A generation of cities was reshaped by IT salaries. The recent layoff waves are testing that foundation. An open-ended look at what history tells us about anchor-industry shrinkage, the AI cost-savings thesis, and whether a company can stay rich when its customers thin out.
a
aiiqlabs academy
·10 min read
For two decades, "the IT job" was treated like a permanent feature of the modern economy — as reliable a foundation for a city as a port or a steel mill once was. Bangalore, Hyderabad, Pune, Dublin, Austin, Seattle, parts of Toronto, Tel Aviv, Kraków, Manila, and dozens of other cities did not just host tech work — they were reshaped by it. Property prices, school catchments, restaurant cultures, even local political constituencies were rewired around a class of well-paid software workers and the multipliers that flowed from their consumption.
The recent waves of layoffs — cumulative and rolling, rather than a single shock — have started to test that foundation. This essay does not try to predict where it lands. It tries, instead, to take the question seriously.
The code was supposed to change the world. Nobody mentioned the world it would change first was the one that wrote it.
Overture — The Cities That Code Built
A timeline of the modern tech labour cycle
Foundation
1991
India liberalises; Bangalore IT corridor takes shape
Birth of a global offshore-IT class.
1995–2000
Dot-com hiring frenzy in the US
First wave of "tech as a regional engine".
First reckoning
2000–2002
Dot-com bust; over a million tech jobs cut in the US
A reminder that software jobs were cyclical.
2008–2009
Global Financial Crisis
Tech recovered fast; many other sectors did not.
The boom
2010–2021
A decade of cheap money and growth-at-any-cost hiring
Big Tech headcount roughly tripled.
The reckoning
2022–2024
Layoff Wave I — hundreds of thousands of tech jobs cut globally
Framed as a post-ZIRP correction.
2024–2026
Layoff Wave II — explicitly framed around AI productivity
Structural in tone, not cyclical.
When an anchor industry stops growing — or shrinks
The historical archive on this is depressingly rich.
Detroit, 1950 → 2013. Peak population 1.85M; bankrupt at under 700,000. The cause was not a single event — it was decades of a quietly shrinking employer base meeting a tax structure and infrastructure footprint designed for the boom.
Pittsburgh, 1970 → 1990. Steel employment collapsed from over 90,000 to a few thousand. The city partially reinvented itself around healthcare, universities, and later software — but "partial" is the operative word. The working-class neighbourhoods that were the social spine of steel did not reinvent themselves; they thinned out.
Manchester (UK), 1850s → 1980s. "Cottonopolis" — the textile capital of the world — became, over a century, a city more famous for its post-industrial decline than for the boom that preceded it, before its later cultural reinvention.
Nokia and Finland, 2007 → 2014. At its peak, Nokia accounted for roughly 4% of Finnish GDP and a quarter of the country's corporate R&D. When mobile collapsed, Finnish growth flat-lined for years. The engineers Nokia released, however, seeded a remarkable startup ecosystem — Supercell, Rovio, and others — and the state had the institutional capacity to soften the landing.
Bangalore, 1991 → present. A city that grew from roughly 4M to over 13M in three decades, on the strength of services exports. The same growth that built tech parks and ring roads also built a property market and a cost base that assume continued IT hiring.
The good and the bad of the boom era
The good is easy to list and easy to forget:
A genuine middle class formed in places that did not have one — particularly in India, the Philippines, Eastern Europe, and parts of Latin America.
Female participation in white-collar work rose sharply in IT-heavy regions.
Public infrastructure — metros, airports, hospitals — got built on the back of the tax revenue.
A culture of upward mobility ("study hard, get into engineering, lift the family") became plausible for tens of millions of households.
The bad is also real:
Regional inequality widened: an IT corridor inside a city often dragged real-estate prices beyond the reach of teachers, nurses, and small-business owners in the same city.
Local economies became monoculture-like: large parts of a city's restaurant, retail, gym, daycare, and real-estate sector existed because of one industry's payroll.
A cohort of workers built their entire financial life — mortgages, school fees, parental support, lifestyle — on the assumption of a 7–10% annual salary trajectory that the industry had taught them to treat as normal. They had built homes on a number. The number was a story.
In many of these cities, IT was not just an employer. It was the indexation mechanism — the thing everyone else's prices, rents, and ambitions were quietly indexed to.
Open questions
When a city's anchor industry shrinks, how long is the lag before tax revenues, school enrolments, and real-estate prices follow? In Detroit it was decades; in Nokia-era Finland it was years. What is the right comparable for an IT-anchored city?
Are we looking at a Pittsburgh-style partial reinvention, a Detroit-style slow hollowing, a Manchester-style multi-generation transition — or something the historical archive does not have a template for?
If IT was the indexation mechanism for a city's entire cost base, what happens to that cost base when the indexation stops, but the debts taken against it (mortgages, civic bonds, school expansions) do not?
And the uncomfortable one: who gets to stay in a city whose cost base was built for a wage class that is shrinking?
Crescendo — The Domino Question
There is a clean, almost too-clean version of the AI investment thesis: a software company replaces a large share of its engineers with AI, ships more product, takes the savings to the bottom line, and compounds. The chart goes up and to the right.
But software does not exist in a vacuum. Walk the cycle deliberately.
The cycle, drawn out
Company A — a software vendor — sells a core banking platform to large banks.
The banks make money primarily by lending: mortgages, auto loans, SME loans, credit cards.
Borrowers are people and businesses who can credibly service that debt — i.e., people with stable, well-paid jobs and businesses with stable, well-paid customers.
A meaningful slice of those borrowers — especially in tech-heavy metros — were precisely the IT workers and the businesses that served them.
The banks' software bills to Company A were ultimately paid out of the spread the bank earned on those loans.
Now collapse step three only partially. Not catastrophically — just partially. Default rates tick up. New loan origination slows. Net interest margin compresses. Risk-weighted assets get re-rated. The bank is not going under; it is just less profitable.
What is the first line item a CFO under pressure looks at? Discretionary tech spend. Then non-core platform contracts. Then renegotiation of "strategic" vendor contracts at renewal — Company A's contract.
The chart that was supposed to go up and to the right starts to bend.
The historical rhyme: 1929
This is not a new pattern. The 1920s in the US were a productivity boom — electrification, the assembly line, scientific management. Output per worker rose dramatically. Corporate profits soared. But wages and consumer purchasing power did not keep pace, and the gap was papered over with consumer credit and stock-market leverage. When the credit cycle turned in 1929, the productivity miracle did not save the economy — because the customers of the productivity miracle were the same workers whose wages had been suppressed by it.
Henry Ford understood the loop — and famously paid his workers $5 a day so they could afford to buy his cars. Whether or not the historical Ford was that altruistic, the economic insight is real.
A company that automates away its own customer base eventually runs out of demand.
The AI version of the same question
The optimistic thesis says: yes, fewer employees per company, but new companies, new categories, new demand. The pessimistic thesis says: capital is now more substitutable for labour than at any prior point in history, and the new categories may not absorb the displaced labour at the same wage level.
A few specific mechanisms make this domino question non-trivial:
Concentration of payers. Software-as-a-service economies depend on a relatively narrow set of well-funded buyers (enterprise IT budgets, in turn funded by enterprise margins). If margins compress across enterprise customers, software vendors feel it before consumer brands do.
Shadow demand from the IT class. A surprising amount of "consumer demand" in tech-heavy cities — premium real estate, private schools, fitness, dining, travel — is funded by the same paychecks AI now argues should be smaller and fewer. Hit that paycheck class and you hit a category of demand, not just individuals.
Banking sensitivity to employment. Banks are exposed not only to defaults but to the option value of future borrowing. A 30-year-old engineer who feels secure takes out a mortgage. The same engineer who is uncertain about their next role does not. Loan books shrink without anyone defaulting.
Government revenue. A non-trivial share of state and municipal tax bases in tech-heavy regions comes from IT payrolls and the property values built on top of them. Public services — including the ones that make a city investable — are downstream of that revenue.
So: can Company A become super-rich on cost savings alone, while its customers' customers thin out?
In a closed system, no — it is the same arithmetic that broke in 1929. In an open system, where new demand emerges from places we have not yet identified, possibly yes. The honest answer is that we do not know which system we are in, and the people most loudly claiming to know are usually the ones with positions to defend.
Open questions
Is the productivity gain from AI being captured primarily as margin (which concentrates wealth) or as price reduction (which redistributes it)? Historically, the answer to that question has determined whether a productivity boom ended in shared prosperity or in 1929.
If enterprise software's end customers are ultimately consumers with jobs, what is the realistic floor on enterprise software revenue when consumer employment weakens broadly?
Can AI-leveraged companies grow their customer count fast enough to offset shrinkage in spend per existing customer?
And the uncomfortable one: does a company that automates away its own customer base have a viable long-term business — even if its short-term P&L is spectacular?
Coda — Dark Mirror
The Weight You Were Never Asked to Put Down
Before we talk about what comes next, let us stop — just for a moment — and say what rarely gets said.
A generation showed up. Not just the engineers — the nurses, the teachers, the delivery drivers, the small-business owners, the parents who stretched paychecks across school fees and parental medicines and EMIs that were sized to a salary band that was supposed to keep rising. They built ring roads with their commutes, schools with their hopes, entire weekend economies with the quiet, unfanfared rhythm of paychecks that simply showed up every month.
You were one of those paychecks. And it mattered more than anyone told you.
The honest part of this transition — the part the spreadsheets cannot capture — is that the people most exposed to it are also the people who have already given the most to build the system that is now changing. That is not a footnote. That is the moral centre of the question. Any answer to "what comes next" that does not begin there is not actually an answer.
Civilisations are not built by algorithms. They are built by people who showed up.
The shape of the dark
It is fair, reading the above, to feel a chill. A generation of workers built lives — mortgages, families, parental responsibilities, civic identities — around a job category the industry now describes as "leverageable," a word that, in plain English, often means "fewer of you."
The dark version is worth saying out loud, because pretending otherwise is its own kind of disrespect:
Cities built on a single high-wage industry tend to take a long time to recover when that industry shrinks, and the people who lived through the shrinkage rarely see the recovery.
Productivity booms that are not paired with broad wage growth have, historically, ended badly.
The political consequences of a hollowed-out aspirational class have, in every century we have records for, been ugly.
And yet.
The Luddites were not wrong about their looms. They were wrong about us.
Every previous transition of this kind — agrarian to industrial, industrial to service, service to digital — was forecast, at the time, in similarly grim terms by people who were not wrong about the immediate suffering but were wrong about the eventual shape of the world. Luddite-era weavers were not wrong that the loom destroyed their livelihoods; they were wrong that nothing would replace them. Nokia engineers were not wrong that their employer was collapsing; they were wrong that Finland would never produce another technology champion.
There are reasons — not predictions, just reasons — for a thin shaft of light:
Human demand is not a fixed quantity. Healthcare, eldercare, education, climate adaptation, infrastructure renewal, and care work are all under-supplied at current prices, and none of them are well-served by AI alone.
Capital that becomes very concentrated tends, eventually and messily, to be redistributed — through taxation, through political reform, or through the simple fact that capital with no customers is not actually wealth.
The same tools that make a small team enormously productive also lower the cost of starting something new. Whatever the next layer of the economy looks like, the cost of bootstrapping it has rarely been lower.
The regions that historically adapted best were the ones that combined strong public institutions, accessible education, and cultures that treated reinvention as ordinary rather than humiliating. Some places have those ingredients. Some do not. That, in the end, is a choice.
None of that is a guarantee. None of it makes the next five years easy for anyone whose mortgage was sized to a 2021 salary band. But it is also true that "this time it ends badly" has been the safest-sounding prediction in every transition — and it has been wrong often enough that humility is warranted in both directions.
History does not promise that things end well. It only promises that they end. What comes after is built by whoever shows up.
Closing questions, for the reader rather than the writer
If the IT job was the indexation mechanism for a generation, what is the next indexation mechanism — and is anyone you know betting on it yet?
If a company can become very profitable while its customer base thins out, is it rich in any sense that survives a full business cycle?
What would your city, your street, your household look like if the assumption of continuously rising IT wages were quietly removed from the spreadsheet?
And — perhaps the only question that ultimately matters — what is the smallest, most boring, most concrete thing you can do this quarter to make your own life less brittle to the answer?
The economy has always been a story about who absorbs the shock. The interesting question is not whether shocks are coming — they always are — but who is preparing to absorb them, and who is preparing to pass them along.
Disclaimer
This article is an open-ended analytical essay — not investment, financial, career, or policy advice. It is intended to provoke thought, not to recommend any specific course of action.
All historical references — Detroit, Pittsburgh, Manchester, Nokia/Finland, Bangalore, the 1929 productivity boom — are illustrative parallels, used to think more clearly about the present rather than to predict the future. Any specific city or industry may, of course, follow a different trajectory.
Statistics cited (layoff totals, employment figures, GDP shares, dot-com-era job losses) are drawn from publicly available reporting and may differ across sources and reporting methodologies. Where ranges are given, they are deliberately approximate.
The questions raised at the end of each section are intentionally open. Where they touch on personal financial, career, or housing decisions, please consult a qualified advisor before acting on any line of thinking inspired by this piece.
