China's AI Offensive Doesn't Start in the Lab — It Starts With Demographics

China's AI Offensive Doesn't Start in the Lab — It Starts With Demographics

· 6 min read

Auf Deutsch lesen →

Whenever China’s AI ambition comes up, the word “arms race” follows almost reflexively: China versus the US, model versus model, chip versus chip. I think that frame hides the real engine. This piece is prompted by a compelling interview with Jörg Wuttke — 27 years with BASF in China, former president of the EU Chamber of Commerce in China. His core thesis, well supported by outside data: China’s rush into AI and robotics is largely a response to a shrinking and ageing working-age population.

Two state-run mega-experiments

Two centrally planned interventions still shape China’s demographics today — and both are practically irreversible in their consequences:

  • The one-child policy (1980–2016): roughly 36 years in its strict form. Meant to curb overpopulation, it permanently skewed the age structure and the sex ratio and created the “4-2-1” structure — one child later supporting two parents and four grandparents.
  • Zero-COVID (2022): the abrupt full stop of the economy, whose aftershocks still linger. It further dampened willingness to marry and have children amid an already pessimistic mood.

You don’t have to overstate the policy as the single cause — the birth-rate decline had structural roots that predate 1980. But the state massively amplified and cemented the trend.

Ageing in fast-forward

All industrial societies age. China does it in fast-forward. The solid numbers:

  • China has been shrinking since 2022 — for the first time in over 60 years. In 2024, roughly 9.5 million births faced about 10.9 million deaths.
  • The working-age population peaked around 2011 and is set to shrink by about a quarter by 2050.
  • Over-60s: around 310 million (about 22 %) at the end of 2024, projected toward 40 % by 2050. The median age climbs from ~40 to ~52.

Wuttke calls it the “Japanification” of society — only without Japan’s wealth cushion and without China’s ability to soften it through immigration. Germany needs roughly 400,000 immigrants a year just to stay stable; for China, that scale is simply unthinkable. His blunt formula: from 2030, China will lose roughly a France-sized population every decade.

The sex imbalance — read correctly

Here’s a correction that gets lost in the interview. The often-quoted “111 men per 100 women” is the ratio at birth (the peak in 2004 was even above 121) — not in the total population. There, the ratio is about 105:100, which concretely means around 35 million more men than women (2020 census). That’s the hardest single figure.

Why the argument still holds: the strongly male-heavy birth cohorts of the 1980s–2000s are now of family-forming age. So there really are relatively few women who could have children — and increasingly they don’t want to. China’s fertility rate sits at around 1.0 (stability would need 2.1). As Wuttke puts it, many women are “almost in refusal mode” — children are seen as expensive, the future uncertain after the property crisis and COVID.

That’s why AI and robots

This is where the circle closes. China’s fifteen-year horizon deliberately targets three levers against demographics: AI (productivity replacing missing heads), robotics (replacing physically missing workers) and biotech (people staying healthier and working longer).

A Japanese government institute (RIETI) sums up the link: China aims for growth “without relying on an expansion of labor input” — i.e. through productivity, not more workers. Important nuance: Beijing itself frames this as an innovation strategy (“new quality productive forces”); the explicit demographic link is drawn mainly by outside economists. But that’s where the real logic sits.

The transition dilemma is telling: today China still has high youth unemployment, AI is already displacing analysts, banks are cutting staff. Right now there are too many people for the jobs — which is why there are even rules against layoffs justified purely by AI productivity (priority: social stability). Wuttke’s forecast: in about ten years this flips. Then China will need exactly this AI to replace the workers it will lack.

Two pictures from the ground

The AI health cube. A booth like an old passport-photo machine: blood, urine, blood pressure — tested automatically and matched against millions of other patients’ data points. According to Wuttke, a very high hit rate (he cites 98 %), with a human giving the final sign-off. The motive is sober: 1.4 billion people aren’t insurable in the classic sense — AI slashes cost and waiting times.

The robot nursing home. In Beijing, a test care home is running where robots are meant to take over routine tasks so carers have more time for people. Plenty still goes “slapstick” wrong — but Wuttke’s expectation is uncomfortably clear: within four to five years, Chinese care robots could stand in European nursing homes. Because the alternative isn’t “an unempathetic robot instead of a human,” but simply no human left to provide care. China’s shortfall of care workers was already around 5.5 million in 2024.

Why China can pull it off (technically)

Briefly, without romanticising it: China has the deepest engineering bench in the world and an extremely customer-close, risk-friendly development culture — “if it’s 80 % right, it’ll do,” rather than German 100 % perfection. Add a massive energy build-out as the backbone of compute-hungry AI. And even US chip sanctions only slow it so much: with DeepSeek, a Chinese model hit top-tier performance in 2025 at a fraction of the cost — and triggered the largest single-day market-cap loss in US history at Nvidia (around 589 billion USD). The lead shows in hard numbers too: in 2024 China installed 54 % of all new industrial robots worldwide (stock over 2 million) and holds more than 70 % of all generative-AI patents.

What this means for us

Three takeaways:

  • The “arms race” frame misleads. Reading China’s AI ambition purely as a US-China power struggle misses the demographic engine behind it.
  • It’s a preview for us. Europe is ageing too; we’re heading into the same care and skills shortages. China is — involuntarily — stress-testing the automation answers that are coming for us anyway.
  • Don’t extrapolate linearly. Wuttke’s core warning cuts both ways: don’t simply project China’s rise or its decline. A China falling below one billion people by 2060 produces geopolitical structures very different from the ones we recognise today.

For me, that’s the real punchline: the most fascinating AI story of our time isn’t a chip statistic — it’s a population pyramid.

If this piece made you think, feel free to share it — and ask yourself whether “AI against labour shortage” is already concrete in your world.

Sources (selection):