The pyramid scheme we call the social contract, and the AI that was supposed to save us from it
I had a lively, sprawling conversation with my in-laws over coffee recently that ended up, as these things sometimes do, on falling birth rates and the need for continued economic growth. It was the kind of debate I always enjoy - generous, curious, properly engaged - and I have been turning the conversation over ever since.
Bear with me here. I have been collecting fragments of this argument from different places for a while - things I have read, things I have half-remembered, things I have been chewing on without ever sitting down to lay them out next to each other. This conversation gave me the excuse. What follows is me trying to see what the fragments make when you put them in a row.
Baby panic
The discourse around birth rates has reached a strange pitch. Not just in the obvious places - JD Vance worrying about childless cat ladies, Elon Musk tweeting about civilisational collapse roughly weekly - but in serious newspapers, in policy papers, and increasingly on the kind of huge mainstream podcasts that shape how millions of people think about these questions. Steven Bartlett and Chris Williamson recently spent a long stretch of The Diary of a CEO hypothesising about why women are choosing not to have children, a conversation that earned them substantial backlash for, among other things, the irony of two childfree men in their thirties weighing in so authoritatively on the matter*. The framing in all of these places is almost always civilisational.
Something is being lost. Something must be done.
The unspoken implication is that women in their thirties have a moral duty they are shirking, and the spoken implication is that the state has a legitimate interest in correcting this.
That the panic has recently reached fever pitch might have something to do with the Resolution Foundation naming 2026 “the first year of a new demographic era in the UK”, with deaths beginning to exceed births by what they call an “ever-widening margin”**.
I want to take the panic seriously for a moment. Not the moral framing (have very little time for that) but the underlying concern. Why, actually, do governments care?
The honest answer is not the one they put in the pamphlets.
Plain-English Economics
Modern economies are built on a particular bit of arithmetic. At any moment, some people are working and paying tax, and other people are drawing on the things that tax pays for - schools, hospitals, pensions, social care, the whole apparatus we have built up over the past century to keep people alive and reasonably comfortable, from cradle to grave. For most of modern history the workers have vastly outnumbered the drawers, and that gap funded the system. As birth rates fall and lifespans extend, the ratio tightens. Fewer workers. More retirees. The maths starts to wobble. In some countries it has already broken. Two plus two no longer makes four. It makes about three and a half, and the missing half is the bit that pays for the care home.
Underneath that, there is a deeper assumption almost no economist will say plainly in public: every modern economy is built on the premise that next year will be bigger than this year. Not stable - bigger. Every government bond, every pension projection, every public spending plan, every mortgage market is collateralised against the assumption of perpetual growth. And one of the most reliable historical engines of that growth has been a growing population. More workers, more consumers, more taxpayers, more demand.
The relationship between population growth and GDP growth is old enough and well-evidenced enough to count as economic furniture - the OECD has been tracking it for decades, and the link runs through both the size of the labour force and the consumer demand it creates***. If the population stops growing, one of the main economic engines stops, and the whole edifice that depends on growth starts making uncomfortable noises.
So when politicians say “we need more babies”, what they mean, underneath, is “we need the growth assumption to keep holding, because everything we have promised everyone is collateralised against it” (not quite as catchy a sound-bite, ey?).
The cleanest illustration of this is the state pension, and I think it is worth a closer look because I think most people believe a story about how it works, that is not the story of how it actually works. I was pretty sure I understood this before I started writing this, but I went and checked the official sources to make sure I had it right. I did.
Once upon a time…
Here is the story most of us were told.
In the UK, you contribute throughout your working life to ‘the pension pot’, via National Insurance. The state invests your contributions, the pot grows over the decades, and when you’re ready to retire you get a state pension funded by the money the pot’s investments have made. Just like a private pension: money in, money out, with your name on it. This is the Beveridge promise - the post-war welfare state designed in 1942 by the civil servant William Beveridge, whose report became the blueprint for the NHS, the benefits system, and the modern state pension - you contribute while working and state looks after you when you’re not.
Here is the actual mechanism. The UK State Pension is unfunded. The Office for National Statistics says this in language so blunt it should be on a poster: there are no assets set aside****. Your National Insurance contributions are not invested for you. There is no pot. There has never been a pot. The contributions you pay in this year are paid out, this year, to current pensioners. The contributions that paid out to your grandparents came from the workers of their day. The contributions that will pay out to you, when your turn comes, will come from whoever is working then. It is a continuous handshake between generations, and the only thing holding it together is the assumption that the next generation will keep showing up*****.
If a private company ran its finances this way, the directors would be in court. We have a name for it. While the state version is not legally a pyramid scheme - the state can compel the next generation to participate, which is supposed to be reassuring and somehow is not - structurally it is exactly the same shape.
The reason we are now panicking about birth rates is that the system we built assumed there would always be more people next year. But the more people are not coming, and rather than redesigning the system we are trying to manufacture the demographic conditions the original design needed.
The panic, it turns out, is fiscal. It is dressed up as something more human, but the machinery underneath is actuarial. We are not really being asked to value families. We are being asked to keep the books balanced.
It is also worth noticing who is doing most of the panicking. The loudest voices on falling fertility belong overwhelmingly to men with very strong opinions about what other people should be doing with their bodies and their decades. I am going to leave that there for now, but it is not a small thing.
Where are you going with this, Laura?
Right. So that is where I had got to, sitting drinking coffee, debating with my in-laws and then debating with myself for the next three days.
And then, because my brain always goes here in the end, I started thinking about AI.
So…If the panic is really about productivity and dependency ratios and the maths of growth… if the demand for more babies is really a demand for more workers and more taxpayers… then maybe we are about to be rescued from the whole problem? Maybe the entire framing is about to become obsolete. Because AI promises us productivity that does not depend on a growing workforce. An economy that could work for the people who already exist, without demanding that more of them get born to keep it running.
Imagine what that could mean. Imagine telling women in their thirties that the state has stopped treating their fertility as some sort of fiscal projection. Imagine the question of ‘whether to have children’ going back to being a question people answer for themselves, on their own terms, with no instrumental pressure from above. Imagine a pension system that does not need to keep recruiting more contributors to honour its promises to the people already in it. Imagine the whole “we need more babies” panic becoming a historical curiosity, like the Victorian fear that women’s bodies could not withstand the speeds of the new railways^.
I want this to be true. I am not being ironic (earnest, perhaps, as my husband would say, but not ironic). I have built my working life around the conviction that this technology can do extraordinary things for ordinary people, and this is exactly the kind of thing I want it to do. The relief of imagining this future, is real.
And then I look at how AI is actually being built right now, and the relief gets harder to hold onto.
More economics
In the first half of 2025, almost all of the United States’ economic growth came from one place. Investment in information processing equipment and software made up only four percent of US GDP - roughly the same share as the entire farming sector - but it accounted for ninety-two percent of all GDP growth^^. To translate that out of economist into English: nine out of every ten dollars of growth in the world’s largest economy came from a tiny slice of activity, concentrated in data centres and AI infrastructure.
On the face of it, these figures look enticing: a small slice of investment generating a wildly disproportionate share of growth is exactly what a productivity miracle looks like. It is the dream every technologist and economist has been chasing for decades: more output from less input. If four percent of the economy is producing ninety-two percent of the growth, the leverage ratio is enormous, and that leverage is precisely what would let you decouple economic growth from headcount. This is the AI-as-rescue-from-the-pyramid scenario. This is the bit that makes me want to believe.
But… while the American economy grew, the American economy as most Americans experience it - jobs, wages, the high street - barely moved.
The analyst James Van Geelen at Citrini Research has a name for this phenomenon. He calls it Ghost GDP - output that shows up in the national accounts and never circulates through the real economy^^^. Data centres do not employ many people once they are built. The construction phase brings jobs; the operational phase needs a small technical team, a security contractor, and not much else. The wealth generated by all that compute does not flow out into local wages, local rent, local spending in the way that a steel works or a car plant or a shipyard used to. The growth is real on paper. It is absent in pockets. And the pattern is not an accident. It is what happens when the gains from a technology accrue to whoever owns the infrastructure, rather than to the workers who used to do the things the infrastructure now replaces.
The truth is that those figures (92% GDP from 4% investment) don’t tell us the whole story on their own. They are a measurement of something extraordinary happening in one place. What that something means for the rest of us depends on what happens next. Specifically, it depends on whether the gains from that AI-driven productivity get shared or captured.
This is the question I keep coming back to in everything I write about AI. Who benefits. The technology is not neutral on this and never has been. The same productivity gain can land in wages, in lower prices, in public services and tax revenue, in the kind of broad-based prosperity that makes the question of birth rates irrelevant - or it can land in shareholder returns, executive compensation, and the balance sheets of a handful of companies whose names you already know. Which of those happens is not a function of the technology. It is a function of the institutional choices we make around it. Tax policy. Competition policy. Labour bargaining power. Ownership structures. The boring stuff that determines whether a productivity miracle becomes shared prosperity or concentrated wealth.
And right now, those choices are pointing firmly in one direction. Wages have not kept pace with productivity in any developed economy for forty years. The labour share of national income has been falling. Corporate taxation has been falling. Competition policy in the tech sector is functionally absent. The pattern of who captures productivity gains has been settled for a long time, and AI is landing on top of that settled pattern, not reinventing it.
So what does this all mean?
Unfortunately I haven’t fully reconciled the two potential futures I see. In one, the thing I was hoping would rescue us from the pyramid scheme, might actually be building a different version of the same trap. The technology that could, in principle, generate the productivity and the systems efficiencies that, applied to our failing public services and changing workplaces, might genuinely change the world for the better. In the other, it looks like AI continues to funnel all the wealth upwards to the top of an even pointier pyramid. A handful of companies and their majority shareholders benefitting disproportionately while the rest of us are told to be grateful for the GDP figures.
I am still working this out. I do not have it resolved. I sat down to write this thinking I was going to land somewhere clean and instead I have ended up somewhere uncomfortable, which (while not all that helpful) at least is honest. The optimist in me still thinks AI could be the way out of our current predicament. The realist in me looks at who is currently building it, and where, and on whose terms, and is a lot less sure.
*Ellie Austin, “What Diary of a CEO, Modern Wisdom podcast hosts got wrong about childless women”, Fortune, 14 January 2026. The episode in question featured Steven Bartlett interviewing Chris Williamson and prompted significant pushback, including from Dr Faye Bate and the TikTok creator known as “The Girl With The List”.
**Resolution Foundation analysis cited in Christian Today, “Britain entering ‘new era’ of deaths overtaking births”, 28 January 2026. The Office for National Statistics’ own projections show a more gradual picture, with deaths and births running roughly in balance through the mid-2020s before tipping decisively from the late 2020s onwards.
***OECD, Economic Policy Reforms: Going for Growth, multiple editions; see also the OECD’s long-running work on the relationship between demographic change, labour force participation, and GDP growth at oecd.org.
****Office for National Statistics, State pension funds, response to FOI request: “The UK State Pension is unfunded, which means that its obligations are not underpinned by an actual fund or funds… There are no assets set aside to generate investment return and there is no net value.” Available at ons.gov.uk.
*****House of Commons Library, National Insurance contributions: an introduction: “The NIF works as a pay-as-you-go fund. Therefore, receipts from contributions in one year are spent in the same year for contributory benefits.” Available at commonslibrary.parliament.uk.
^For anyone interested: Fanny Kemble was the woman who disavowed the Victorians of this particular notion. A twenty-year-old actress and the celebrity of her day, she was offered a preview ride on George Stephenson’s Rocket on 26 August 1830 by her host Lady Wilton at Heaton Hall, three weeks before the official opening of the Liverpool and Manchester Railway. She wrote a long, vivid letter about the experience shortly afterwards, comparing the locomotive to a horse and describing the sensation of speed as flying. She remained an enthusiast for railways, an actress, a published writer, and later in life an abolitionist and outspoken critic of slavery in the American South. Her womb stayed where it was.
^^Jason Furman, Harvard Kennedy School, analysis circulated September 2025; reported in Fortune, “Without data centers, GDP growth was 0.1% in the first half of 2025”, 7 October 2025.
^^^James Van Geelen and Alap Shah, The 2028 Global Intelligence Crisis, Citrini Research, February 2026. The piece is explicitly framed as a thought experiment rather than a forecast, but the Ghost GDP concept it introduced has since entered wider circulation.