Zhishen Robotics at WRC 2026: 15,000 Robots Shipped, 'Legs First, Hands Later' vs. the Embodied AI Demo Bubble

Forum topic · 小凯 · 2026-08-26

Summary

At WRC 2026, Zhishen Robotics (智身科技) co-founder Liu Yulong challenged the industry's 'demo bubble' with hard numbers: 15,000+ robots cumulatively mass-produced, monthly capacity of 5,000+ units, thousands of quadrupeds deployed in security patrols, 100+ power inspection projects, and coverage of 100+ fire brigades. The company's strategy — 'build the legs that run first, then attach the hands that work' — prioritizes quadruped robots for all-terrain mobility and autonomous navigation, monetizing mature modules in security, emergency response, and power inspection before returning to humanoid manipulation intelligence. Key engineering specs include the CHAMP P85MAX-S joint module (1 kg driving 90 kg loads, 1 million units/year capacity), the Gangdeng L2 quadruped (7:1 load-to-weight ratio, IP66 rating, underwater operation), and a 'chain leader + module output' business model with 300+ integrator partners. Liu argued the industry's biggest gap is not technology but the absence of shared standards for performance metrics, evaluation, and certification, and that 'laying eggs along the way' — revenue-funded iteration — offers a viable alternative to burning cash while waiting for perfect humanoids.

Zhishen Robotics at WRC 2026: 'Legs First, Hands Later' — Piercing the Embodied AI 'Demo Bubble' with 15,000 Mass-Produced Robots

> Beijing time, 2026-08-27 · Zhichai AI · Embodied Intelligence column

Source: China.com interview with Zhishen Robotics co-founder Liu Yulong, Aug 26, + WRC 2026 coverage

Key points

While much of the industry showcases staged demos, Zhishen Robotics (智身科技) co-founder Liu Yulong presented verifiable figures at WRC 2026: 15,000+ cumulative mass-produced units, 5,000+ monthly production capacity, thousands of robots deployed in security, 100+ delivered power inspection projects, and coverage of 100+ fire brigades.

His strategy in one line: 'Build the legs that can run first, then attach the hands that can work.' The company focused on quadrupeds because all-terrain mobility and spatial navigation are already solved; once those modules prove themselves in real security, emergency, and power scenarios, the company will return to humanoid manipulation intelligence.

This stands as a mirror image of the WHRG 2nd World Humanoid Robot Games (theatrical embodied AI) — Zhishen represents embodied AI in the real world: 24-hour community patrols, fire brigade solutions, routine substation inspection.

The 'Demo Bubble': 100,000 units on PPT vs. hundreds in reality

At WRC 2026, over 400 robot products were exhibited, with claimed annual output of 100,000 units.
Unitree chairman Wang Xingxing was blunt: 'In fixed scenes, task success rates approach 100%; change the object or environment and success rates collapse.'
Zhishen's booth instead showed quadrupeds climbing open-grid stairs, mounted arms operating precisely, and joint modules running continuously 1 meter underwater.
Liu: 'We are not "planning to enter" these scenarios — we are already running in them.'

'Laying eggs along the way': 15,000 units is not a slogan

Most peers bet everything on waiting for a perfect general-purpose humanoid — continuous funding, no revenue. Zhishen decomposes embodied intelligence into modules (all-terrain mobility, spatial cognition/navigation, manipulation). Once two modules mature, quadrupeds can already do tedious, repetitive, dangerous work.

Hard data as of June 2026:

| Dimension | Figure | |---|---| | Cumulative production | 15,000+ units | | Monthly capacity | 5,000+ units | | In-house joint modules | 1M units/year | | Security deployments | Thousands (residential/malls/scenic coastlines) | | Power inspection projects | 100+ | | Emergency product matrix | 20+ models | | Fire brigade coverage | 100+ | | Ecosystem partners | 300+ | | Cities/provinces covered | 100+ |

Three gates: from 'can be built' to 'works well'

1. Technical gate — stable reproducibility, not one-off demos: IP66 outdoor rating; Gangdeng L2 static load 150 kg, sustained carry 16 kg; P85MAX-S joint drives 90 kg load at ~1 kg weight; joint reliability, thermal rise, fall recovery; demonstrated by joint modules running 1 m underwater. 2. Scenario gate — completing the customer's task, not just reaching the site: integrating body, navigation, payloads, comms, algorithms, and business rules into a definable, verifiable deliverable. 3. Sustained delivery gate — folding recurring needs back into standardized products via standard interfaces, SDKs, and modular iteration.

Capability dimensions: strength, environment, intelligence

Strength: P85MAX-S joint (1 kg driving 90 kg); Gangdeng L2 (7:1 load-to-weight ratio); humanoid Yinyi NE01 single-joint peak torque 180 N·m.
Environment: 2nd-gen in-house 'Qilin alloy' joints, IP66 industrial outdoor rating; continuous underwater operation.
Intelligence: 'Big brain + cerebellum' — automotive-style rule-stack + end-to-end navigation, and a perception-control unified model for autonomous industrial stair climbing.

Business model: chain leader + module output

Upward: self-developed CHAMP high-power-density joint modules, 1M units/year, supplying core actuation to the industry.
Downward: 300+ integrators, universities, and industry partners handling scenario delivery.

The point: separate building robots from using robots — the embodied AI company handles core components and brains; integrators handle scenarios.

Three verticals with checkable numbers

Security: thousands of quadrupeds on routine patrol in residential communities, commercial districts, industrial parks, and coastlines; all top-tier Chinese security integrators are partners. One robot replacing one night-shift guard (~5,000 RMB/month savings) × mass deployment = substantial labor-cost reduction.
Emergency: 100+ fire brigades in Beijing, Tianjin, Gansu, etc., with a 'robot dog + drone/off-road vehicle' multi-robot solution; 20+ product models covering explosive, toxic, smoke, and heat scenarios.
Power: 100+ delivered inspection projects covering substations and energy storage stations — storage sites with high temperature, humidity, and gas-leak risks are ideal robot territory.

'Lay eggs along the way' vs. 'wait for perfection'

| Dimension | Wait for perfection | Lay eggs along the way | |---|---|---| | Revenue start | After full humanoid maturity | Now, via quadrupeds | | Unit cost | High (R&D on few units) | Medium (amortized by volume) | | Customer validation | Closed-door R&D | Continuous feedback loop | | Cash flow | Continuous fundraising | Revenue funds R&D | | +6 months | Lab prototypes | Thousands of units running | | +12 months | Struggling to commercialize | Large deployments + new modules |

Liu on why robots deserve to exist: protect lives (machines enter danger first), fill gaps in tasks humans cannot sustain, raise efficiency by standardizing experience-based work, and advance industry by giving sensors and algorithms a mobile carrier into the real world.

The real bottleneck: industry standards, not technology

Liu's overlooked WRC point: 'The industry still lacks mature standards — performance metrics, functional metrics, evaluation systems, certification.' No standards means no scale, no cost advantage, no real deployment. Whether China's embodied AI sector moves from fragmented private standards to shared industry standards in the next 12–18 months may decide whether valuations shift from PPT-based to revenue-based.

Key insights recap

| Dimension | Zhishen | Peer demo approach | |---|---|---| | Product | Quadruped first, humanoid later | All-in on humanoids | | Production | 15,000+ / 5,000 monthly | PPT 100k / actual hundreds | | Scenarios | Security + emergency + power | Staged booth demos | | Customers | Tier-1 integrators + fire brigades | Investors + media | | Revenue | Rolling | Continuous fundraising | | Model | Chain leader + module output | Build and sell alone | | Moat | 1M joint modules/year | Concept + story |

References

1. China.com, 'Interview with Zhishen Robotics co-founder Liu Yulong' (2026-08-27) — https://news.china.com/socialgd/10000169/20260827/49701024.html 2. Sina Finance / China Quality News (2026-08-26) — https://finance.sina.cn/2026-08-26/detail-iniprqcu2379993.d.html 3. Consumer Daily (2026-08-26) — https://www.xfrb.com.cn/article/kj/13300518997982.html 4. Tencent / CNR Finance interview (2026-08-26) — https://new.qq.com/rain/a/20260826A05XJU00 5. NetEase (2026-08-26) — https://www.163.com/dy/article/L59Q6FQN0556C2XD.html