$GXO

Summary

The analyst is bullish on humanoid robots as a multi-year thematic and expects widespread adoption, citing rising labor costs and falling robot prices, with logistics companies AMZN, GXO, and MELI running trials with Agility's robots.

Reasoning

Structural labor scarcity and rising wages make automation via humanoid robots economically attractive as robot costs fall, leading to a capital allocation decision favoring adoption.

Paradis Research | Economics of Humanoids As many of you know, I'm bullish on humanoids as a multi-year thematic. While the technology looks cool with robots doing backflips and folding laundry...the root of my conviction rests on the boring (but reliable) macroeconomics of labour scarcity. Let me start with probably the most underrated fact in all of forecasting: the workers of 2050 have already been born. This is why working age population projections over a ~25 year horizon are some of the most reliable numbers in economics. Think of it this way - central banks and governments can shift the demand for labour (via stimulus) but they can't magic up new humans of working age from their asses. And simply....that structural labour supply picture is pretty bad (putting it mildly). Here's some of that data from key economics: China: - Population fell for a fourth year in a row - down 3.39 million in 2025 to 1.405 billion. - Just 7.92 million births in 2025, the fewest since 1939. - The core working cohort (16-59) shrank by ~6.62 million in a single year to 851 million. - Working age population (15-64) is projected to fall from 984 million (2024) to 745 million (2050)...a loss of over 25% of the labour force. Japan: - Face a shortfall of 11 million workers by 2040, concentrated in nursing care (25%), transport/drivers (24%), and construction (22%). - Working age population falling from ~75 million in 2020 to ~59.8 million by 2040, and the rapid decline beginning in 2027. - The 2025-2026 labour market already shows strain, with more job openings than applicants (a job : applicant ratio of 1.25). Korea: - Fertility rate of just 0.8 in 2025, which is the world's lowest and roughly 1/3 of the 2.1 replacement rate. - and even that, up from 0.75 in 2024 on a 6.8% rise in births to 254,500, was seen as a 4 year high "rebound." Germany: - Their statisticians now expect 1/4 residents to be aged 67+ by 2035 - versus 1/5 in 2024 - with the working age base (20-66) shrinking through the mid-2030s as the baby boomers retire - the freshest 2026 estimate in June (IW Cologne institute) puts that decline at 4.3 million workers by 2036. US: - In the US, there could be up to 2M unfilled manufacturing jobs by 2033, out of 3.8M needed (Deloitte). Overall, it's not looking good for the major economies. Cost of living is rocketing = people are having fewer kids. Then there's the price of labour which is a second-order derivative of the above. When labour gets scarce and expensive (as the above highlighted), the factory managers stop asking "can I hire?" and starts asking "what can I automate?" And we're seeing wages rocket in key economies like Japan (5.3% YoY) and China (8% CAGR). In the US, we see a similar trend in the manufacturing industry where worker salaries crossed $30 / hour for the first time ever in April. That then naturally leads us to looking at the actual cost of replacing a human for a humanoid: - On the labour side, US manufacturing total employer compensation averaged about $46.30 per hour worked in Q4 2025. - This loads out to well over $100,000 a year, and closer to $150,000-$156,000 once overhead is added on industry estimates. - The average humanoid price has collapsed from ~$85,000 in 2023 toward about $50,000 in 2025 (this number varies depending on if you include China). - Looking at Agility's investor presentation, BOM costs are set to fall drastically from economies of scale. Goldman Sachs have also noted the cost trajectory explicitly: "Where our analysts had expected a decline of 15-20% per annum, the cost declined 40%," which it said "could speed up the timeline to factory applications by a year." An average humanoid can run close to 20 hours / day, has no benefits, and costs around $15,000 a year to operate (depending on the OEM and if you buy it outright rather than lease). When the machine costs less than a year of one worker's fully loaded wage and can cover two shifts, payback compresses toward a year or less....evidenced by Agility's current payback being ~1 year on an expensive BOM cost basis currently due to low production volumes. That is then basically the whole thesis compressed: As wages rise and the cost of a robot falls, the two curves cross. Once they cross, adoption becomes a capital allocation decision...rather than a sci-fi one. In my opinion, we're about to see widespread humanoid adoption. With logistics companies like $AMZN, $GXO and $MELI all running trials in test sites with Agility's robots. And as humanoid safety improves, we'll start to see them interacting more with humans. Which opens up significant TAM in areas like last-mile delivery and elder care.

Return by horizon