Skip to content
Botlero

The state of humanoid robots, 2026

Everything below is counted from the 31 humanoid robots documented on Botlero — 22 manufacturers, 1,967 published values resting on 7,925 sourced statements from 45 websites, each one carrying the page it came from, a verbatim quote and the date it was checked. These are counts of what makers and directories publish, not an estimate of the industry: 31 robots is not every humanoid robot in the world, and where a measurement is published by only part of the catalogue the section says so and gives the denominator.

There is no market size on this page, no growth rate and no forecast. Nothing in this database holds one: every such figure would come from a report we have not read, and what Botlero refuses to publish covers it. How this page is built is at the foot.

Prototype is the most common lifecycle stage: 10 of 31 robots

7 stages across 31 robots

A lifecycle stage is only shown when a source states it. Prototype leads with 10, and the other 17 robots with a stated stage are spread across 6 further stages: General availability 4, Preorder 4, R&D 4, Mass production 3, Announced 1 and Early access 1. 4 robots publish no lifecycle stage at all, which is why the bars add up to 27 rather than 31. The bars are read against the whole catalogue, so a bar's length is that stage's share of all 31 robots.

Robots by lifecycle stage
Prototype
10
General availability
4
Preorder
4
R&D
4
Mass production
3
Announced
1
Early access
1

China accounts for 20 of the 31 robots documented here

7 design countries

Counted on the country a source names as the robot's design origin, not its place of assembly — a separate measurement with thinner coverage. 20 of 31 are designed in China, more than every other country put together; the other 6 countries account for 11.

Robots by design country
China
20
United States
6
India
1
Japan
1
South Korea
1
Spain
1
United Kingdom
1

19 of 31 robots move on bipedal locomotion

legs or wheels

Locomotion is the one facet this site treats as a hard boundary: a compare table only ever puts two robots of the same class side by side, because a walking speed and a wheeled base are not the same measurement. The split is Bipedal 19 and Wheeled 12, and a robot whose locomotion no source states appears in neither bar.

Robots by locomotion type
Bipedal
19
Wheeled
12

Research is the most common use, stated for 17 of 31 robots

11 use classes

A robot may be stated for several uses, so these bars sum to more than 31. The count unions the application classes a source lists with the user type it states — the same union the use hubs and the browse filter read, so the three cannot disagree about who belongs where.

Robots by stated use
Research
17
Professional
16
Education
8
Logistics
6
Manufacturing
5
Hospitality
4
Household
3
Consumer
2
Healthcare
2
Retail
2
Entertainment
1

Published heights run from 107 cm to 210 cm

31 of 31 publish a height

31 robots publish a height, with a median of 166 cm — within a few centimetres of an adult. The span in the heading is the lowest figure published to the highest, and one robot publishes a range rather than a single figure, so its top end sets that upper limit. The bands below are 20 cm wide and place each robot at the first number its own page prints, so an adjustable robot sits at its lowest height; the highlighted band is the one the median falls in.

Published heights, in 20 cm bands
100–119 cm
3
120–139 cm
3
140–159 cm
3
160–179 cm
18
180–199 cm
4

29 of 31 robots have a published starting price, from $4,900 to $350,000

5 of 29 from the maker

A price is published for 29 of 31 robots, and only 5 of those figures appear on a maker's own page — the rest are reported by third-party directories, which is why every price on this site carries a chip saying who stated it and whether it is a list price, a report or an estimate. No currency is converted: a figure not published in US dollars is not published here.

Published prices that come from the maker's own page5 of 29 · 17%

Published starting prices run from $4,900 to $350,000.

The median robot publishes 64 of 195 specifications

1,967 of 6,045 possible

This is a measure of disclosure, not of the robots: it counts how many of the 195 specifications this site tracks have at least one sourced value. The best-documented robot publishes 104 and the thinnest 42; across the catalogue 1,967 of a possible 6,045 are filled, or 33%.

Robots by number of specifications published, in bands of 20
40–59
14
60–79
13
80–99
3
100–119
1

167 of the 195 specifications tracked are published by at least one robot

28 are published by nobody

The other way round: of the 195 measurements in this site's registry, 167 are published by at least one robot and 28 by none at all. The gap is vendor silence rather than a gap in the pipeline — the groups furthest down this list are the ones makers do not describe, and the bars show how many of each group's keys anybody publishes.

Specification groups by how many of their keys anybody publishes
Autonomy
6 of 6
Body
10 of 10
Commercial
5 of 5
Compute
7 of 7
Connectivity
6 of 6
Environment
6 of 6
Home
2 of 2
Joints
19 of 19
Load
5 of 5
Mobility
6 of 6
Organization
4 of 4
Overview
13 of 14
Perception
13 of 14
Hands
12 of 13
Power
18 of 20
Software
12 of 15
Safety
5 of 7
Service
7 of 10
Privacy
5 of 9
Security
3 of 6
Compliance
3 of 11

18% of the 7,925 sourced statements behind the catalogue come from a maker's own page

the rest are third-party

Of 7,925 sourced statements, 1,436 were read from a page belonging to the robot's own maker and the rest from third-party directories. Where a maker and a directory disagree the maker's figure is the one this site publishes, and the other is kept and shown beside it rather than discarded.

Sourced statements read from a maker's own page1,436 of 7,925 · 18%

Not one of the 7,925 sourced statements has been independently measured

nothing here is a Botlero measurement

Botlero measures nothing itself and has no laboratory: a value here is proof that a page said it, on a page about that robot, not proof that it is true. No value in the catalogue has been independently measured or filed with a regulator — every number is a claim by a maker or by a directory, and the chip beside it says which.

Sourced statements independently measured or filed with a regulator0 of 7,925 · 0%

The 7,925 sourced statements come from 45 websites

top 10 of 45

Every value is traceable to the page it was read from, so the catalogue can be counted by source. 45 distinct websites are cited across 7,925 statements; the ten most-cited are below. A maker's own domain and a third-party directory are both ordinary rows here — what separates them on a robot's page is the chip, not the count.

The ten most-cited source websites, by number of sourced statements
roboatlas.ai
1,021
aparobot.com
648
robotoverflow.com
628
originofbots.com
608
humanoid.press
466
humanoid.guide
463
unitree.com
283
guijicankao.com
273
humanoid-robots.io
271
robozaps.com
266

Sources disagree on 580 of the 1,967 values this site publishes

29% of published values

This is the number no other directory in this category can publish: on 580 of 1967 values, this site holds at least one source stating a different figure from the one it shows — a value being one robot's figure for one measurement, with 7,925 statements standing behind the 1,967 of them. Nothing is overwritten — both are kept, the stronger claim is shown, and clicking a value lists every source with its quote. The measurements that disagree most are below.

Measurements on which sources disagree most often, by number of robots affected
Applications
30
Sensors
30
Degrees of freedom
23
Announced
22
Compute
22
Height
22
Stage
22
User type
20

683 published values have been withdrawn since 2026-09

1 month of history

A withdrawn value is one this site stopped publishing, kept with the reason recorded at the time rather than deleted. 683 have been withdrawn across 1 month; every one is listed, with its reason, on corrections.

Values withdrawn per month
2026-09
683

How this page is built

method last, and on the page

N = 31 robots, which is the whole of Botlero's catalogue and not the whole industry. Every count above is computed from this site's own published values when the page is rendered — there is no stored statistic, no hand-entered total and no figure taken from a third party. If a value changes on a robot's page, the number in the heading above changes with it on the next rebuild.

One value per robot per measurement. Where several sources state a measurement, the figure counted here is the one the robot's own page shows: the maker's own where a maker publishes one, otherwise the figure the most sources state. That is the same rule the robot page, the comparison table and the records page all read, so a number counted here is a number you can click through to.

Robots with no value are excluded, and the exclusion is stated. A section that covers part of the catalogue prints its own denominator — 31 of 31 publish a height, 29 a price — rather than treating a missing figure as a zero. Nothing is interpolated, averaged across sources or converted between currencies.

The charts are drawn on the server, from the same numbers. Each one is plain SVG or an HTML bar in the page itself: no chart library loads, no chart fetches data, and a reader without JavaScript sees every figure. Each chart carries its values as text and each row has its own anchor, so a single datum can be linked rather than the page. Length is the only thing a bar encodes; the accent colour marks the datum the heading is about and means nothing else.

What is not here. No market size, growth rate, forecast, share-of-market or valuation aggregate: this database stores none of them, and a number sourced from a report nobody here has read is exactly what the rest of the site refuses to publish. No score, no ranking of makers and no verdict of our own — records ranks published values and says so, and about & method states what a value on this site does and does not prove.