The most consequential number in China’s humanoid-robot competition may not be a sprint time or a benchmark score. It may be the price at which a machine can perform useful work. Zhao Tongyang, founder and chief executive of Shenzhen-based ENGINEAI, says its per-unit cost has dropped below 100,000 yuan for a humanoid intended to handle practical general tasks. TechNode reported the claim after Zhao’s comments around the World Robot Conference.
The figure is a company statement, not independently audited manufacturing data. It also refers to cost rather than the selling price a customer would pay. Even so, the claim is strategically important. If practical humanoids can be produced at that level while maintaining reliability, China’s large manufacturing base could move the industry away from expensive prototypes and toward more realistic deployment economics.
Below 100,000 Yuan Is a Manufacturing Claim, Not a Retail Offer
Zhao said comparable machines cost more than 1 million yuan three years ago, according to TechNode. The comparison suggests a steep decline in component and production costs, but it should be interpreted with care. A robot’s bill of materials is only one part of its cost. Companies must also pay for research, software, data collection, testing, integration, support, safety systems, facilities, and the losses that come with early production runs.
The question for customers is not whether a robot can be made cheaply in one factory. It is whether it can complete tasks reliably enough that the total cost of ownership makes sense. That includes electricity, maintenance, spare parts, remote supervision, worker training, insurance, and the cost of downtime. A humanoid that needs frequent assistance may be less economical than a simpler automation system, even if its headline unit cost is attractive.
Still, lower production cost changes the range of possible pilots. A business may be willing to test a robot in a warehouse, store, factory, or controlled service setting if the capital risk is manageable. It becomes easier to gather data from real work, identify failure modes, and improve the next version. That feedback loop is central to physical AI because robots learn from environments that are more variable than a web page or a benchmark dataset.
China’s Supply Chain Is Becoming Part of the AI Argument
ENGINEAI’s argument rests on China’s established manufacturing base. Zhao told TechNode that the domestic humanoid supply chain has become more complete, although moving traditional manufacturing into precision robot components still requires experience and technical development. Sina’s reprint of IT Home says the company’s Shenzhen intelligent-manufacturing base opened in May and that ENGINEAI has said its line can produce one T800 humanoid every 15 minutes. That production-rate figure is also a company claim and should be treated as an indication of ambition rather than a proven output record.
What makes the cost debate important is that it links AI directly to industrial organization. Better models help robots understand scenes and coordinate actions. But widespread deployment also depends on motors, gearboxes, sensors, batteries, cameras, compute modules, assembly processes, and suppliers that can deliver consistent quality. China’s advantage may lie in combining model development with a large ecosystem for making hardware at volume.
The recent Unitree story on low-cost dexterous robotics showed why price has become a core competitive message. Companies are trying to persuade buyers that humanoids can become useful before they become perfect. ENGINEAI’s claim pushes that argument further by setting a visible threshold for general-purpose machines rather than specialized robotic arms or tightly constrained devices.
Practical Work Remains the Hardest Benchmark
Price alone will not decide the market. A robot must perceive objects under changing light, avoid people, handle unexpected conditions, operate safely over long shifts, and recover when a task goes wrong. Those problems are why the humanoid field remains a mixture of impressive demonstrations and cautious deployment plans.
China’s policymakers are also beginning to focus on the conditions for real deployment. The Beijing robot conference report showed why common tests for capabilities, interfaces, safety, and industrial operation are becoming necessary. Standards may not make a robot cheaper, but they can help customers understand what a lower-cost machine can actually do.
ENGINEAI’s 100,000-yuan claim should therefore be viewed as an opening bid in a larger debate. The relevant measure is not the cheapest possible machine, but the cost of a robot that can complete a defined task repeatedly, safely, and with a service model that customers can support. If Chinese companies can meet that test, they could accelerate adoption in manufacturing and services. If they cannot, lower unit costs will not overcome the gap between a conference demonstration and a dependable worker.
The importance of Zhao’s statement is that it moves the conversation toward economics. China’s humanoid sector has already shown that it can attract investment and generate public attention. The next phase requires evidence that a machine can create more value than it costs. That calculation will determine whether physical AI becomes a broad industrial market or remains a high-profile technology experiment.