accounting. Administrative and clerical work. Content production. Entry-level data processing. The tasks that AI performs most competently today are the tasks that, until recently, constituted the training path for the next generation of UK services professionals. Those training paths are narrowing visibly, in real time, in the job posting data.

Wage growth is slowing in parallel. The ONS reported that annual regular pay growth fell to 3.6% and total pay growth to 3.8% in the three months to February 2026, with both now below 4% for the first time since 2020. In real terms, annual pay growth was just 0.7% including bonuses and 0.4% excluding them. The UK labour market is not yet producing wage compression in the sense the scenario section describes. But the engine of pay growth has clearly slowed, and the composition of the workforce has begun to tilt toward unemployment and inactivity, precisely in the age groups and occupational categories where the AI effect would first appear.

None of this, in isolation, proves that AI is compressing the UK labour market. Unemployment rises in slowing economies. Graduate hiring is cyclical. HR1 forms measure potential redundancies, not actual job losses. Employer surveys capture expectations and self-reporting, not always realised outcomes. Higher employer National Insurance, minimum wage increases, weak demand, and regulatory uncertainty are all part of the current labour-market story. An honest analysis has to acknowledge that.

But the pattern is still consistent with the thesis. The weakness is showing up first in vacancies, payroll employment, redundancy planning, and entry-level hiring. It is showing up most visibly in exactly the kinds of white-collar, junior, analytical, administrative, and professional-services roles that AI exposure models identify as vulnerable. That does not prove causation. It does mean the risk has moved from theory into observable labour-market data.

The housing market, for now, is holding its nominal value. Savills’ £9.18 trillion figure is a record. House prices are not falling. The thing that underwrites house prices, the earning power of the workforce that services the mortgages, is already showing signs of the kind of structural shift the scenarios in this essay describe. The gap between the two is the space in which the trade still lives. It is not yet visible in the prices. It is increasingly visible in the labour market data that determines those prices.

The rebuttal is that AI will boost UK productivity and therefore UK wages. The Organisation for Economic Co-operation and Development’s central estimate is that AI could add between 0.4 and 1.2 percentage points to annual UK labour productivity growth over the next decade, placing the UK second only to the United States among G7 economies for AI-driven productivity upside.16 This is a genuinely significant number, and the UK government has already taken to citing it as a reason for optimism about the fiscal and growth trajectory.

There are two problems with relying on it.

The first is that mortgages do not clear on aggregate productivity. They clear on individual incomes. A productivity gain that flows to shareholders, founders, and capital holders does not underwrite a mortgage in Birmingham. It underwrites a dividend stream to investors who may or may not be UK taxpayers, a share price appreciation for equity holders who may be concentrated among the already-wealthy, and a growth in GDP that shows up in aggregate statistics without showing up in household income. If the productivity gains are captured by capital rather than labour, which is what the recent history of technology-driven productivity gains in the UK and US suggests they may be, the gains do not reach the wage base that underwrites the housing market. They reach somewhere else.

The second problem is distributional within labour. Even the UK government’s own January 2026 assessment, which emphasises the productivity upside, acknowledges that the distributional effects are uncertain and likely to concentrate gains among a subset of workers. The IMF’s research on early AI adopter labour markets found that wage gains concentrated at the high-skill and low-skill ends and bypassed middle-skilled workers. The middle-skilled are the cohort most likely to have a mortgage, to have bought recently, and to be highly leveraged against their income. A productivity gain that skips them does not help clear the market. It may actually accelerate the compression, by making the firms deploying AI more profitable while simultaneously hollowing out the salaried cohort whose mortgages anchor residential property valuations.

The blast radius extends well beyond house prices.

Buy-to-let yields depend on rent, which depends on tenant wages. Compress the wages and the yields break, forcing sales into a falling market. Pension funds holding UK bank equity take a mark-to-market hit as the banks’ mortgage books deteriorate. Council tax receipts soften as property revaluations work through the system. Treasury receipts decline as stamp duty, capital gains tax on property, and inheritance tax revenues all fall simultaneously. The fiscal projections that the Office for Budget Responsibility produces every six months assume rising property-linked revenues. A sustained fall in property values would require the OBR to revise its medium-term fiscal forecasts downward, which would in turn force the government to choose between tax rises, spending cuts, and higher borrowing. Each of those choices has second-order effects on the housing market, which amplify the original shock.

Every major UK asset class is, in one form or another, indirectly linked to services wages. When services wages are the thing that is revalued, many things revalue together. The housing market is the canary in the mine, because it is the most leveraged and the most visible. It is not the only bird in the mine.

And yet the silence is total.

Politicians cannot say this because the median voter is a homeowner, and telling homeowners that their primary asset is structurally mispriced is not a strategy for getting elected. The Conservative Party has spent forty years building its electoral coalition around homeownership and cannot be seen to question the foundation of that coalition. The Labour Party has spent the last decade courting the same voters and has no interest in alarming them about the value of the houses they own.

Banks cannot say this because their capital requirements are calibrated to their current asset valuations. Publicly acknowledging that the mortgage book is mispriced would trigger regulatory action, capital raises, and equity sell-offs. The incentive structure of every major UK bank is to maintain confidence in the valuation, not to question it.

Pension trustees cannot say this because their members’ retirements depend on the valuations being honoured. A trustee who warned of a revaluation would be accused of creating the panic they were warning against. The professional incentive is to stay quiet, manage the exposure at the margin, and hope the problem resolves itself.

The Treasury cannot say this because its own long-term fiscal projections assume rising property-linked receipts. Stamp duty, capital gains tax on property, inheritance tax on property, and the council tax base all flow through the housing market into government revenue. A revaluation of UK housing is, automatically, a downward revision of the fiscal forecast, which the Treasury has no incentive to publish.

The Bank of England has stress-tested the banks against a 28% fall. It has not publicly modelled an AI-driven services wage compression as a distinct scenario, because no institution has yet produced the underlying labour market data in a form that the Bank would consider sufficiently evidenced. The risk exists in the financial landscape, but it is not yet on any official balance sheet.

The silence is not a conspiracy. It is the emergent property of a system in which every major actor has a structural incentive to stay quiet. That is how markets misprice things. Not through collusion, but through the accumulation of individually rational silences into a collective blindness.

Britain did not choose this trade explicitly.

No politician stood on a platform and promised to underwrite the nation’s pension, banking, and housing wealth against the wages of a labour force that would shortly be targeted by the first general-purpose cognitive automation technology in human history. No Treasury official sat down and modelled the scenario. No bank risk committee formally approved the exposure. The trade was made one mortgage at a time, over four decades, through a combination of deindustrialisation, services expansion, mortgage market liberalisation, pension reform, and the steady professionalisation of the London and South East economy. Each individual decision was rational on its own terms. The aggregate outcome is a household balance sheet unusually exposed to the wages of a labour force that is, in turn, unusually exposed to the first AI wave.

And the technology that most directly targets it is, by all serious accounts, still in its early deployment phase.

The children of the middle-class households that currently hold the bulk of UK property wealth will reach their peak mortgage-servicing years in the 2030s and 2040s. The AI wave they will be working through is unlikely to spare the occupations that underwrote their parents’ houses. Some of them will emigrate. Some of them will take lower-paying work. Some of them will find AI-complementary roles and prosper. But the aggregate labour market in which they operate will look meaningfully different from the one in which their parents’ mortgages were underwritten, and the houses their parents own will clear at whatever price those future wages can support.

The trade is still open. It has not yet moved. But it is not a trade that any rational household would choose to make if they understood its terms.

When the call comes, it will not come from the Treasury, the Bank of England, the Office for Budget Responsibility, the pension industry, the banking sector, or the political class. All of them are structurally unable to call it.

It will come from the data. From the first quarters of PAYE receipts that undershoot forecasts in AI-exposed sectors. From the first year-over-year decline in average services wages in categories that once reliably rose. From the first housing transactions that print at prices the mortgage market will no longer support. From the first bank stress test that incorporates a labour automation scenario because it can no longer credibly exclude one.

Some of those early signals are already here.

Payrolled employment has already begun to contract. Redundancy planning is already running at a level not seen since the financial crisis, even without a crisis of the same kind. Some AI-exposed early-career categories, especially technology graduate roles, have already fallen by close to half in two years, while broader UK entry-level postings are down by roughly a third since ChatGPT’s launch. Both regular and total pay growth have now fallen below 4% for the first time since 2020. The labour market data is already moving in the direction the scenarios in this essay describe, though the data does not yet prove AI is the cause.

The housing market has not yet followed. But housing markets do not respond instantly to labour market shifts. They lag. Transactions are slow to complete. Sellers hold out at aspirational prices. Mortgage books continue to perform until arrears begin to rise. The gap between a change in the underlying wage base and a revaluation of the asset it supports is typically measured in quarters rather than days. What the early labour market data suggests is that the lag has started.

When the call comes in full, the trade will move fast, because it is a trade that everyone made individually and nobody was hedged against collectively. There will be no orderly revaluation. There will be a discovery moment, and then a reassessment, and then a price discovery process that will play out across housing, banking, pensions, and the fiscal base simultaneously.

The numbers in this essay are from the Office for National Statistics, the Bank of England, Savills, the International Monetary Fund, the Organisation for Economic Co-operation and Development, Goldman Sachs, the UK Parliament’s House of Commons Library, the Insolvency Service, the Institute of Student Employers, the Chartered Institute of Personnel and Development, Boston Consulting Group, the World Economic Forum, and the UK government’s own January 2026 assessment of AI and the labour market.

Most of them came from Whitehall.

Nobody in Whitehall is connecting them.

£9.18 trillion rests on the continued earning power of one of the most AI-exposed labour forces in the developed world. Nobody chose the trade. Nobody is pricing the trade. Every institution with a view into the trade has a structural reason not to describe it.

Britain bet the house on wages that AI is coming for. The bet is still live. The market has not yet been called.

But the labour data has started to move.

I have been writing for several weeks on X now about the systems that underpin modern Britain and the ways in which each of them is failing. The housing market that no longer clears on income. The care system that cannot staff itself. The fiscal pillars that rest on assumptions no longer supported by the underlying demographics. The slow, steady emigration of young British professionals to countries that did not pay for their training. Each of these posts has been an attempt to describe one part of a larger picture.

I have mentioned HS2 in passing more than once during this series, but always as an example rather than a subject. It deserves a post of its own, because HS2 is not a failed transport project. It is something altogether stranger and more revealing than that. HS2 is a mirror. A lens through which the condition of modern Britain can be seen with unusual clarity, rendered not in statistics or fiscal projections but in concrete, steel, and unspent billions. Every failure mode of the country that I have been writing about separately is visible in this single railway. Hold it up to the light and the whole system comes into focus.

The numbers are where any honest account has to begin, because they are so extreme that they do most of the argumentative work by themselves. HS2 was developed from 2009 with early estimates of roughly £31 billion to £36 billion for the full Y-shaped network that would connect London to Birmingham, then split northward to reach Manchester on one side and Leeds on the other. By 2013 the official funding package had risen to £42 billion.1 By 2019 estimates were in the high fifties. By the time of the Oakervee review in 2020, they were above £80 billion, with some scenarios reported at £106 billion.2 In 2021 the eastern leg to Leeds was substantially curtailed. In October 2023 the northern leg to Manchester was cancelled by Rishi Sunak in a speech delivered, with a kind of unintentional cruelty, at the Conservative Party conference in Manchester itself.3 What remains is a radically reduced project whose opening stage is focused on Old Oak Common in west London to Birmingham Curzon Street. The government says it remains committed to the onward connection to Euston, and tunnelling between Old Oak Common and Euston began in 2026, but the Euston station itself now depends on a still-evolving public-private financing model.4 The Phase 1 estimate published in June 2024 was £54 billion to £66 billion in 2019 prices, an estimate the Department for Transport itself says it does not consider reliable. In current-price terms, public reporting and parliamentary scrutiny have put the likely final cost closer to the high tens of billions.5 Nobody inside government is prepared to commit publicly to a final figure.

Per kilometre, these numbers place HS2 among the most expensive high-speed railways ever attempted. The Economics Observatory, drawing on the Transit Costs Project, has put the cost of the London-Birmingham section at roughly $232 million per kilometre. By way of comparison, Germany’s Wendlingen-Ulm high-speed line, which runs through mountainous terrain and required extensive tunnelling, was delivered at around $71 million per kilometre. The Lyon-Turin line, which tunnels through the Alps, has cost about $130 million per kilometre so far. France’s most expensive high-speed line ever built cost less than a fifth of HS2 per kilometre. Spain has been building a national high-speed network for decades at roughly a tenth of HS2’s unit cost.6 These are not obscure examples. They are the comparators every infrastructure economist reaches for. HS2 is an order of magnitude more expensive than the projects it is directly comparable to, built in peer European economies with advanced engineering capacity.

It is worth pausing to place this in a longer historical frame. The Victorians built the first stretches of the London Underground in the 1860s using pick, shovel and steam shovel, through the most densely populated city on earth, funded by private capital. A century and a half later, HS2 is being built through mostly empty countryside, funded by the state, using tunnel boring machines of a sophistication the Victorians could not have imagined. What has changed is not the technology and not the ambition of the engineers. What has changed is the country itself. HS2 is not expensive because it is difficult. HS2 is expensive because it is British.

Once that fact is admitted, HS2 begins to function as a lens. Every pathology of modern Britain that I have been writing about separately becomes visible through the specific details of this one project.

Consider first what HS2 reveals about the country’s ability to cost its own ambitions honestly. Every significant British infrastructure project of the last two decades has come in late and substantially over budget. Crossrail was delivered around £4 billion over its original budget and years behind schedule.7 Hinkley Point C, the first new nuclear power station Britain has attempted to build in a generation, has moved from an original estimate of around £18 billion in 2015 prices to scenarios as high as £46 billion in 2024 prices, with first power now expected around the turn of the 2030s.8 The proposed Lower Thames Crossing is estimated at around £10 billion, a price that has itself become emblematic of Britain’s unusually high infrastructure costs.9 The National Programme for IT in the NHS was dismantled in the 2000s after a forecast cost of nearly £10 billion, with far less delivered than promised.10 These are not isolated failures. They are a consistent national pattern. The institutions that exist specifically to enforce discipline on public projects, the Treasury, the Office for Budget Responsibility, the National Audit Office, the Infrastructure and Projects Authority, produce report after report warning that costs are spiralling, and the projects continue regardless. The Infrastructure and Projects Authority labelled HS2 itself “unachievable” in July 2023, the most serious rating it issues, defined as a project whose successful delivery no longer appears possible.11 The project continued. The warnings, it turns out, are not mechanisms for intervention. They are mechanisms for the record. They exist so that when everything eventually goes wrong, the system can point to the paperwork and demonstrate that somebody, somewhere, said something. The warnings are part of the theatre.

Consider next what HS2 reveals about the country’s ability to hold a decision. Since the project was approved in 2012, HS2 has cycled through a succession of chief executives, chairmen and transport ministers. Six different prime ministers and eight different chancellors have held office during this time. Every significant design decision the project has ever made has subsequently been revisited, reopened, or reversed. The route has been redrawn in several sections. The Euston terminus has been repeatedly replanned, paused, and rescoped. The eastern leg to Leeds was cancelled in 2021. The northern leg to Manchester was cancelled in 2023. Whatever HS2 was originally supposed to be, it is not that thing now, and there is no longer a stable version of the project that matches the one originally sold to the public.

This is not an HS2 problem. It is the country. British energy policy has been reversed at every change of government for two decades. Airport expansion around London has been debated continuously for fifty years without resolution. Social care reform has been legislated, cancelled, re-legislated and indefinitely deferred. Net zero targets have been set as legally binding commitments and then quietly relaxed within a single parliament. Industrial strategies have been announced, formally launched, and quietly abandoned at intervals of roughly three years throughout the century. A country whose political system cannot hold a decision across a single electoral cycle cannot build anything that takes longer than five years to complete. Which turns out to be most of the things that matter.

Consider next what HS2 reveals about the country’s relationship with its own planning and consent machinery. The chair of HS2 Ltd has said the project required more than 8,000 separate consents for Phase 1 alone.12 Each of these has required surveys, consultations, legal sign-off, and in many cases appeals processes. The project has, in a quite literal sense, generated more documentation than track. The single clearest illustration of this phenomenon is the bat protection structure at Sheephouse Wood in Buckinghamshire, a one-kilometre covered structure designed to protect a colony of approximately 300 Bechstein’s bats from the disturbance of passing trains. The original 2019 cost estimate for the structure was £40 million. By 2024 the figure had reached £100 million. In June 2025 the Department for Transport confirmed, in a letter to the Public Accounts Committee, that the total cost of the one-kilometre section of railway at Sheephouse Wood, including the bat structure and associated civil works, was approximately £216 million in current prices.13

Natural England, the public body that was widely blamed for requiring the structure, has publicly stated that it did not require HS2 Ltd to build this tunnel, did not advise on its design, and did not influence its costs.14 HS2 Ltd chose the route. HS2 Ltd chose the mitigation. HS2 Ltd chose the size of the structure. The local planning authority, Buckinghamshire Council, initially refused permission, and four years of engagement followed before a Planning Inspector overrode the council’s refusal. One kilometre of railway through Sheephouse Wood, including a bat protection structure. Around £216 million. Eight years of planning. The same pattern holds at Euston, where an Architects’ Journal Freedom of Information request in 2023 revealed that design fees alone had exceeded £289 million before any station had been built.15 The terminus has been redesigned three times by different architectural teams. The original Grimshaw and Arup consortium was replaced in 2017 by a WilkinsonEyre and WSP team, only for the original team to be brought back the following year. The number of platforms has been reduced from eleven to ten. Construction has been repeatedly paused. More than £2 billion had been spent on the Euston site by the end of 2022, with £1.5 billion of that sum going on land purchases and preparatory works that serve a station that has not yet been built, to a design that has not yet been finalised.16

This is not an HS2 problem. It is the country. It is a planning system that can generate five thousand pages of documentation to support a single decision and then produce no building at the end of it. No major new public water-supply reservoir has opened in England for decades, despite a growing population and worsening water security. Grid connections for new industrial and energy projects are now queued into the 2040s. Housing targets have been announced and missed every decade since the war. Nuclear power stations have been announced and subsequently abandoned with a regularity that suggests the announcing is now a more significant national activity than the building. Britain cannot build houses, reservoirs, power stations, railways, or roads at anything like the rate its population and economy require. What it can build, in more or less unlimited quantity, is paperwork about why the thing has not yet been built.

Consider next what HS2 reveals about the relationship between the British state and its citizens. Along the full Y-shaped route of the original project, HS2 Ltd compulsorily purchased thousands of homes, felled ancient woodland that had stood for centuries, destroyed hedgerows, and divided communities. The justification for all of this was always the same. The public good of the completed line. But the line is now radically reduced, the eastern and northern legs cancelled, and the destruction nevertheless stands. Homes were taken for a railway that will never reach the places they were taken to connect. Woodland was felled for trains that will not run. Compensation was paid for disruption that will never produce the promised benefit. Along parts of the cancelled route, land was acquired, properties were disrupted, and local plans were reshaped for infrastructure that will now not be built as promised. The state paid the compensation, acquired the land, and then abandoned the project. The disruption is permanent. The benefit will not arrive.

This is not an HS2 problem. It is the country. It is a social contract in which the state continues to demand sacrifices of its citizens for promises it no longer keeps. Young people take on £45,000 of student debt for degrees that no longer lead where they were sold as leading. Homeowners pay tax on frozen thresholds for public services that no longer function at anything like the level the signage implies. Savers followed the rules for thirty years and are now told that their pensions are inadequate. Patients paid into an NHS that now keeps them on waiting lists for months or years. Workers trained for jobs that have been automated, offshored, or simply evaporated. The state has not simply broken one promise. It has broken the pattern of keeping promises, and in doing so has broken the basic bargain between state and citizen that every functioning country depends on.

Consider next what HS2 reveals about the country’s capacity to imagine its own future. HS2 was, arguably, the last transport infrastructure vision of truly national, generational scale that Britain attempted to sell to itself. There is no HS3. There is no successor project at anything like comparable ambition. The political class has stopped proposing at that level of ambition, because proposing has become a political cost without a corresponding political reward. To see how much has been lost, consider what the previous three generations of this country managed to do. In the 1960s Britain planned and built the first sections of the M1 motorway within a few years of breaking ground. In the 1970s it built the Humber Bridge, at the time the longest single-span suspension bridge in the world. In the 1980s it completed the M25 around London. In the 1990s it finished the Channel Tunnel to France. The Channel Tunnel was itself famously over budget and a financial disaster for its investors, but it was completed, and it has operated continuously ever since. The Victorians and Edwardians built us the railways and much of the Underground. The post-war generation left us the motorways, the National Health Service, and the welfare state. Each generation for nearly two centuries left something substantial behind for the next generation to inherit and build upon. Our generation is leaving a one-kilometre section of railway in Buckinghamshire, including a bat protection structure, that cost around £216 million, a terminus at Old Oak Common that was never meant to be the main terminus, and an extensive paper trail explaining why the project no longer resembles what was originally promised. Archaeologists of some future century will find the abandoned earthworks of the cancelled northern leg and wonder, reasonably, what kind of country this was that could begin something on such a scale and then simply not finish it.

Consider finally what HS2 reveals about the country’s capacity to tell the truth about itself. HS2 will never be called a failure by any government, of any party, at any point in the future. Successive ministers have already begun describing it variously as a success, a partial success, a strategic realignment, a transformation, and an optimisation. When trains eventually begin running between Old Oak Common and Birmingham, a ribbon will be cut. A minister will speak. The signage will never be updated to match the reality on the ground. And this is not an HS2 problem. It is the country. The NHS will never be officially called broken, although the waiting lists are now measured in years. The justice system will never be officially called collapsed, although Crown Court cases are being scheduled into 2028.17 The state pension will never be officially called unaffordable, although the OBR’s long-term projections show ageing-related spending, including pensions, putting sustained pressure on the public finances.18 The housing market will never be officially called a rationing system by inheritance, although it clears on parental wealth rather than income. Each failing institution will continue to be described in the language of challenges, reforms, and headwinds. The actual word for what is happening will never be used by anyone in power, because using the word would require action, and the action is beyond what the political system will tolerate. So the signage stays, and the gap between what is claimed and what is delivered widens every year.

HS2, then, is not a transport project at all. HS2 is a self-portrait. The inability to cost. The inability to decide. The inability to build. The inability to protect. The inability to imagine. The inability to admit. Every major failure of modern Britain is visible in this one railway, and anyone who wants to understand the state of the country does not need to read the fiscal projections, watch the gilt market, or study the migration statistics. They can drive up the M40 and look at the half-built viaducts of a project that no longer exists in the form it was sold as. They can look at a £216 million section of railway in a Buckinghamshire wood designed, in part, to protect a colony of bats. They can look at the portals of tunnels that lead to places the line will no longer reach. They can look at £289 million of design fees for a station that has been redrawn three times and still not built. Britain has, in effect, caught sight of itself in concrete and steel and unspent billions, and has not yet worked out how to look away.

The railway will eventually open between Old Oak Common and Birmingham. Trains will run. A ribbon will be cut. A minister will speak. And the country that stands on the platform on that day will not be the country that was promised the railway sixteen years earlier. It will be smaller, poorer, more divided, more indebted, and more exhausted. It will have lost much of what it thought it had. It will not have gained what it thought it was building. The children who were told, in schools in Leeds in 2010, that they would one day catch a high-speed train to London, will by then be in their mid-thirties. They will not be catching it. Some will have emigrated. Others will be living in homes they still cannot afford to own, served by a rail network that still depends on Victorian infrastructure, working in jobs that are being quietly automated, paying taxes into a fiscal system that cannot honour the promises made to their parents, let alone the promises made to them.

HS2 will run. It will be slower than the network it was originally meant to join. It will not reach where it was originally meant to reach. It will cost what Britain was told it would never cost. And it will be the exact shape of the country that built it.

That is the thing that is worth sitting with, after all the numbers are accounted for and all the comparisons have been drawn. HS2 is not a warning about what Britain might become. HS2 is a portrait of what Britain has already become, rendered at enormous expense in materials that will outlast the political careers of everyone involved in its delivery. It is a self-portrait in concrete. An admission, inscribed in engineering, of what this country has allowed itself to become over the last several decades.

The question is not whether HS2 will eventually open. It will. The question is whether the country that finally opens it will still remember, by that point, what it was originally trying to build. Or whether the signage on the station, bright and new and proclaiming a high-speed future, will simply be the final touch of decoration on a monument to everything we have forgotten how to do.