A UBS research figure titled “Mapping out major humanoid robot participants” splits a humanoid robot into six layers, from the companies that build the whole machine to the makers of its hands and joints. The figure was shared on X on 24 October 2026; we transcribed it on 25 October 2026.
By our count, the figure lists 50 Chinese and 38 overseas names across the six layers, or 87 distinct companies, since Leadshine appears twice. Chinese names outnumber overseas ones in five of the six layers; computing is level at five each.
Business Pill · PHYSICAL AI
A short explainer of physical AI: AI systems that act in the physical world, such as robots. It teaches the general idea only and says nothing about the UBS figure.
The key insight: As we read it, the figure is a map of who is named, not who is winning: by our count Chinese names outnumber overseas ones in five of six layers, and computing is the one layer where the two sides are level.
The Six Layers of a Humanoid Robot
The figure labels perception the head, computing the brain, and battery and structure the torso; dexterous hands and actuators at the joints are the parts that move.
| Layer | China, as listed | Overseas, as listed | Count |
|---|---|---|---|
| The maker (OEM) | Unitree Robotics, AgiBot, UBTECH, Galbot, Dobot, Geek Plus, Estun (Codroid), Leju Robot, Fourier Intelligence, Stardust Intelligence, LimX Dynamics, Galaxea AI, EngineAI Robotics | Tesla, Boston Dynamics (Hyundai), Hexagon, Figure AI, Agility Robotics, Apptronik, 1X Technologies, Sanctuary AI, NEURA Robotics, PAL Robotics, Engineered Arts | 13 / 11 |
| Perception (head) | Hesai, RoboSense, Orbbec, Sunny Optical, OPT, Omnivision, DJI | Keyence, Cognex, Ouster, Luminar, Sony, RealSense | 7 / 6 |
| Computing (brain) | Horizon Robotics, Cambricon, Black Sesame, Rockchip, Huawei (Ascend) | NVIDIA, Qualcomm, AMD, Intel, Physical Intelligence | 5 / 5 |
| Battery and structure (torso) | CATL, EVE Energy, Sunwoda, Minth, Joyson, Lens Technology, WeLion | LG Energy Solution, Samsung SDI, Panasonic | 7 / 3 |
| Dexterous hands | Zhaowei, Moons’ Electric, Leadshine, LingXin Dexterous Hand, INSPIRE Robots, BrainCo | ROBOTIS, WONIK Holdings, Maxon, Shadow Robot, SCHUNK | 6 / 5 |
| Actuators (joints) | Inovance, Hengli Hydraulic, Tuopu, Sanhua, Kedali, Leadshine, Leaderdrive, Laifual, Guomao, Shuanghuan, Zhongda Leader, Kunwei | Hyundai Mobis, Nidec, Schaeffler, Harmonic Drive Systems, Nabtesco, Hiwin, THK, NSK | 12 / 8 |
Source: the UBS figure “Mapping out major humanoid robot participants”, as shared on X on 24 October 2026, our transcription. Counts are China / overseas and are our arithmetic; they count names on the figure, not market share.

What Was Posted on X
@StockMKTNewz lists some of the companies UBS names across humanoid robot makers, computing, sensors, batteries and joints, in a post that also promotes a sponsor’s webinar.
What Each Layer Does
The makers, or OEMs, design and assemble the finished robot. The overseas list includes Tesla, Boston Dynamics, which the figure ties to Hyundai, and Figure AI; the Chinese list starts with Unitree Robotics, AgiBot and UBTECH, in the figure’s order.
Perception is how the robot sees: cameras, depth sensors and lidar. The figure puts Hesai and RoboSense on the Chinese side and Keyence, Cognex and Sony overseas.
Computing is the brain that runs the robot’s AI models. The overseas list is NVIDIA, Qualcomm, AMD, Intel and Physical Intelligence; the Chinese list includes Horizon Robotics, Cambricon and Huawei’s Ascend.
The battery and structure layer powers and holds the body. CATL is listed first on the Chinese side; LG Energy Solution, Samsung SDI and Panasonic are the overseas names.
Hands and actuators are the moving parts. Actuators combine reducers, drivers and motors at each joint; the overseas list includes Harmonic Drive Systems, Nabtesco, Schaeffler and Nidec, and the Chinese list includes Inovance and Sanhua.
Where the Figure Sits in UBS’s Numbers
The figure does not say which UBS report it comes from or when that report was published. A UBS humanoid report from February, as summarised by Wallstreetcn on 12 February 2026, projected global demand of 30,000 humanoid robots in 2026 in its base case.
That summary also said UBS put 2025 global shipments at about 18,600 units, with Unitree and AgiBot each shipping about 5,000, and that UBS analyst Phyllis Wang expected real shipment volumes to be concentrated in 2027 to 2028.
Update (25 Oct): an investor’s view on adoption speed. Andrew Kang, chief executive of RoboStrategy, a closed-end fund that invests in robotics and embodied AI companies, argued in a 23 October essay that “There won’t be any ChatGPT moment for Robotics.” Kang writes that robots doing physical work must perform well “often >99.9% of the time” and that whoever buys a robot pays for it whether it is working or not, so adoption will come over years rather than weeks, with commercial work such as laundry, machine tending and material handling before homes.
The essay names UPS, BMW and Samsung among companies putting more resources into robot automation. Kang sits on the boards of Dyna Robotics and Standard Bots, according to RoboStrategy’s team page, and the essay discloses that the fund’s portfolio may include companies it discusses.
Update (26 Oct): a CATL-backed humanoid maker files for Hong Kong. Noetix Robotics, a Beijing humanoid robot maker backed by CATL, has filed confidentially for a Hong Kong initial public offering that could raise several hundred million dollars, Bloomberg reported on 22 October, as carried by The Standard. The listing could come as early as 2027, and the size and timing could change, Humanoids Daily reported from Bloomberg’s account. Noetix, officially Songyan Power Beijing Technology, raised 1 billion yuan in March in a round led by Chendao Capital, an investment platform affiliated with CATL, The Standard reported.
Bloomberg reported that Hong Kong IPOs have raised more than $50 billion this year, and that Unitree’s shares have fallen almost 40% from their peak since its August Shanghai IPO, according to Humanoids Daily. Noetix did not respond to Bloomberg’s request for comment. CATL appears in this piece’s UBS map in the battery and torso layer.
Connecting the Coverage
Musk put a run rate of 20,000 Optimus robots a week a year or more away, per Axios, as our Optimus piece of 21 October reported, while Tesla said Optimus production began in Fremont.
Unitree, the first maker on the Chinese list, was pricing a Shanghai IPO that implied a valuation floor near 42 billion yuan, as our Unitree piece of 6 August reported. On the computing layer, our Figure compute piece of 4 September covered a Nscale deal for up to roughly 100,000 NVIDIA Vera Rubin GPUs for Figure, the company behind the Helix robot model.
What is new here: one map that places 87 named companies across the parts of a humanoid robot, with China listed more often in five of six layers.
The Structural Read
As we read it, the widest gaps on the figure are in battery and structure, seven names to three, and in actuators, twelve to eight: four names each, by our count.
As we read it, the computing layer is where the overseas side holds level: NVIDIA, Qualcomm, AMD, Intel and Physical Intelligence against Horizon Robotics, Cambricon, Black Sesame, Rockchip and Huawei’s Ascend.
As we read it, the February numbers set the scale: Wallstreetcn’s summary put UBS’s 2026 base case at 30,000 robots, and said Unitree and AgiBot each shipped about 5,000 in 2025.
Title of the UBS research figure, as shared on X on 24 October 2026
“Mapping out major humanoid robot participants”
Three Implications
FOR MAKERS The figure lists 13 Chinese and 11 overseas OEMs, by our count, from Unitree Robotics to Tesla.
FOR PARTS SUPPLIERS By our count the figure names 12 Chinese and 8 overseas actuator makers, the largest parts layer on the map.
FOR CHIP MAKERS NVIDIA, Qualcomm, AMD and Intel are the overseas computing names on the figure, beside Physical Intelligence.
The Business Engineer Lens
This story maps onto the Business Engineer framework The AI Supercycle.
The framework puts it this way: “The supercycle runs on rare earths before it runs on tokens. Every GPU cooling fan, every data-center motor, every robotic actuator, every EUV lithography machine depends on neodymium-iron-boron magnets.”
As we read it, the UBS figure’s actuator layer is where that physical floor meets the robot: the joints. The framework does not discuss this figure.
What Is Not Established
We read the UBS figure only as an image shared on X; we did not read the UBS report it comes from and do not know its date or title. The crop does not include a legend, so we do not explain the asterisks or the grey names. Our counts are of names on the figure; they say nothing about each company’s revenue, volume or market share. The February forecast comes from a Wallstreetcn summary, not the UBS report itself. This is not investment advice.
The Bottom Line
A UBS figure maps 87 companies across six layers of a humanoid robot, from NVIDIA in the brain to CATL in the battery, and by our count lists Chinese names more often in five of the six.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. This piece rests on a UBS research figure as shared on X on 24 October 2026, which we transcribed and counted ourselves; Wallstreetcn’s 12 February 2026 summary of a UBS humanoid report; and our own earlier pieces, all read on 25 October 2026. We did not read the UBS report behind the figure and did not contact UBS or the companies named.
Sources: UBS figure as shared on X, 24 Oct 2026 · Wallstreetcn via Longbridge, 12 Feb 2026









