The global expansion of China’s robotics industry is taking a novel turn in Southeast Asia. Rather than simply exporting hardware, Chinese companies are beginning to export continuous algorithmic capabilities. In a significant development for the region’s manufacturing sector, Hangzhou-based YingShen Intelligence has signed agreements to deploy hundreds of its embodied robots across multiple footwear factories in Vietnam, utilizing an innovative token-based service model.
The agreements, signed with Vietnamese manufacturers Power Loong, Thuan Phi, and Vietnam Garment, represent a combined order value in the tens of millions of yuan. The robots, powered by YingShen’s proprietary 4D world model, will be deployed in factories in Hai Phong, initially focusing on complex, labor-intensive processes such as gluing and sole pressing.
The 4D World Model Advantage
The deployment highlights the rapid maturation of “embodied AI”, artificial intelligence that interacts with the physical world. Unlike traditional industrial robots, which blindly execute pre-programmed, repetitive motions, embodied robots are designed to perceive their environment, understand variations, and adjust their actions in real-time.
This adaptability is crucial in footwear manufacturing, an industry notorious for its resistance to automation. Shoe materials are flexible and easily deformed, and production lines must frequently switch between different styles, sizes, and material combinations. Traditional automation systems struggle with this variability, requiring costly and time-consuming reprogramming for every change.
According to Asia News Network, YingShen addresses this challenge using a “4D world model.” This approach trains the AI using data that captures not just the three-dimensional structure of objects, but also how they change over time (the fourth dimension). This allows the robots to understand physical interactions, such as contact, force, friction, and material deformation, more intuitively.
“Language is the native language of humans, while the native language of machines exists in the three-dimensional physical world,” explained Min Wei, founder of YingShen Intelligence. By training models on 4D physical data rather than just compressed 2D images or text, the robots can dynamically adjust their gripper posture and movement paths when faced with variations in shoe shape or material placement.
Pioneering a Token-Based Service Model
Perhaps more significant than the hardware itself is the business model YingShen is employing. The company is shifting away from the traditional paradigm of one-off equipment sales. Instead, it treats the robot as a physical platform through which its world model operates, offering a Robotics-as-a-Service (RaaS) model powered by tokens.
Under this arrangement, the Vietnamese factories are not just buying machines; they are subscribing to ongoing model inference, algorithm updates, and scenario-specific optimizations. The usage is metered and billed via tokens, similar to how cloud computing or API access to large language models is priced.
This model aligns the incentives of the robotics provider and the manufacturer. YingShen generates recurring revenue and continuously receives real-world operational data to refine its models. The factories, in turn, benefit from robots that become progressively smarter and more capable over time, without needing to constantly invest in new hardware.
As EastFrontier previously analyzed the commercialization of China’s robotics sector, finding sustainable business models is as critical as solving the technical challenges. The token-based approach offers a compelling pathway for scaling embodied AI in industrial settings.
Strategic Implications for Regional Manufacturing
The deployment of these robots in Vietnam is highly strategic. Vietnam has emerged as a major hub for global footwear and apparel manufacturing, absorbing significant production capacity that has shifted out of China due to rising labor costs and geopolitical tensions.
However, Vietnam is now facing its own labor shortages and rising wage pressures. By introducing advanced embodied AI into these factories, Chinese technology companies are helping to sustain the competitiveness of the Southeast Asian manufacturing base, while simultaneously embedding their technological standards and platforms into the region’s industrial infrastructure.
This dynamic creates a complex symbiotic relationship. While physical manufacturing may migrate across borders, the underlying intelligence and algorithmic control remain anchored in China. The export of embodied AI ensures that Chinese tech firms continue to capture a significant portion of the value generated by global supply chains, even when the factories themselves are located elsewhere.
The Future of Flexible Manufacturing
The initial deployment of YingShen’s ShadowGlue and ShadowPress robots is just the first phase. The partners plan to extend the technology to other processes, such as sewing and packaging, with the ultimate goal of building a benchmark factory for embodied intelligence-enabled footwear manufacturing in Vietnam.
If successful, this deployment will serve as a powerful proof of concept for the viability of embodied AI in highly variable, labor-intensive industries. It demonstrates that the future of manufacturing lies not in rigid, pre-programmed automation, but in flexible, intelligent systems capable of adapting to the messy realities of the physical world. For China’s robotics industry, the token-based export of these capabilities represents a sophisticated new phase of global expansion.