Japan Is Building a Robot Workforce That Could Change the World Forever!

Robotics and Automation
4 followers
AI Generated·May contain errors
While the world argued about chatbots,
Japan decided to manufacture its replacement workforce
— 10 million AI robots by 2040 — in hospitals,
airports, construction sites, and nursing
homes.
You'll find out in the next 60 seconds.
Japan has about 450,500 industrial robots after 50 years
of building them.
The target is 10 million in 14 years,
22 times more, one robot for every 11 or 12 people,
And Japan has no other choice.
By 2040 it expects to be 11 million workers short,
so we are going to test whether the plan survives
contact with reality. You will see a dead
car factory in Kyoto turned into a humanoid assembly
line, a robot loading luggage at Haneda airport,
and why the passport of that robot matters more
than anything else in this story.
A 150 kilogram caregiver that gently rolls a man onto his side.
A hydrogen-powered robot horse that drew
an estimated 1.2 billion social media impressions
before it could carry a rider.
And a robot monk giving sermons in a Kyoto temple.
And underneath it all, the real contest.
Not Silicon Valley, but Tokyo against Beijing.
Over who builds the machines that do the world's physical work.
By the end, you will know whether Japan can pull it off.
And why one of its own factories, running flat out,
would need more than eight centuries
to reach the goal.
Start with the announcement itself.
On June 30, Japan's Trade Minister, Ryohei Akazawa,
unveiled a revised national robotic strategy.
Roughly 10 million AI-equipped robots deployed by 2040,
spread across 18 sectors.
This update added restaurants,
food manufacturing and medical care to the list.
Now notice the word most headlines get wrong.
It says robots, not humanoids.
Plenty of coverage pictures 10 million walking metal people.
The strategy is much wider.
Industrial arms, mobile machines, medical systems, logistics,
inspection, disaster response.
Humanoids are in the plan where they make sense,
but they are one piece of it, not all of it.
And the words AI-equipped matter just as much.
The industrial robots Japan has perfected since
the late 1960s are brilliant but blind,
bolted down, usually fenced off,
repeating one motion with extreme precision.
The new generation is supposed to see, decide and adapt,
working in places that change every hour and standing next to people.
That is the leap from a machine that follows
a script to a machine that follows a situation.
And it is why a plan built
on a technology Japan has mastered for half a century
Is still a moonshot.
So what does 10 million actually mean?
Remember, Japan is not a beginner.
The International Federation of Robotics ranks it
as the second-largest industrial robot market
on Earth. It installed 44,500 industrial robots in 2024,
it has 450,500 in operation,
And in 2023, it produced 38% of all
the industrial robots the world made.
So the new target is not growth, it is a different universe.
22 times the current fleet.
In a country the United Nations expects
to have 112 million people by 2040.
Picture a robot for every dozen citizens.
Why would a famously cautious country gamble like this?
Because the real driver isn't technology, it's arithmetic.
By 2040, Japan's population is projected
to be about 6.2% smaller than in 2030.
The median age will be 54. More than 35% of people will be 65 or older.
In 2024, Japan recorded fewer than 687,000
births against about 1.6 million deaths,
with a fertility rate of 1.15,
when you need 2.1 just to hold a population steady.
Feed that into the labor market,
and you get the number behind this whole story.
A projected shortfall of around 11 million workers,
according to the RecruitWorks Institute.
11 million jobs with nobody to fill them. You can already see it.
By one survey, 84%
of Japanese employers say they cannot fill open positions, the third
worst rate in the world.
In construction, there are nearly seven openings for every applicant,
seven employers fighting
over one person.
Most countries in this position would open the door to immigration.
Some reports say Japan would need more than
6.5 million foreign workers by 2040 to hit
its growth goals.
Today, foreign workers are fewer than 3% of Japan's caregivers,
about 57,000 people.
For cultural and political reasons, Japan has picked the other road.
If you can't find the workers, build them.
But a robot is a brain and a body,
and the plan treats those as two separate problems.
Let's start with the brain, because
that is where Japan did something unusual.
On the same day as the announcement,
the government commissioned a brand new company
to build it. It is called Noitra, and look at who owns it.
SoftBank, NEC, Sony Group and Honda
hold the majority, with Fujitsu
and Rakuten reportedly considering joining.
Four of the most powerful technology companies in Japan,
normally rivals, pulled into one national
project. Imagine the United States forcing Apple,
Google, Microsoft and Tesla into a single robotics
consortium. That is the scale of what just happened.
Their assignment is a homegrown AI model,
and not for chatting, for what engineers call physical AI.
Here is the difference in plain
English. A chatbot lives inside a screen,
and when it is wrong, it writes a bad sentence.
Physical AI looks through cameras and sensors,
decides what to do, and moves a body in the
real world. And when it is wrong, something gets dropped,
or someone gets hurt. The government has
committed up to 1 trillion yen over five years,
about 6.1 billion dollars, with roughly 387
billion yen this fiscal year alone.
A first version of the model is planned before the
fiscal year ends in March,
built together with the National Research Institute AIST.
And $6 billion is only one line item.
It sits inside a 14-year, ¥370 trillion growth strategy
that also covers chips, quantum computing and nuclear fusion.
Japan isn't dipping a toe in the water.
The minister has said the country's winning card is accumulated data.
Decades of experience in elder care,
disaster response, factory automation, and the cleanup
of the Fukushima Daiichi nuclear plant.
A brain trained on some of the hardest real-world jobs on Earth.
And brains need bodies, which is where a dead factory comes in.
Mitsubishi Motors had a problem most car makers would recognize.
An unused engine production line at its Kyoto plant.
Idle machinery that normally ends up as scrap.
Instead, Mitsubishi announced a partnership to turn
that floor into a mass production line for humanoid robots,
targeting 1,000 per month, with production starting in early 2027.
And here is the twist inside the twist.
The partner is not a tech giant.
It is Highlanders, a startup founded in 2023 by University of Tokyo
graduates, barely three years old,
now holding the keys to what Mitsubishi describes as the
first collaboration between a car maker
and a humanoid maker for mass production anywhere
in the world. The robot is based on a model unveiled
in 2025 with 19 degrees of freedom,
electric actuators for smooth movement,
and an AI that perceives its surroundings.
Plans its motions and responds to voice commands.
And the rollout is clever.
The first robots off the line will not be sold.
They will work inside Mitsubishi's own engine operations,
doing real tasks and generating real data
under automotive-grade quality control.
Only after that validation do they go on sale
to other manufacturers desperate for labor.
The robots literally earn their way out the door.
Other giants are attacking other weak points.
Panasonic is going after the humanoid's Achilles heel, battery life.
Its chief AI officer, Akira Sakabara,
says the company wants to use its battery expertise to
extend operating time,
starting with moving materials and tending machines, then moving
on to warehouse picking and packing.
Kawasaki, which has built industrial robots since 1969,
showed the ninth generation of its humanoid,
the Kaleido 9, at the end of 2025.
It stands 190 centimeters tall, weighs 99 kilograms,
builds a live 3D map of its surroundings
with laser sensors and stereo cameras,
spots the edge of a stair and adjusts its stride,
and can run on its own or be piloted remotely through a VR headset.
Kawasaki's roadmap is
refreshingly modest,
humanoids working in controlled places like factories around 2030,
supervised remotely at first and fully independent later.
It also runs a physical AI center in San
Alongside Nvidia, Microsoft, Analog Devices,
and Fujitsu, Toyota, the country's biggest
manufacturer, has humanoid prototypes
on a components line for evaluation,
and its research institute is working
with Boston Dynamics to put new AI brains into the Atlas
robot. And a Tokyo startup called Zeals unveiled the D1,
a 129-centimeter humanoid on wheels,
Priced from about 5 million yen, roughly $32,000.
It plans 100 units by March 2027
and 10,000 hours of real-world running time.
Notice the design logic.
It is sized to fit through standard Japanese door frames and ride
elevators, so hospitals and factories don't have to renovate.
That is what a maturing robot industry
looks like. Not flashy backflips.
A robot designed around the width of a door. So why
build robots in human shape at all?
Wheeled machines are simpler, cheaper, and more stable.
The answer is hiding in a baggage hall.
Tokyo's Haneda Airport handles more than
60 million passengers a year,
and behind the scenes it runs on brutal physical labor,
loading bags, handling cargo,
squeezing into tight spaces around aircraft, lifting heavy loads in
heat, cold, and rain. It is exactly
the work young Japanese people are no longer lining up for,
just as tourism explodes.
With more than 7 million visitors in the first two months of 2026,
After a record 42.7 million the year before.
So in May, Japan Airlines, its ground handling arm,
and a company called GMO AI and Robotics
began a trial that puts humanoid robots to work beside aircraft.
It is planned to last two to three years,
with the robots taking on more real work,
and humans stepping back as the system proves itself.
The reason for the human shape is JAL's entire logic.
An airport is built for people. Containers,
ramps, vehicles, doors. A humanoid slots
into that system without anyone redesigning aircraft or
ground equipment. The robots
in the demonstration stand about 130 centimeters tall. One pushes a
cargo container onto a conveyor beside an aircraft,
then turns and gestures to nearby staff.
Cameras and 3D sensors help it navigate
an apron where nothing stays still, and it runs
for two to three hours per charge.
JL Ground Services president said it is about easing
the burden on workers, and GMO's executive
pointed out that airports look automated from the passenger side,
while the back end still
leans heavily on human muscle.
The same logic applies on construction sites.
Shimizu Corporation, one of Japan's largest builders,
has started field trials and
Aiming for humanoids doing painting
and plastering by around fiscal 2030,
at a Tokyo site, a humanoid walked
an active construction zone autonomously at 1 meter per
second. No cage, no rails, no fixed track,
just a route and its own judgment about conditions that
change every day. A Shimizu executive admitted
that earlier construction robots needed human
help to move, kept needing rework,
and everyone said a person could do the job faster.
What changed is the learning.
Alongside the humanoids,
Shimizu is testing robotic arms that learn painting by watching
and imitating human workers. Not programmed. Taught like an apprentice.
and a startup called Donut Robotics showed
a 130 centimeter humanoid that picks up new movements
from video that is the difference between a robot
that follows instructions and one that learns a
job and it is what makes the 10 million number thinkable at all
because a robot you can teach
by showing is a robot a small business can afford to deploy
so the logic holds together
If the world is built for humans,
the easiest robot to deploy is one shaped like a human.
But there is a detail in those airport photos
that I want to come back to,
and it changes how you read this entire story.
The humanoid in those Haneda photos is not Japanese.
It is a Unitree, built in Hangzhou,
China. Japan's flagship airport trial is running on Chinese hardware,
supplied through a Japanese
integrator, with a second Chinese machine,
from UBTECH, assigned to the heavier jobs.
Now look at the scoreboard.
China has more than 2 million industrial robots in operation,
over four times Japan's fleet.
In 2024, China alone accounted for 54% of all industrial robot
installations in the world, 295,000 machines, against Japan's 44,500.
On humanoids, China has
set a target of 10,000 commercial units by the end of 2026,
working across 100 real-world scenarios.
And Japan, the country that gave the world ASIMO,
watched Honda retire it back in 2018.
So is Japan already losing?
Not so fast, because Japan is trying to play a different game.
China's strategy is volume, flood the world with affordable hardware.
Japan is betting that hardware turns into a commodity,
and the profits go to whoever
owns the intelligence. It is the smartphone lesson.
The money didn't go to whoever stamped
out the most handsets. It went
to whoever controlled the operating system.
That is what a sovereign AI model is for.
Add alliances, with reports that Japan's
industrial giants are teaming up inside Nvidia's
physical AI ecosystem. And add proof in the
toughest environments on Earth, like the Fukushima cleanup,
and the shape of the plan appears.
Japan doesn't need to win the hardware race if it can own the brain
and prove it where it matters.
The government is backing that bet with targets and rules.
Reports say the trade ministry wants Japan
to capture around 30 percent of the global physical
AI market by 2040, and that industrial heavyweights like Fujitsu
and Fanuc agreed in July to team up on
robot development to stay ahead of Chinese rivals.
Even the rulebook is being rewritten.
Japan's current industrial safety laws assume robots live behind cages,
not in hallways beside
nurses and baggage handlers,
and new guidance for mobile humanoids is reportedly on the way.
Which brings us to the hardest job of all.
It isn't on a tarmac. It's at a bedside.
In a laboratory at Waseda University in Tokyo,
a 150-kilogram humanoid called Erek leans over a man lying on his back.
It places one hand on his knee and the other on his shoulder
and gently rolls him onto his side.
If you have never worked in care, that looks trivial.
It isn't. It is the maneuver used to
change diapers and prevent bed sores,
and caregivers perform it dozens of times a day,
wrecking their backs in the process.
Getting a machine to do it on a fragile human body,
with exactly the right pressure,
is one of the hardest problems in robotics.
The professor behind it, Shigeaki Sugano,
made a point that cuts through the humanoid hype.
Humanoids are being built all over the world, he said.
But they mostly do chores and factory work.
They rarely touch people.
The moment a human is in the picture, everything changes. Safety,
precision, adapting to every individual body.
That is the frontier AIREC is attacking.
And the name says it all.
Short for AI-driven robot for embrace and care.
Beyond rolling patients, it can help a person sit up,
put on socks, cook scrambled eggs and fold laundry.
Sugano expects it to be ready for nursing facilities around 2030,
at a starting price of at least 10 million yen, roughly $67,000.
A government moonshot research program is now training
the next generation of these systems in NVIDIA-powered simulation,
so a robot can practice repositioning
a patient thousands of times while accounting
for each person's physical condition before it ever touches a real one.
Expensive?
Sure!
Now consider the alternative.
In Japan's nursing sector,
there is one applicant for every 4.25 open jobs.
The entire baby boom generation turned 75 or older by the end of 2024.
A care worker who watched the demonstrations
said he doubts robots can understand everything
about nursing, but hopes for a future where robots
and humans work together.
And that is the real vision. The part Western coverage keeps missing.
In Japan, the robot isn't coming for your job.
It is filling the job nobody is left to do.
The shift where a bedridden patient waits longer than they should
because there is no one to help.
That cultural comfort is why Japan can even attempt this.
Hollywood grew up on Terminators.
Japan grew up on Astro Boy, robots as helpers, companions,
even something close to sacred.
In February, a Kyoto temple introduced a robot monk called
Buddha-bot, performing rituals with its hands pressed together,
powered by an AI trained
on Buddhist scriptures that answers personal questions
the way a monk would. It was built
Built with researchers at Kyoto University.
And then there is Corleo.
Kawasaki's hydrogen-powered four-legged robot horse
was shown as a concept at Expo 2025 in Osaka,
with an arrival date of around 2050. Then the internet noticed.
Now, Kawasaki has a dedicated team reporting straight to its president.
Rideable prototypes aimed at Expo 2030 in Riyadh,
and a sales target around 2035.
A 150cc hydrogen engine powers four independent legs,
the only exhaust is water, and you steer
by shifting your weight like a horse.
You may one day buy a robot horse
from the company that builds Ninja motorcycles.
Fun, but remember why all of this exists.
Now give the skeptics their full case, because the math is brutal.
Japan needs roughly 9.5 million more robots in 14 years,
around 680,000 a year, starting now.
In 2024, it installed 44,500.
That is 15 times today's pace.
And that is my back-of-envelope math.
But it shows the size of the climb.
Mitsubishi's showcase line,
1,000 humanoids a month, makes 12,000 a year.
At that pace, one factory would need more
than eight centuries to reach 10 million.
Of course, nobody expects one factory to do it.
And most of those robots won't be humanoids.
But it tells you where the bottleneck is.
There is also the gap between a demo and a deployment.
In the Honda footage, one trade outlet noticed
that the conveyor belt the robot pushes cargo onto
still appears to be run by a person.
The trial's own plan starts with mapping where robots can
work safely, then simulations, and only then real operations.
And adoption is still thin.
One survey in May found only around 4%
of Japanese firms already running AI robots.
5% planning deployments and about a quarter considering them,
although among transportation
equipment makers, 80% were using them or studying them.
The wave is building from the factory floor up,
but it has not broken yet.
Then there is the payoff, which nobody has proved.
The Korean outlet Digital Today noted
that real gains in productivity, cost, and safety
still have to be verified through large-scale deployment at worksites.
Insurers are already preparing
for accidents involving workplace robots,
with Mitsui Sumitomo Insurance buying
into a physical AI developer in July.
And then price, at roughly $32,000 for a wheeled helper
and $67,000 for a caregiver.
These machines are not cheap,
and the ones working most visibly today, at Honda, are not Japanese.
Even Spain's Exatica pointed out that the plan leaves open questions,
like how the 10 million will divide among industrial,
medical, and service machines.
The supporters' answer is not weak.
The target counts robots broadly, not humanoids alone.
And Japan already has half a million in operation.
Prices fall as production scales.
The Mitsubishi model answers the demo problem directly.
Robots prove themselves inside their owner's own factory first.
Where a failure costs a bad shift, not a lawsuit.
And the brain, if Noetra delivers it, can run on hardware from anyone,
which is exactly the point.
So can Japan pull it off?
Go back to the number from the start.
22. If you judge the plan by headcount,
a 22-fold jump in 14 years against a 15-fold jump
in annual installations, I would not bet
on the number itself. But I would not bet against the plan,
because the number isn't
what Japan is really testing.
It is running the largest experiment any country has attempted.
Whether an aging society can engineer
its way out of demographic gravity,
whether machines can absorb the work
of 11 million missing people without tearing the social fabric,
whether a robot handing a cup of tea
to a 90-year-old can feel like care and not abandonment.
You don't have to wait until 2040 to find out. Mark the calendar.
By next March, Noetra's first physical AI model is due.
In early 2027, Mitsubishi's
Kyoto line should start producing
and Zeals should have its first 100 robots.
In 2028, the Haneda trial wraps up.
Around 2030, AREC, Kawasaki's humanoids,
and Shimizu's painting robots are all due,
and a robot horse is meant to gallop in Riyadh.
If those dates hold, the number starts to look less crazy.
If they slip, it was a slogan. And watch one more thing.
Japan says it will make the NOTRA model available
to domestic developers and businesses,
which would turn a national project into a platform
that every small robot maker in the
country can build on. And the stakes reach far beyond Japan.
South Korea, Italy, Germany,
and China are all sliding down the same demographic slope,
just a few years behind.
The country that ages first and solves it first writes the playbook
and collects the royalties.
If it fails, we learn that some kinds of human work can't be replaced,
and the rest of the world had better plan for that.
Japan didn't just shock the world with what it is building.
It showed us our own future, and it is arriving right on time.
So I want your verdict. Is 10 million a visionary target or a slogan?
And the question that sits underneath all of it,
would you let a robot care for your parents?
Tell me in the comments,
because Japan's answer is already rolling off the assembly line.
If you want more stories that follow the money
and the machines behind the next decade,
subscribe, and I will see you in the next one.
