As of 2026, humanity has likely fallen below the fertility rate needed to replace itself, a first that survived neither wars nor pandemics before now. Economists Jesús Fernández-Villaverd e and Patrick Norrick lay out the case in a new paper posted this month, arguing the decline is concentrated among poorer, less educated women in low- and middle-income countries, not the wealthy nations demographers long expected to lead the fall.

The paper lands as venture investors are already repricing demographic risk into return models. Robotics funding reached $18.8 billion globally through mid-2026, ahead of the $15 billion raised in all of 2025, and physical AI is now a headline thesis at firms including Bessemer, whose portfolio spans Anduril's valuation of $60 billion and Saronic's $1.75 billion round. VanEck frames aging populations as a direct driver of reshoring and automation demand, and Pictet puts robot density in Japan, South Korea and China at 40 per 1,000 workers, nearly four times the global average. Eldercare capital has moved slower. AgeTech investors surveyed by MedCity News called the category too nascent for a proven business model.

To build their case, Fernández-Villaverde and Norrick fit a single-factor model to 236 countries and territories dating to 1950. The common global driver of fertility peaked in 1978, they find, but the sharper story is country-specific trends: 219 of 236 nations are declining, losing an average of 0.062 births per woman a year, with none leveling off. The spread is wide. Mexico's fertility rate hit 1.51 in 2025, below non-Hispanic white Americans at 1.57, despite Mexican GDP per capita running near 28% of the US level. Tunisia posted 1.45 with GDP per capita at 18% of the US figure, and Thailand fell to 0.87. No society in history, the authors note , has reached the fertility levels now seen in South Korea, China, Thailand, Colombia and Chile.

The math behind those numbers is unforgiving; because children arrive in whole numbers, the paper shows a cohort where 15 in 100 women stay childless, most have two children, and one in six has three or four still produces a cohort's fertility rate of 1.8. Raising childlessness to 26, shifting a few more women to one child, and emptying out four-child families drops that same cohort to 1.3, the threshold for lowest-low fertility, even though three-quarters of women behave exactly as before.

The authors reject the usual suspects; China's one-child policy cannot explain much, since Taiwan's fertility collapsed further than mainland China's despite never restricting births. Smartphones accelerate the decline, the paper argues, but are not its cause. Instead, Fernández-Villaverde and Norrick point to modernity itself , a mix of formal credentialing, rigid career ladders and shrinking kin networks that makes a third child expensive and childlessness cheap.

The paper also challenges the UN's own fertility model, arguing its Bayesian projections assume an automatic recovery unsupported by recent national registries in Colombia and Egypt, where 2023 births were overstated by as much as 36.7% and 17.7%. On that basis, the authors put the peak of world population around 2056, decades earlier than the UN's official estimate near 2084.

The growth math compounds as Japan's economy grew just 0.79% a year from 1991 to 2023, versus 2.60% in the US, but adjusted for working-age population , the gap nearly closes, to 1.33% against 1.73%. Swap the two countries' labor growth rates and the ranking flips entirely, with Japan outperforming the US by a full percentage point a year.

The stakes compound beyond any one sector or VC fad. Fewer people mean fewer researchers and slower innovation-driven growth , the authors write, unless AI offsets a shrinking pool of scientists and engineers. That reframes the automation trade many general partners are already making: less a bet on labor cost savings alone, more a hedge against a shrinking base of workers, consumers and inventors across South Korea, China, Italy and much of Latin America. For anyone underwriting decade-long return models, population, not productivity, may be the binding constraint on growth.