BusinessChina

Dobot's embodied AI robot sales jump fivefold to RMB20 million as its cobot business pays for the humanoid push

Published 5 min read
Source
Dobot's embodied AI robot sales jump fivefold to RMB20 million as its cobot business pays for the humanoid push (Illustration by MEKAWIRE)

Hong Kong-listed cobot maker Dobot grew 2025 revenue 31.7% to RMB492.2 million, with embodied AI robots rising to RMB20.0 million from RMB3.9 million. Its net loss narrowed to RMB84.0 million, but the adjusted loss widened to RMB49.6 million as 39.3% of research spending went to embodied AI.

Shenzhen Dobot, the Hong Kong-listed maker of collaborative robot arms, said revenue from its embodied AI robots rose to RMB20.0 million in 2025 from RMB3.9 million a year earlier, according to the annual report it filed with the Hong Kong stock exchange on 30 March 2026. Group revenue grew 31.7% to RMB492.2 million and the net loss narrowed to RMB84.0 million from RMB95.4 million.

The report shows a company still financed by its arm business while it spends heavily on humanoids. Gross profit rose to RMB226.7 million, and gross margin slipped to 46.1% from 46.6%. Research and development expenses climbed 59.7% to RMB114.7 million, and Dobot said RMB45.1 million of that, or 39.3%, went to embodied AI. On its adjusted, non-IFRS measure of loss attributable to shareholders, the loss widened to RMB49.6 million from RMB36.5 million.

Where the money came from

Six-axis collaborative robots remained the engine. Sales of those arms rose 44.7% to RMB302.2 million and made up 61.4% of revenue. Integrated cobots, which pair an arm with other hardware, grew 27.3% to RMB67.6 million. Sales of four-axis desktop arms, the products Dobot built its name on in classrooms and labs, fell 3.0% to RMB93.0 million. Embodied AI robots contributed 4.1% of revenue.

Dobot revenue by product line (RMB million)
  • Six-axis cobots
    2025
    302.2
    2024
    208.9
    Change %
    44.7
    Share of 2025 revenue %
    61.4
  • Four-axis cobots
    2025
    93
    2024
    95.8
    Change %
    -3
    Share of 2025 revenue %
    18.9
  • Integrated cobots
    2025
    67.6
    2024
    53.1
    Change %
    27.3
    Share of 2025 revenue %
    13.7
  • Embodied AI robots
    2025
    20
    2024
    3.9
    Change %
    418.8
    Share of 2025 revenue %
    4.1
  • Others
    2025
    9.4
    2024
    12
    Change %
    -21.7
    Share of 2025 revenue %
    1.9
  • Total
    2025
    492.2
    2024
    373.7
    Change %
    31.7
    Share of 2025 revenue %
    100

From Dobot's 2025 annual report. Change % for Others derived as 2025 / 2024, minus 1; all other percentages as disclosed.

As of Oct 1, 2026

Dobot described 2025 as its first year of mass production and commercialisation for embodied AI. Its range includes the ATOM humanoid and a wheeled version, ATOM-W, sold under what the company calls a "one brain, multiple bodies" strategy, in which one software stack drives arms, humanoids and legged robots. As a showcase deployment, the report cites a humanoid making popcorn at a K11 site at a rate of 1,000 cups a day. Dobot also said cumulative global shipments of its robots passed 100,000 units and that it has served more than 80 Fortune Global 500 companies. The chairman's statement says Dobot's 2025 shipments ranked first worldwide; the report does not name a third-party source for that ranking.

Company background

Dobot was founded in 2015 in Shenzhen and listed on the Main Board of the Hong Kong exchange in December 2024. It then sold new H shares in placings in July and November 2025, which the annual report says added RMB1,705.9 million to share capital and premium. Shareholders' equity rose to RMB2,622.6 million at the end of 2025 from RMB967.3 million a year earlier. The company had 768 employees at 31 December 2025.

How it compares

The filing landed within the same 48 hours as results from three other Hong Kong-listed Chinese robot makers. UBTech Robotics reported revenue of RMB2,001.0 million, of which RMB820.6 million came from 1,079 full-size humanoid robots, and a loss for the year of RMB789.8 million. Estun Automation, the industrial robot maker, posted revenue of RMB4,888.01 million and a return to profit, with RMB41.34 million attributable to shareholders. Warehouse robot maker Geekplus grew revenue 31.6% to RMB3,171.0 million and cut its loss for the year to RMB10.4 million.

Dobot vs Hong Kong-listed Chinese robot peers, fiscal 2025 (RMB million)
  • Dobot
    Stock code
    2432.HK
    Revenue 2025
    492.2
    Revenue 2024
    373.7
    Revenue growth %
    31.7
    Gross margin %
    46.1
    Net result 2025
    -84
    R&D expenses 2025
    114.7
    R&D / revenue %
    23.3
  • UBTech
    Stock code
    9880.HK
    Revenue 2025
    2001
    Revenue 2024
    1305.4
    Revenue growth %
    53.3
    Gross margin %
    37.7
    Net result 2025
    -789.8
    R&D expenses 2025
    507.5
    R&D / revenue %
    25.4
  • Estun Automation
    Stock code
    2715.HK
    Revenue 2025
    4888
    Revenue 2024
    4008.8
    Revenue growth %
    21.9
    Gross margin %
    28.95
    Net result 2025
    41.3
    R&D expenses 2025
    418.7
    R&D / revenue %
    8.6
  • Geekplus
    Stock code
    2590.HK
    Revenue 2025
    3171
    Revenue 2024
    2409
    Revenue growth %
    31.6
    Gross margin %
    35.5
    Net result 2025
    -10.4
    R&D expenses 2025
    335.1
    R&D / revenue %
    10.6

Figures as disclosed in each company's fiscal 2025 annual report or results announcement filed on 30 or 31 March 2026, not estimated. Net result is loss for the year, except Estun, where it is profit attributable to equity shareholders. Geekplus gross margin = gross profit / revenue. R&D / revenue = R&D expenses / revenue. Growth = 2025 revenue / 2024 revenue, minus 1.

As of Oct 1, 2026

Dobot is the smallest of the four by revenue but has the highest gross margin, 46.1%, against 37.7% at UBTech, 35.5% at Geekplus and 28.95% at Estun. Its R&D intensity of 23.3% of revenue is close to UBTech's 25.4% and well above Estun and Geekplus. The difference is where that spending already pays. UBTech booked 41% of its revenue from full-size humanoids; Dobot booked 4.1% from its embodied AI line.

The technology in plain terms

A collaborative robot is an arm built to work next to people without a safety cage, usually with force limits and sensors that stop it on contact. Six-axis arms can reach and orient a tool much like a human arm, which suits assembly, inspection and machine tending. Four-axis arms are simpler and cheaper and work mostly on a flat plane. Embodied AI refers to robots whose perception and movement are driven by learned AI models rather than hand-written programs. Dobot's bet is that the same model and data pipeline can serve its arms and its humanoids.

What it means

For buyers in Europe, Japan and Korea, the numbers show a Chinese cobot maker with gross margins in the mid-40s that is reinvesting much of that margin in robot intelligence. That puts price and feature pressure on Universal Robots, FANUC's CRX line, Doosan Robotics and Rainbow Robotics in the arm market, where Dobot's six-axis sales are growing fastest.

Risks and open questions

Selling and distribution expenses rose 32.1% to RMB182.3 million, faster than revenue, while administrative expenses fell to RMB73.2 million from RMB88.8 million. The widening adjusted loss means the narrower headline loss leaned on items outside core operations. Dobot did not disclose how many embodied AI robots it sold, so the RMB20.0 million cannot be split into units and price. The four-axis decline also leaves the education business exposed.

MEKAWIRE analysis

Dobot's humanoid push is being paid for by a profitable cobot franchise, and 2026 will show whether embodied AI can grow faster than the cost of building it.

The evidence is in the split. Six-axis cobots added RMB93.3 million of revenue in 2025, more than the whole embodied AI line, while Dobot spent RMB45.1 million on embodied AI research to earn RMB20.0 million of embodied AI sales, a ratio of about 2.3 to 1. Equity of RMB2,622.6 million gives it room to keep that ratio for several years.

The strongest counter-argument is that UBTech has already shown a Chinese company can turn humanoids into a large revenue line, selling 1,079 full-size units in 2025. Dobot entered later, and in humanoids the scale of deployment matters more than margin on arms.

Bull case. Dobot's installed base of more than 100,000 robots and its customer list give it data and sales channels that humanoid start-ups lack. If one model can run arms and humanoids, every cobot sale feeds the humanoid business, and embodied AI could reach a double-digit share of revenue in 2026.

Bear case. Rivals with deeper pockets, including UBTech and Unitree, are raising money on public markets. Dobot could end up carrying rising research and selling costs while its embodied AI line stays small, and price competition in cobots could erode the 46.1% margin that funds the plan.

Signals to watch:

  • By 31 August 2026, Dobot's interim results, due within two months of the half-year end under Hong Kong rules, and whether they give embodied AI unit numbers.
  • Unitree's STAR Market listing process after its 20 March 2026 filing, which will set a public valuation benchmark for Chinese humanoid makers.
  • By 31 March 2027, the 2026 annual report, and whether embodied AI passes 10% of revenue.

Good or bad news for the company?

 

Key facts
Filing
Annual report, HKEX 2432, 30 March 2026
Net loss
RMB84.0 million (2024: RMB95.4 million)
Employees
768 at 31 December 2025
Gross margin
46.1% (2024: 46.6%)
R&D expenses
RMB114.7 million (+59.7%), RMB45.1 million on embodied AI
Revenue 2025
RMB492.2 million (+31.7%)
Embodied AI robot revenue
RMB20.0 million (2024: RMB3.9 million)
Cumulative robot shipments
More than 100,000 units
Adjusted net loss (non-IFRS)
RMB49.6 million (2024: RMB36.5 million)
Sources
MEKAWIRE Intelligence
  1. 01

    Why it matters

    Dobot is one of the few listed Chinese companies moving into humanoids from an established robot business with high gross margins, so its accounts give an unusually clean read on what the transition costs. The 2025 report puts a number on it. Embodied AI took 39.3% of a research budget that grew 59.7%, while the same line earned 4.1% of revenue.

    The filing also arrived alongside UBTech, Estun and Geekplus, giving investors four audited data points on Chinese robotics in one week. Dobot stands out for margin, not scale, which makes it a test of whether margin discipline or sheer volume wins the humanoid market.

  2. 02

    Rival analysis

    In humanoids, UBTech is the reference point, with RMB820.6 million from 1,079 full-size units in 2025. Dobot does not disclose units, which makes direct comparison impossible and suggests volumes are still small. Unitree, which filed for a Shanghai STAR Market listing on 20 March 2026, is the other domestic benchmark.

    In cobots, Dobot's six-axis growth of 44.7% competes directly with Universal Robots, FANUC's CRX series, and Chinese rivals such as JAKA and Elite Robots. Estun's return to profit points to firmer industrial automation demand in China in 2025, the same customer base Dobot's industrial cobots serve.

  3. 03

    Valuation context

    Dobot's equity rose to RMB2,622.6 million after placings that added RMB1,705.9 million, against annual revenue of RMB492.2 million. The market is therefore funding Dobot far more for its embodied AI option than for its cobot earnings, which remain loss-making on both reported and adjusted measures.

    The widening adjusted loss, RMB49.6 million against RMB36.5 million, matters for valuation because it strips out items that flatter the headline. Investors pricing Dobot as a humanoid company will want evidence that embodied AI revenue can scale faster than the RMB45.1 million research bill.

  4. 04

    Supply-chain implications

    Dobot's cobots and humanoids draw on the same Chinese component base of servo motors, reducers, encoders and controllers, much of it available from suppliers near its Shenzhen base. A 46.1% gross margin on hardware suggests it captures value in integration and software rather than in commodity parts.

    For Japanese reducer and sensor suppliers, the shift of Chinese cobot makers toward humanoids widens the number of joints they can sell into, but also invites domestic substitution as Chinese makers chase cost targets for humanoids.

  5. 05

    Signals to watch

    First, the interim report due by 31 August 2026 and whether embodied AI revenue rises as a share of the total. Second, whether Dobot discloses humanoid unit volumes, which would allow comparison with UBTech. Third, gross margin; a fall below 45% would suggest price pressure from domestic cobot rivals.

    Also watch selling expenses, which grew 32.1% in 2025. If overseas channel building keeps outpacing revenue, the path to profit lengthens.

  6. 06

    Analyst view

    Thesis: Dobot is a credible second-tier humanoid entrant whose real strength is a high-margin cobot business, and its embodied AI line will remain small in 2026. Confidence: medium.

    The financial data are audited and clear, but Dobot gives no unit numbers or 2026 targets, and the humanoid market is moving fast enough that a single large order could change the picture. The shipment ranking claim in the chairman's statement is unsourced and carries no weight here.

  7. 07

    Questions you should be asking

    How many ATOM and ATOM-W units did Dobot ship in 2025, and at what average price? What share of the six-axis growth came from overseas customers? Why did the adjusted loss widen while the reported loss narrowed?

    How much of the RMB1,705.9 million raised in 2025 placings is earmarked for embodied AI, and over what period? Does the shared software stack actually run on both arms and humanoids today, or is it a roadmap?