The embodied intelligence community in China was set ablaze last week after Xpeng, best known for making electric vehicles, unveiled its humanoid robot Iron on November 5.
The controversy began when clips of Iron walking went viral. The robot’s gait appeared so lifelike that some viewers claimed to spot the outline of a human ear and what seemed like the contour of a woman’s undergarment beneath its back covering. Speculation erupted online, with many questioning whether a real person was inside the robot.
In response, Xpeng founder and CEO He Xiaopeng posted a video to dispel the rumors. In it, he cut through Iron’s leg covering with a pair of scissors to expose its internal frame, then unzipped the back panel to reveal a cream-colored, mesh-patterned robotic torso.
Yet curiosity persisted. How could Iron move with such fluid, humanlike grace, so unlike the stiff and mechanical motion typical of most robots?
According to 36Kr, the answer lies in a combination of breakthroughs: a bionic spine, joints offering 82 degrees of freedom, and a large model trained on human gait data. Perhaps the most critical factor is a new material used to form the robot’s “muscles.” That material is known as elastomer.
“When people start questioning whether a robot might actually be human, it means they have begun to measure it by human standards,” said Xiao Bowen, CMO of PollyPolymer, in an interview with 36Kr. “That kind of debate shows how far the industry’s imagination has come.”
PollyPolymer, a supplier in the elastomer materials field, produces high-performance materials through 3D printing at speeds 20–100 times faster than conventional manufacturing. Its clients in robotics include Agibot and UBTech Robotics.
Vitalbridge Capital, an early investor in PollyPolymer, believes materials innovation has become a key factor in the evolution of embodied intelligence. Traditional robots rely on rigid metal frames, while 3D printing enables flexible, lightweight elastomers. If a future where humans and robots coexist is inevitable, humanoid robots will need more than an intelligent brain. They will also need bodies that are safe, coordinated, and efficient, qualities elastomers can help provide.
Building robots with elastomers
Elastomers occupy a middle ground between rigid materials and silicone. They can be squeezed, stretched, or even thrown without suffering damage.
Initially developed for aerospace applications, elastomers were later used in military helmets as shock absorbers. In recent years, they have found their way into consumer goods such as shoe soles, bicycle grips, mattresses, dental aligners, and wearable devices.
This balance of strength and flexibility makes elastomers particularly well suited for humanoid robots.
According to Xiao, traditional robot shells are made from rigid metals like stainless steel or aluminum, which restrict movement. Elastomers, by contrast, respond continuously and flexibly, mimicking the way human muscle groups contract and stretch. This allows robots to move more naturally, an essential trait for use in homes, elder care, and service settings.
Weight is another advantage. At equal volume, elastomers are roughly 60–70% lighter than metals, allowing humanoid robots to redistribute mass to critical components such as joints, hands, and heads. The result is lower torque and energy consumption and greater safety in physical interactions with humans.
Continuous motion also generates heat, but the lattice-structured elastomer dissipates it efficiently because of its breathable design.
Xiao predicts that elastomers will become the standard material for robotic “muscles.” Beyond Xpeng’s Iron, other notable examples include UBTech’s Walker S2 (elbow joints), Agibot’s Lingxi X2 (arms, chest, and legs), and the Figure 02 (joint cushioning).
Still, building artificial muscles is complex. Flexible materials introduce new challenges, such as micro-deformation control, that require more precise algorithms and feedback systems.
Manufacturing adds another layer of difficulty. The lattice structure of elastomers cannot be molded using conventional methods, which lack the precision needed to form intricate internal geometries.
Instead, elastomers are fabricated through 3D printing. Under controlled light exposure, photosensitive resin filaments interweave layer by layer to form Iron’s elastic body, echoing in some ways how human muscles develop.
Ultimately, it is this convergence of materials and technologies that gives Iron its remarkably lifelike motion.
RELATED ARTICLE
Why Xpeng’s strategy hinges on bionics
Xpeng has long emphasized the “human” in humanoid. From its dexterous bionic hand and vision system that surpasses the human eye to its refined gait and now lifelike “muscles,” the company’s robots are evolving toward human resemblance.
Why does bionics matter so much? CEO He told reporters that robots resembling humans are more approachable, which drives user acceptance and supports economies of scale. This, in turn, helps lower costs and close the business loop.
Since 2020, Xpeng has been developing a closed ecosystem connecting its automotive and robotics divisions. He has said that the company’s humanoid robots will first be deployed in its own factories, where they will assist in vehicle production. Beyond improving efficiency, this environment doubles as a testing ground for real-world motion data.
Automotive factories are designed for human workers, making them ideal settings for humanoid robots to collect high-quality behavioral data. This information feeds back into the design process, accelerating iteration and creating a data flywheel. In this loop, bionics is not just a design principle but a mechanism that sustains continuous improvement.
Compared with other robotics players, Xpeng holds a structural advantage. Aside from components such as joints and synthetic muscles, nearly 70% of its core technologies overlap with its automotive business.
In a recent interview, He outlined his long-term outlook: once robotics crosses its technological and product inflection point, he expects the industry to enter a rapid growth phase. Over the next 10–20 years, he believes the global robotics market could reach USD 20 trillion, roughly twice the size of the automotive sector.
“I don’t know how many robots we’ll be selling ten years from now,” he said. “But I’m certain it’ll be more than cars.”
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This article was written by Qiu Xiaofen for 36Kr.