BotBeat rated 6 Chinese humanoid makers. The scores tell you who is real vs vapor:
Unitree 21/25: real revenue ($248M), real margin (60%), 5,500+ units shipped
UBTech 18/25: public company, 1,079 units, 54% humanoid margin
Galbot 15/25: $800M raised, zero revenue disclosed. /t.co/JtpkpvHHqK
Unitree 21/25: real revenue ($248M), real margin (60%), 5,500+ units shipped
UBTech 18/25: public company, 1,079 units, 54% humanoid margin
Galbot 15/25: $800M raised, zero revenue disclosed. /t.co/JtpkpvHHqK
0
0
1
37
China cut humanoid joint costs 65% in 2 years.
Unitree: $1,800/actuator in 2023. $620 today.
That collapse is the whole story. At $620/joint, a 20-DOF humanoid costs $12,400 in actuators alone. At $1,800 that same robot costs $36,000.
Who is building comparable actuator cost tps://t.co/JtpkpvHHqK
Unitree: $1,800/actuator in 2023. $620 today.
That collapse is the whole story. At $620/joint, a 20-DOF humanoid costs $12,400 in actuators alone. At $1,800 that same robot costs $36,000.
Who is building comparable actuator cost tps://t.co/JtpkpvHHqK
0
1
1
41
6 Chinese humanoid makers. 1 has real revenue. 3 have real units. 2 are vaporware.
Find out which is which here: open.substack.com/pub/botbeat247…
Find out which is which here: open.substack.com/pub/botbeat247…
0
0
0
18
Vision-only is a bet that data beats sensors. Tesla made it work for cars. Asian labs are watching closely. If Foundation Robotics ships at scale, every lidar-heavy humanoid roadmap in Korea and Japan gets a hard rethink.
Foundation Robotics founder and CEO Sankaet Pathak says their humanoid relies solely on cameras; adding other sensors introduces noise and complexity that far outweigh the benefits.
0
0
1
48
Open-source is how Asia closes the gap. Japan and Korea have KAIST and AIST publishing full stacks. China is now doing it commercially. ROBOTO ORIGIN going full open-loop means Western labs lose the replication monopoly faster than they think.
Let’s make humanoids reproducible. 🤖🛠️
Humanoid robotics doesn’t need more hype—it needs baselines that people can actually rebuild. RoboParty has officially fully open-sourced the entire engineering loop for ROBOTO ORIGIN, from sourcing and assembly to calibration and motion -grade biped went from zero to a 3 m/s (6.7 mph) run in just 120 days. It’s a complete, verifiable reference stack designed to stop the industry from constantly reinventing the wheel and start building on shared infrastructure.
The team is calling for "co-builders" to fork the project, document failure modes, and share benchmarks. It’s time to move past cool stunts and focus on real-world utility through radical transparency.
GitHub: github.com/Roboparty/robo… Docs: roboparty.com/roboto_origin/…
#Humanoid #OpenSource #Robotics #BuildInPublic #EmbodiedAI #RoboParty #LeggedLocomotion #ROBOTO_ORIGIN
0
0
0
77
AgiBot crossed 10K total units in March. Halfway there (5,100) came from just one quarter. The acceleration curve is steeper than most supply chain analysts modeled. botbeat247.substack.com
Interestingly, global humanoid robot shipments reached nearly 13,000 units in 2025 (Omdia report), with AGIBOT delivering more than 5,100 units,ranking first worldwide.
But beyond rankings, what truly matters is that thousands of real robots have been deployed in commercial generated high-quality data—especially data tied to real business workflows and cross-task coordination,something that around 90% of companies may find difficult to achieve.
Within the industry, there is a widely shared view on data quality:
real-world environments > teleoperation + motion-capture data > simulation data > human video data.（Become stronger by working with customers, not in the laboratory.）
This means that AGIBOT’s thousands of humanoid robots are actively helping it build a real-world flywheel that runs from commercial scenarios to product iteration, and then back into models and data.
In other words, AGIBOT is trying to establish an early advantage in commercial humanoid robot applications—moving ahead of companies such as Tesla, Figure, and Boston Dynamics. Examples include the deployment of Genie 2 in factory warehouses at Joyson Electronics, Fulin Precision, and Damon Group, as well as the X2 appearing across major tourist attractions and stage performances in China.
To be honest, this is only the beginning. Tesla’s Optimus V3 is expected to launch in Q1, alongside preparations for million-unit production lines; Boston Dynamics’ new Atlas is entering mass production this year; and UBTECH already holds more than RMB 1.4 billion in orders, with shipments projected to reach the ten-thousand-unit scale this year. All of these companies are, to varying degrees, entering the real world,but truly representative, large-scale use cases remain rare.
We are still in the early stages of humanoid robotics, but the changes coming in 2026 will be significant.
Are U ready?
0
0
0
22
Chinese humanoids debuted at Korea's biggest industrial fair. AW 2026 Seoul: 2,300 booths, record 36-year attendance. AgiBot X2/G2, Unitree G1, Leju Kuavo 4th Gen all exhibiting. China's humanoid makers are pitching Korean manufacturers in their own backyard. Are Korean firms buying? botbeat247.substack.com
0
0
0
29
AgiBot hit 10,000 humanoid robots on March 30. The 5K-to-10K jump took 3 months -- 4x faster than the prior phase. 39% global market share in 2025. It took 2 years to reach the first 1,000 units. The production curve is now exponential. What breaks the acceleration? botbeat247.substack.com
0
0
0
16
Chinese humanoids debuted at Korea's biggest industrial fair. AW 2026 Seoul: 2,300 booths, record 36-year attendance. AgiBot X2/G2, Unitree G1, Leju Kuavo 4th Gen all exhibiting. China's humanoid makers are pitching Korean manufacturers in their own backyard. Are Korean firms buying? botbeat247.substack.com
0
0
0
27
73.6% of Unitree revenue from research/education is the key risk buried in this filing. Most commercial success is universities and labs buying for R&D, not factories deploying for production. The real commercial ramp starts 2026. Watch the customer mix shift. x.com/Charles_T_Sun/…
The prospectus confirms 73.6% of humanoid robot revenue came from “research and education” customers. At least five of those universities sit on U.S. or Canadian export control lists. One, North University of China, historically known as the “First School for People’s Ordnance,” described its Unitree purchase as intended for “weapon science and technology” applications.
0
0
0
15
LimX Luna at fashion show is brand play. The real story: Shenzhen's commodity actuator ecosystem ($400-$800/unit motors) enables 140+ humanoid manufacturers. Western humanoids can't source at that price.
0
0
0
15
The data gets starker with the forecast. Morgan Stanley doubled its China humanoid estimate to 28,000 units for 2026 in January. That is not a rounding error. That is a market structure. Meanwhile Western companies are still pre-revenue. x.com/IntelligenceHQ…
Humanoid Robots by Actual Sales
→ China dominating in early innings
Unitree’s R1 robot ($5,900 each) leads with 5,500!
But Tesla’s Optimus is warming up…
Expected unveil of Optimus V3 is expected in the next few months.
The first model meant for mass production.
A slow this year.
1M Optimi/year is the target capacity for their Fremont factory in 2027
But the price is expected to be higher: $20k-$25k each
Have you seen any humanoids in person yet?
1
0
3
46
Atlas in production era while UBTech has 1,079 humanoids already deployed and revenue-generating. The gap is not capability. It is deployment velocity and unit economics. Hyundai scale-up is the right move. China still has a 2-year head start in the field. x.com/JarsyInc/statu…
⚡️Robotics Brief⚡️Boston Dynamics Atlas Enters Production Era with Hyundai Scale-Up
The production version of Boston Dynamics' humanoid Atlas robot is ramping into 2026 with major manufacturing partnerships. Hyundai has secured orders for the platform and is targeting 30,000 units annually by 2028, marking the transition from robotics research to mass production.
The collaboration includes technical partnerships with Google DeepMind to advance AI capabilities on the Atlas platform. This ecosystem approach combines Boston Dynamics' hardware expertise with world-class AI research, accelerating the timeline for practical humanoid deployment.
Why it matters: Humanoid robots at scale represent a generational shift in industrial automation. Reaching 30,000 units yearly by 2028 suggests mainstream adoption timelines are compressing faster than most forecasts predicted.
Source: Automate.org
@BostonDynamics
0
0
0
26
Hyundai just made the largest single corporate robotics bet in history.
$86.7B through 2030. 71% earmarked for robotics, AI, hydrogen. Target: 30,000 robots/year by 2028.
An automaker betting its entire future on not being one.
botbeat247.substack.com
$86.7B through 2030. 71% earmarked for robotics, AI, hydrogen. Target: 30,000 robots/year by 2028.
An automaker betting its entire future on not being one.
botbeat247.substack.com
0
0
0
5
China shipped 11,747 humanoid robots commercially in 2025. The entire Western humanoid industry shipped fewer than 20.
Morgan Stanley just doubled its 2026 China forecast to 28,000 units.
The scale gap is not narrowing. Its accelerating.
Full breakdown: botbeat247.substack.com
Morgan Stanley just doubled its 2026 China forecast to 28,000 units.
The scale gap is not narrowing. Its accelerating.
Full breakdown: botbeat247.substack.com
0
0
0
13
Galbot just became China's highest-valued unlisted humanoid company.
$362M raised in March 2026. $570M total. Valuation: ~$3B.
Backed by China's state "Big Fund" + Sinopec + CITIC. Founded May 2023 (under 3 years old).
Orders from CATL, Bosch, Toyota, BAIC, SAIC Motor.
China s critical infrastructure. When state oil companies and state banks invest in a robotics startup, the signal is unmistakable.
Full breakdown Friday: botbeat247.substack.com
Is $3B fair for a sub-3-year-old humanoid company?
$362M raised in March 2026. $570M total. Valuation: ~$3B.
Backed by China's state "Big Fund" + Sinopec + CITIC. Founded May 2023 (under 3 years old).
Orders from CATL, Bosch, Toyota, BAIC, SAIC Motor.
China s critical infrastructure. When state oil companies and state banks invest in a robotics startup, the signal is unmistakable.
Full breakdown Friday: botbeat247.substack.com
Is $3B fair for a sub-3-year-old humanoid company?
0
0
0
46
This Week in Robotics:
UBTech shipped 1,079 humanoids in 2025. Revenue up 2,204%.
Unitree told Korea: cooperate on our terms or get left behind.
A Chinese factory builds one humanoid every 30 min.
Korea launched a 250-company national AI alliance.
Thread:
UBTech shipped 1,079 humanoids in 2025. Revenue up 2,204%.
Unitree told Korea: cooperate on our terms or get left behind.
A Chinese factory builds one humanoid every 30 min.
Korea launched a 250-company national AI alliance.
Thread:
1
0
0
55
Shipment scoreboard:
China humanoid shipments (2025 est): ~11,700+
USA humanoid shipments (2025 est): <20
This is not a competition. It is a rout.
The gap is not closing. It is accelerating.
China humanoid shipments (2025 est): ~11,700+
USA humanoid shipments (2025 est): <20
This is not a competition. It is a rout.
The gap is not closing. It is accelerating.
1
0
0
15
Full decoded analysis with charts and sourced data in this weeks free newsletter:
botbeat247.substack.com/p/this-week-in…
What story are you watching closest?
botbeat247.substack.com/p/this-week-in…
What story are you watching closest?
0
0
0
5
UBTech just proved Chinese humanoids are a real business.
3 units in 2024. 1,079 in 2025. 360x increase.
Humanoid revenue: $112.5M, up 2,204%.
Gross margin expanded 9 points while scaling.
Boston Dynamics has never had a profitable year.
3 units in 2024. 1,079 in 2025. 360x increase.
Humanoid revenue: $112.5M, up 2,204%.
Gross margin expanded 9 points while scaling.
Boston Dynamics has never had a profitable year.
2
0
0
24
Unitree pitched "cooperation" in Seoul. The decoded message: buy our robots or fall behind.
G1 price in Korea: $13,500.
Korean alternatives: $100,000+.
A 7-8x cost gap no industrial policy can close fast enough.
Korea faces the humanoid version of the EV dilemma it already lost.
G1 price in Korea: $13,500.
Korean alternatives: $100,000+.
A 7-8x cost gap no industrial policy can close fast enough.
Korea faces the humanoid version of the EV dilemma it already lost.
0
0
0
12
