One month ago,
Chinese humanoid robot company Unitree went public in Shanghai.
On its first day, the stock price surged by over:
5 times.
At that time, the market bet on humanoid robots as potentially the next:
smartphone,
new energy vehicle,
or even a major new AI industry.
But just a month later,
the scenario suddenly changed.
Reuters reported today that
Chinese regulators are slowing down IPOs of some humanoid robot companies.
And Unitree’s stock price
has fallen by about
55% from its highest point since listing.
The key question is not how much the stock has fallen,
but rather the regulatory focus on a question the humanoid robot industry will inevitably have to answer:
Are these companies' revenues really driven by genuine market demand for robots?
To be clear: China has not banned humanoid robot listings
Reuters quotes insiders saying
Chinese regulators have recently employed
window guidance,
an informal advisory mechanism,
to slow down IPO processes for some humanoid robot companies.
One source described the humanoid robot IPO situation as essentially
paused.
Another source emphasized that this is not a formal ban,
but rather
an increase in scrutiny on this sector.
The China Securities Regulatory Commission has not responded to Reuters’ inquiries.
So it’s inaccurate to say
“China has completely banned humanoid robot IPOs.”
The more accurate interpretation is that
IPO standards are being raised.
Why is Unitree a turning point?
Unitree provides a direct example.
This company produces
humanoid robots
and quadruped robot dogs.
When it went public in August,
its stock price jumped over fivefold on the first day.
SasaDaily previously wrote that
the market was betting not so much on Unitree’s current earnings,
but rather on
whether China could be the first to mass-produce humanoid robots.
But Reuters notes today that
Unitree’s stock has now fallen about
55% from its peak.
Such volatility makes regulators worry even more whether
the market is moving too far ahead of real commercialization.
At least half a dozen more companies still aiming for IPO
Reuters reports that
around
six
Chinese humanoid robot companies are still preparing IPOs.
These include
Deep Robotics,
X Square Robot,
and AGIBOT.
That means after Unitree’s listing,
the question becomes
whether their revenue quality can stand the scrutiny.
This is precisely what regulators are now investigating.
The sensitive issue is not technology, but "where the revenue comes from"
Humanoid robot companies can generate revenue.
But
not all revenue necessarily reflects genuine market demand.
Reuters cites sources close to the humanoid robot investment market saying regulators are especially concerned about two sources of income:
robot data-collection centers
and
joint ventures supported by local governments.
What is a data-collection center?
It can be understood as
a large site where many robots repeatedly
walk,
grab objects,
carry items,
and operate equipment,
allowing companies to collect real physical data,
which they then use to train
embodied AI.
This activity certainly has value.
The issue is
whether customers are truly paying for robots to work, or if government-backed projects are artificially generating industry data.
These two types of income have completely different implications for valuation.
Some joint ventures see local government covering 80–90% of initial investment
Reuters reports that
in some joint ventures,
local governments may provide
80–90% of the initial investment.
These projects generate
orders,
revenues,
and training data,
which can also help companies reach the scale needed for listing.
But regulators are now asking
if this government support were removed, would customers still pay for the robots themselves?
This is the real commercial demand.
Some estimates suggest valuations may drop 60–70% if these revenues are excluded
Reuters cites a source close to humanoid robot investors estimating that
if revenues related to
data-collection centers
are excluded from valuation models,
valuations for some robot companies
could drop by 60–70%.
Note that
this is not an official figure from Chinese regulators,
nor does it imply the whole industry will shrink by this amount.
It is one market participant's estimate.
But it highlights the clear issue:
if a company’s valuation relies on “future large-scale commercial use,” the market now demands proof that this future is becoming reality.
Coinciding with this, the first global humanoid robot data offers an answer
Interestingly, today the
International Federation of Robotics (IFR)
released the first global humanoid robot market statistics.
In 2025,
the actual number of humanoid robots sold worldwide was about
7,000 units.
This number covers
industrial
and
professional service
applications,
but excludes
consumer,
military,
and medical robots (the latter classified separately).
7,000 sounds like a lot,
but placed in the context of the entire robot market,
it shows how small humanoid robot deployment still is.
Traditional industrial robots number 542,000 annually
For comparison,
in 2024, global installations of traditional industrial robots were about
542,000 units.
Professional service robots sold about
199,000 units.
Humanoid robots account for roughly
7,000 units.
In other words,
media hype might lead people to think
"factories will soon be full of humanoid robots,"
but actual deployments remain a very small segment within the robot industry.
And not all of those 7,000 units are truly "on the job"
IFR Secretary General Susanne Bieller told Reuters that
many of the humanoid robots sold in 2025 were not
actively conducting daily production work in factories.
Many were purchased by
research institutions,
AI companies,
and corporate R&D departments
for
research,
testing,
and
collecting AI training data.
This aligns exactly with the issues Chinese regulators are now investigating.
The real question is no longer
"Are customers buying robots?"
but rather
"What are the robots being used for?"
Car factories mostly still have only a few test units
The automotive sector has long been considered the most likely early large-scale adopter of humanoid robots,
due to its
highly standardized factories,
many repetitive tasks, and
existing familiarity with robot automation.
However, IFR notes that current humanoid robot pilots in some automotive companies
often involve only
single-digit
or
teens
in unit counts,
which is far from the scale of hundreds or thousands of machines in a single factory.
This explains why "running and jumping" alone isn’t enough
In recent months, the most eye-catching scenes of Chinese humanoid robots have included
running,
boxing,
dancing,
and completing obstacle courses.
These skills are important because they represent advances in
balance,
motion control,
vision, and
AI planning.
But factory managers’ real questions are not
"Can it throw a punch?"
but rather
How many hours can it work in a day?
How often does it break down?
What is the failure rate per pick?
Does switching tasks require an engineer to rewrite software for hours?
Compared to traditional robotic arms,
which option is more cost-effective over a year at a workstation?
These numbers determine
genuine orders.
China’s government has not changed its strategic direction on humanoid robots
The stricter IPO review does not mean
China is backing away from humanoid robots.
On the contrary,
China still regards
embodied intelligence,
the AI that can perceive the real world and act within a physical environment,
as an important emerging industry.
The government, local authorities, and private capital continue to invest heavily.
What is happening is more like
technology development continues, while the capital markets are demanding evidence.
Both can occur simultaneously.
From "everyone’s hype" to "selective belief"
S&P Global Market Intelligence analyst Ruiying Zhao used a precise phrase.
The investment mood for humanoid robots is shifting from
blanket euphoria
to
selective rationality.
Simply put,
it used to be
"As long as you build humanoid robots, I'll invest."
Now it’s become
How many units have you sold?
Who are your customers?
Are there repeat orders?
How long have the deployments lasted?
What tasks are they performing?
Is the revenue tied to government projects?
Are there sustainable gross margins?
This is a natural stage in the industry’s maturation.
Some private market valuations have already dropped 30–50%
Reuters quotes investors saying this humanoid/embodied AI boom
has been even stronger than China’s previous
internet
and
new energy vehicle
investment booms.
Some founders attracted dozens of prospective investors in a short period,
even refusing traditional due diligence.
But as the market cools,
valuations for some private market projects
are reported to have dropped
30–50%.
This reflects capital reexamining
"Are you selling technology or a story?"
But this doesn’t mean humanoid robots have no market
Having only 7,000 units now
does not necessarily mean
“Humanoid robots are a bubble.”
Because the industry is still very early stage.
Bank of America Global Research estimates
global humanoid shipments could reach
90,000 units in 2026,
and
1.2 million units by 2030.
But these are forecasts,
not confirmed sales.
So the real focus should be
not which forecast is largest,
but how many units are actually delivered each year, and how many remain working consistently.
Humanoid robot KPIs may soon be very unglamorous
To assess the commercial value of a humanoid robot company,
the real key indicators might not be
demo video views,
running speed,
or dancing smoothness,
but rather:
deployment count—how many units are actually deployed,
utilization—how many hours a day they work,
task success rate,
downtime—how often they break down,
recurring orders—whether customers buy again,
and
independent commercial customers—how many orders are not government-supported projects.
These metrics might be far less flashy than robots doing backflips,
but are closer to
real business.
The contrast between Unitree a month ago and today is complete
On August 19,
Unitree’s first day closing price soared
460%.
The real story then was
capital markets pricing in the future potential of China's humanoid robots.
On September 21,
the story is
Unitree’s stock has fallen 55% from its peak,
regulators are slowing other humanoid IPOs,
and revenue quality is under scrutiny.
On the same day,
the first global sales statistics showed:
about 7,000 humanoid robots actually sold in 2025.
Putting these three facts together,
the humanoid robot industry is entering its next phase:
moving from believing in the future to verifying the present.
This could signify the real start of industry maturity
A new tech industry’s earliest stage typically focuses more on:
demos,
concepts,
funding, and
valuation.
Next comes competition based on:
customers,
orders,
cost,
reliability, and
ROI.
The companies that survive
won’t necessarily be the best performers in robot demonstrations,
but rather
those that can prove customers are willing to pay repeatedly to have the robot perform the same job every day.
So today’s real signal is not
“China is losing faith in humanoid robots.”
but rather
“Humanoid robots are finally being held to real business standards.”
Today, grow a little with AI.
Learn one AI skill each day.
Save a bit of time daily.
Improve a bit every day.
SasaDaily, growing with you.
