Humanoid robots are transitioning from digital concepts to real-world applications, crossing the threshold of mass production.
On December 8, Zhiyuan Robotics announced the completion of its 5,000th general-purpose embodied robot, the "Lingxi X2," marking a milestone in mass production. Zhiyuan’s co-founder, President, and CTO Peng Zhihui revealed that, as of now, 1,742 units of the Expedition A1/A2 series, 1,846 units of the Lingxi X1/X2 series, and 1,412 units of the Elf G1/G2 series have rolled off the production line.
However, mass production only signifies that the technology has passed initial market validation. Long-term reliability, operational cost control, and performance in more complex scenarios must still be tested through broader real-world deployments.
### Humanoid Robots Cross the Mass Production Threshold 2025 is hailed as the "first year of humanoid robot mass production." Breakthroughs in core technologies such as high-torque-density joint motors and multi-modal perception fusion algorithms are propelling humanoid robots from R&D to large-scale manufacturing.
For years, doubts persisted about whether humanoid robots could move beyond labs and achieve mass production. Now, production data from multiple companies is turning this vision into reality.
Recently, Zhiyuan Robotics disclosed that its three product lines have collectively produced 5,000 units: 1,742 in the Expedition series, 1,846 in the Lingxi series, and 1,412 in the Elf series. Peng Zhihui announced that Zhiyuan has officially entered the mass production phase.
Peng emphasized, "Producing 5,000 units validates our scalable manufacturing system. This gives us confidence to plan for 10,000 units next year and eventually 100,000 units."
At Zhiyuan’s newly opened production bases for the Expedition, Lingxi, and Elf series, rigorous testing—from durability checks for hundreds of components to software installation and multi-hour scenario simulations—ensures robots can operate reliably in unpredictable environments post-production.
Beyond Zhiyuan, other domestic robotics firms are accelerating mass production. UBTECH ROBOTICS (优必选) aims for an annual industrial humanoid robot capacity of 5,000 units by 2026, scaling to 10,000 by 2027. Unitree Robotics has achieved stable mass production and global sales in consumer and industrial quadruped robots, with humanoid robots entering trial production and delivery phases.
Peng also noted that mass production and delivery of general-purpose embodied robots will drive further cost reductions, paving the way for broader market adoption and creating a positive feedback loop for industry growth.
Additionally, many Chinese robotics companies are adopting a "self-developed + domestic supply chain" approach to break through core component bottlenecks, reducing reliance on expensive imported parts and further cutting costs.
Jiang Lei, chief scientist at the National Humanoid Robot Innovation Center, previously stated that the cost of humanoid robot electric joints has dropped from around 50,000–60,000 yuan in 2018 to approximately 500–600 yuan today.
### From Mass Production to Real-World Application: Bridging the Lab-to-Reality Gap Mass production is just the first step; the real challenge lies in large-scale deployment. Long-term reliability, operational costs, and performance in complex environments must be validated through extensive real-world use.
In lab settings, robots learn from limited, pre-set data. In real-world scenarios, they must adapt to dynamic, noisy environments. Rare but critical "long-tail problems" only emerge and can be addressed through widespread deployment.
An industry expert noted that collecting real-world data to continuously train perception, motion control, and interaction models will be key to the next phase of competition.
Peng Zhihui highlighted that multi-scenario deployments create a "data flywheel" effect: each additional robot in operation generates real interaction data, refining algorithm models through iterative training.
Currently, embodied intelligence applications are expanding along two main paths: industrial production and commercial services.
Industrial applications, with their structured environments and well-defined tasks, are often the first to adopt robotics. For example, Zhiyuan and UBTECH are piloting robots in automotive parts handling, precision assembly in 3C industries, and warehouse sorting and logistics.
Commercial service robots—used in roles like guiding, entertainment, and customer service—offer high-frequency, diverse human-robot interactions, providing valuable data to enhance perception and decision-making algorithms.
UBTECH’s CBO Tan Min previously stated that the core challenge lies in AI’s current inability to enable fully autonomous robot decision-making.
"Robots must first master repetitive tasks in controlled environments to train their 'brains,' advancing AI toward AGI (Artificial General Intelligence). Only then can they enter homes as service assistants," Tan explained.
Comments