Behind Wujie Dongli's 500 Million RMB Overseas Order: The Data-Algorithm-Order Triangle of Chinese Embodied AI

Forum topic · 小凯 · 2026-08-15

Summary

In early August 2026, Wujie Dongli (Beijing) signed a 500 million RMB order with Envision Technology, the first hundred-million-level overseas order in China's embodied intelligence manipulation sector, with the first robots destined for a French battery gigafactory. The company's total global orders reached 700 million RMB, including partners like ZF LIFETEC. Its competitive edge lies in infrastructure: the Shijingshan humanoid robot data training center, China's first and largest full-size bipedal humanoid data facility, produces nearly 20 million high-quality data entries yearly across ~700 scenarios using VR teleoperation, data gloves, and wearable haptic sensors. Experts estimate global real-robot data stands below 500,000 hours while generalizable models require tens of millions of hours. Wujie Dongli's latent-space world model (MWATM) bets on unified multimodal reasoning rather than the common VLA route, and its K15 robot holds EU CE industrial certification. Separately, Shangwei New Materials reported 160 million RMB in consumer robot R&D spending with 210 million RMB in pre-orders for its Qiyuan Q1 and T1, while Zhiyuan Robotics mass-produced 15,000 units in H1 2026. New national standards, cross-border data platforms, and output exceeding 40,000 humanoids in H1 round out a coordinated push across data, algorithms, orders, products, and policy.

In early August 2026, Wujie Dongli (Beijing) signed a 500 million RMB order with Envision Technology — the first hundred-million-level overseas order in China's embodied intelligence manipulation industry. The first batch of robots destined for a French battery gigafactory has already shipped. The company simultaneously announced total global orders of 700 million RMB, with partners including automotive supply chain giants such as ZF LIFETEC and Omover. How does a company less than two years old land international orders of this scale?

The Data and Certification Infrastructure

The answer lies in two places in Beijing: Shijingshan and Haidian.

The Shijingshan Humanoid Robot Data Training Center is China's first and currently largest full-size bipedal humanoid data training base. With three construction phases completed, it produces nearly 20 million high-quality data entries per year covering nearly 700 fine-grained scenarios.
It uses three collection channels: first-person perspective (VR headset + controller teleoperation), teleoperation (data gloves directly controlling dexterous hands), and wearable tactile sensors. The data feeds large models training both "perception" and "manipulation precision."
In Haidian's Embodied Intelligence Innovation Industrial Park, Wujie Dongli's pilot platform runs a closed loop from raw material inspection to full-machine testing. Its second-generation in-house robot, the K15, has obtained the EU CE industrial full-directive certification — effectively a "passport" into European industrial production lines.

The Global Data Gap

Zhu Kai, general manager of the Shijingshan training center, offered a stark comparison:

> "The total amount of real-robot data collected worldwide does not exceed 500,000 hours. Yet training a general-purpose model that lets robots adapt to complex environments requires at least tens of millions of hours of high-quality data."

That gap — 500,000 hours versus 10 million hours — is the most direct picture of today's global embodied-AI data supply-demand imbalance.

Wujie Dongli's latent-space world model architecture (MWATM) departs from the industry's common VLA route, emphasizing unified representation and reasoning of multimodal information within latent space — a key technical bet on packing "perception + physical laws + causality" into one model. Orders from both an energy giant (power batteries) and the automotive supply chain (chassis and electronics) mean their "general-purpose brain" is receiving real-world feedback from at least two industrial scenarios.

Consumer-Side Validation

A parallel consumer line runs through Shangwei New Materials. Its H1 2026 report, disclosed August 14, showed roughly 160 million RMB in R&D investment for consumer embodied intelligence robots, two products launched (Qiyuan Q1 and T1), 210 million RMB in pre-orders, an R&D team expanded to 227 people, and 273 cumulative patent applications. Still in a "business validation" phase, the 160M R&D / 210M pre-order structure signals preparation for mass production.

Add Zhiyuan Robotics, which mass-produced 15,000 units in H1 2026 (half-size humanoids accounting for the largest share), and China's embodied manipulation sector now shows simultaneous leaps along three axes: B-end orders-and-data, C-end product R&D, and manufacturing-line data.

Policy and Cross-Border Data Channels

Supporting policy landed in step:

The "Humanoid Robot and Embodied Intelligence Standard System (2026 Edition)", released in February 2026, covers the full industrial chain, full technology stack, and all application scenarios.
MIIT Deputy Minister Ke Jixin reported China produced roughly 20,000 humanoid robots last year, over 40,000 in H1 2026, with a full-year forecast exceeding 100,000.
At the August 2026 CIFTIS fair, Shijingshan will launch a "technical service platform for cross-border circulation of embodied intelligence data elements under compliance frameworks" — the interface design for data flowing from factories to overseas markets.
Seven national standards for humanoid robot test methods are already underway.

Conclusion

Data, algorithms, orders, C-end products, policy, and cross-border channels — five layers stepped up simultaneously for Chinese embodied AI in a single summer. Each item alone is industry news; stacked together, they form an "infrastructure week": starting from data elements, Chinese embodied intelligence linked industrial orders, overseas orders, consumer pre-orders, policy bills, and the national standards system in one stroke.