A Greying World, and a Continent Asleep at the Wheel

The world has reached peak child. AI and robotics are maturing fast enough to help care for ageing societies. China and the US are building. Europe, the continent that needs it most, is barely in the race.

There are roughly two billion children under fifteen on Earth. According to the UN World Population Prospects 2024 (the most recent revision; the next is due July 2027), that number peaked in 2020 and will barely budge for the rest of the century. The under-fives peaked in 2017. The under-twenty-fives, probably in 2023. Hans Rosling, the Swedish statistician who spent his career making population data legible, coined the term peak child around 2014. He was a few years early. We are now squarely past it, and the data from Our World in Data confirms the plateau.

Two billion children today. If current trends hold, roughly the same in 2100. Projections can shift, especially in sub-Saharan Africa, but the broad shape of the curve is settled: humanity is no longer expanding from the bottom of the pyramid. It is swelling from the top. The twenty-first-century demographic question is not “how many will we be?” but “how old?”

The projections, and why they disagree

The UN (WPP 2024) projects a global peak around the mid-2080s, roughly 10.3 billion, followed by slow decline to 10.2 billion by 2100. Global fertility, now at 2.3 children per woman, would cross below the replacement level of 2.1 (the total fertility rate, or TFR, needed to keep a population stable from one generation to the next) in the 2050s and drift to 1.8 by century’s end.

The IHME published a competing model in The Lancet in March 2024 with a much earlier peak: 9.7 billion by 2064, declining to 8.8 billion by 2100, with 76% of countries below replacement by 2050 and 97% by 2100. By their projections, only six countries (Samoa, Somalia, Tonga, Niger, Chad, and Tajikistan) would still have a TFR above 2.1 at century’s end.

Jorgen Randers, co-author of Limits to Growth, has pushed further: a peak at 9.5 billion by 2050, steep decline toward 6 billion by 2100.

The spread between these scenarios (6 to 10 billion at century’s end) reflects deep uncertainty about sub-Saharan Africa, the only region where fertility remains well above replacement. The IHME projects that by 2100, one in every two children born on the planet will be born there. But the direction everywhere else is the same: humanity ages, irreversibly. By the late 2070s, people aged 65+ will outnumber those under 18 globally. By the mid-2030s, people over 80 will outnumber infants under one.

So the pyramid inverts. What happens to economies when it does?

Acemoglu’s inversion: ageing drives automation

The usual answer is decline: fewer workers, lower output, fiscal collapse. Daron Acemoglu and Pascual Restrepo (MIT/Boston University), in a series of papers culminating in a 2022 publication in the Review of Economic Studies, found something that complicates this answer. Across 60 countries and over two decades of data, ageing turns out to be one of the strongest predictors of industrial automation. Ageing alone explains 35% of cross-country variation in robot adoption. US metro areas where the population aged faster invested more in robots. And there was no negative correlation between ageing and GDP growth across countries.

The mechanism is what they call “directed technological change”: when middle-aged workers (36-55) become scarce, firms invest in automation rather than accept lower output. The effect concentrates in industries with the greatest technical opportunities for automation. Productivity rises, labour share declines, and the economy adapts, though unevenly and with distributional consequences.

Germany, Japan, and South Korea are not robotics leaders despite their demographics. They are robotics leaders because of them. The US and UK lag in part because their populations are not ageing as fast. As Acemoglu told MIT News: “Quite a bit of investment in robotics is not driven by the fact that this is the next ‘amazing frontier,’ but because some countries have shortages of labour.”

An earlier paper (2017) had already flagged the broader puzzle: despite everything we assume about ageing killing growth, the data showed no such thing, provided technology responded to the demographic shift. The qualifier matters. Charles Goodhart (LSE) and Manoj Pradhan, in The Great Demographic Reversal (2020), argue that technology may not respond fast enough. The low-inflation environment of the past three decades, they contend, had more to do with demographics and globalisation than with monetary policy. As working-age populations now contract globally, they predict inflation returns, interest rates rise, and government debt becomes harder to service. Their warning to technology optimists: unless AI and robotics deliver a “massive increase in productivity per worker,” growth may stall. Japan, they note, had a “global escape valve” while its workforce shrank. “These options simply will not be available as the entire global manufacturing complex ages together.”

Whether Acemoglu or Goodhart proves closer to right is, in a sense, the central bet of the next two decades.

Japan pioneered the experiment. China is scaling it. AI may be the missing layer.

Japan is where the care-robot experiment started. A third of its population is over 65. By 2040, the country faces a shortage of 570,000 care workers. The nursing sector has one applicant for every 4.25 open positions.

Japan has been building care robots for over twenty years. By 2018, the government had spent well over $300 million on R&D. The Moonshot R&D Programme allocates $440 million through 2050. Stanford’s APARC project has been tracking outcomes. What Japan has to show for it is modest. MIT Technology Review’s investigation found that care robots have struggled to gain real adoption. Many devices sit unused. The famous Robear lifting robot (2015) remains a prototype. The struggles trace to three levels:

1. Operational. Each device demands maintenance, training, and workflow adjustments that overburdened staff cannot absorb. Robots tend to create more work for caregivers, not less.

2. Economic. The robots were too expensive for individual facilities, even with government subsidies, and too bespoke to reach the production volumes that would have driven costs down.

3. Trust. This is the layer engineers tend to miss. Caring for a vulnerable person requires trust, and trust builds through recognition, tone, memory, the accumulation of small shared moments. Sherry Turkle (MIT), who has spent decades studying human-robot interaction in nursing homes, observed that elderly people will sometimes confide in a robot precisely because it does not judge them. But they also know, at some level, that nothing is listening. The elderly resist machines for intimate tasks, bathing, toileting, dressing, not because the machine is clumsy (though it often is) but because vulnerability demands a witness, and a witness must be capable of caring back. Intergenerational care is not just labour. It is emotional transmission: stories told, dignity maintained, a hand held at the right moment.

Japan approached the problem as a tailored engineering challenge. What it missed was that the harder problem was always UX and trust, and twenty years of R&D barely touched either.

China, on the contrary, is approaching robotics the way it approached EVs and solar: as a manufacturing-scale problem. In 2025, Chinese companies accounted for roughly 90% of global humanoid robot shipments. Unitree shipped 5,500 units, AgiBot around 5,100, while Tesla and Figure AI shipped about 150 each. Unitree’s G1 sells for $13,500; its newer R1 starts at $5,900. Chinese production capacity could reach 50,000 to 100,000 units per year by the end of 2026. Cities including Beijing, Wuhan, and Shanghai are opening training sites with simulated environments that include elderly care facilities, retail outlets, and smart homes, harvesting standardised data at scale.

Over 140 Chinese humanoid manufacturers have entered the market, most of which will fail. As Time noted, many of these robots cannot reliably perform skilled human tasks. But the pattern is familiar from every major technology transition: nascent industries need a buyer of first resort, and in China, the state plays that role. Massive public procurement, subsidised deployment in care facilities and factories, and preferential financing for robotics companies create the initial demand that no private market would generate on its own. This is how solar panels went from $76 per watt in 1977 to under $0.20 today, how Chinese EVs went from curiosity to global threat in a decade, and how the US military’s early purchases of semiconductors in the 1960s gave Silicon Valley its first market. The data advantage compounds on top: whichever country deploys more robots collects more training data, which produces better models, which enables better deployment.

Can this approach work for care? Manufacturing a car battery at scale and helping an 85-year-old with dementia get dressed are different problems. What makes care hard is not the physical tasks. It is that an elderly person in distress needs a human face, a familiar voice, a hand that adjusts its grip. You can automate the lifting and the pill-dispensing. But the relational core of care, the presence, the patience, the capacity to read distress in a face or hesitation in a voice, has until now been something only a human could provide.

AI changes both sides of this equation. On the hardware side, it makes robots dramatically more capable: a $6,000 humanoid that can barely walk becomes useful once its perception, language, and decision-making are driven by a foundation model trained on millions of hours of human interaction. On the trust and UX side, AI may be the first technology that can begin to meet vulnerable people where they are. Large language models adapt to the user rather than requiring the user to adapt to them. A 78-year-old who cannot handle a multi-step authentication flow can hold a conversation. And a growing body of clinical evidence suggests that AI handles the relational dimension better than expected. A randomised, double-blind study published in Nature Medicine in March 2026 found that AI agents delivering cognitive behavioural therapy, built on foundation models with a specialist clinical reasoning layer developed by Limbic, scored higher than licensed human therapists on standard clinical metrics. 74.3% of AI-powered sessions scored above the top 10% of human therapy sessions. In a real-world analysis of nearly 20,000 therapy transcripts, users with the highest AI exposure showed a 51.7% recovery rate, versus 32.8% for those with lower exposure.

AI excels at patience, consistency, availability, and adaptation to individual speech patterns and cognitive levels. It does not tire, does not lose composure, does not have a bad shift. It speaks any language, adjusts its vocabulary to the person in front of it, and when a conversation stalls (a common occurrence with cognitively impaired users), it can rephrase, simplify, circle back. These metacognitive abilities, knowing when communication has broken down and trying a different approach, are precisely what undertrained human caregivers often lack. For an ageing population, where loneliness, cognitive decline, and chronic disease management are the dominant care needs, conversational AI may turn out to be at least as useful as, if not more than, any humanoid robot.

Where is Europe?

China has 140 humanoid robot manufacturers and is running the EV playbook on an ageing society. The US has Tesla, Figure AI, and the world’s best general-purpose foundation models. Japan, despite its difficulties in care robotics, has two decades of institutional knowledge about what does and does not work with elderly populations, plus a $440 million Moonshot programme.

Europe has the oldest population on the continent after Japan, a silver economy estimated at €5.7 trillion, and almost nothing to show for it in terms of companies built to address it. Europe’s first age-tech accelerator launched in 2022. One. A 2025 study in the Journal of Business Research put it bluntly: “business awareness and strategic mobilisation remain limited; many firms underserve older consumers and underutilise older workers.” The McKinsey/Draghi diagnostic points to an annualised gap of €580 billion in corporate investment and over €300 billion for startups relative to the US. European venture funding as a share of GDP sits at 0.17%, well below the US.

Our continent will age fastest after East Asia. We have the clinical expertise, the regulatory frameworks for health tech, the public health infrastructure, and the social welfare systems that could serve as distribution channels. What we lack is capital deployed at scale, founders building for the 70-year-old, and a political class that treats ageing as more than a pension line item and a housing problem.

The current approach, across most European governments, treats demographic ageing as an infrastructure burden: more hospital beds, more trained nurses, higher retirement ages, adapted housing. All of which is necessary.

None of which is sufficient. What is missing is the recognition that ageing societies are also an extraordinary opportunity to build a desirable future and the companies that make it possible: health platforms that keep people healthy, engaged, and autonomous into their 70s and beyond, AI companions that fight loneliness and cognitive decline at scale, voice-first agents that make complex financial and medical decisions accessible, robotics that free human caregivers for what only humans can do. Andrew Scott (London Business School) estimates the “longevity dividend” at $56 trillion if societies adapt. In Europe, 90% of the increase in employment over the past decade, 17 million additional workers, already came from people over 50. The demand is here. The demographic clock is running. And for the first time, the technology is mature enough, or very close to it: foundation models that understand natural language, robotics approaching production economics, diagnostic AI that outperforms specialists on specific tasks.

Defence tech became a European venture category overnight when the geopolitical context demanded it. The demographic context is now demanding the same urgency. The ingredients are all European: world-class hospitals, universal health coverage as a distribution channel, a regulatory culture that the rest of the world tends to follow on health and data. Until recently, building for ageing populations was a hard sell. The technology was not ready, the market was fragmented, the returns were uncertain. That is changing. The convergence of foundation models, rapidly improving robotics (in both cost and capability), and a demographic wave that is no longer approaching but arriving makes this a moment where founders and capital allocators can build together on solid ground. Europe has every reason to lead here: the clinical depth, the regulatory credibility, the public health infrastructure, the social contracts that could serve as go-to-market channels.

Founders, it is time to build for the greying world. Capital allocators, it is time to finance it.