Don’t be fooled by “mass production next year”! From costs and safety to supply chains, understand the real industrialization timeline.
1. A Harsh Reality: There Is No True “Mass Production” Globally
As of November 2025:
- Tesla Optimus: Gen-2 is only in small batches for internal factory testing and not for external sales;
- Utree G1/H1: Approximately 6,000 units shipped in total in 2025, mainly for research, developers, and pilot customers;
- UBTECH Walker X: Deployed in closed environments such as power grids and exhibition halls, with a unit price exceeding 500,000 yuan;
- Figure 01: Testing in collaboration with BMW, but the cost per unit still exceeds 200,000 USD.
💡 Definition Clarification: “Mass production” ≠ “can be produced”, but refers tocontrollable costs, stable yield, customer repurchase, and large-scale delivery (over 10,000 units per year). Currently,no one globally meets the standards.
2. Five Core Challenges That Block Industrialization
Challenge 1:High Costs – One Robot = One Luxury Car
| Component | Current Cost (Yuan) | Mass Production Target (Yuan) | Gap |
|---|---|---|---|
| Dexterous Hand (12 Degrees of Freedom) | 38,000 | ≤8,000 | 4.75 times |
| High Load Joint Module | 12,000/unit × 28 units ≈ 336,000 | ≤3,000/unit | 4 times+ |
| Main Control Chip + Sensors | 65,000 | ≤20,000 | 3.25 times |
| Battery System | 28,000 | ≤8,000 | 3.5 times |
| Total BOM Cost | ≈580,000 | ≤150,000 | Nearly 4 times |
📌 Key Conclusion: Even without making a profit, the current selling price is hard to drop below 400,000 yuan, far exceeding the psychological threshold for industrial customers (≤200,000 yuan).
Challenge 2:Weak Supply Chain – Domestic Production Rate Below 40%
Despite claims of “full-stack self-research”, key links still rely on imports:
- Hollow Cup Motors: High-end models are still sourced from Germany’s FAULHABER and Switzerland’s MAXON;
- Harmonic Reducers: Green’s harmonic has broken through, but the yield for high-torque models is only 70%;
- Torque Sensors: 90% rely on the US’s ATI and Japan’s NMB;
- AI Training Chips: The MUSA ecosystem is in its early stages, with most still using NVIDIA A100/H100.
💡 Risk Point: If geopolitical tensions escalate, the delivery of complete machines may be interrupted.
Challenge 3:Safety and Regulatory Gaps – No One Dares to Let It into Homes
- No Mandatory Certification Standards: China has yet to issue national safety standards for humanoid robots;
- Ambiguous Responsibility: If a robot injures a worker, who is responsible – the manufacturer, the user, or the algorithm?
- Lack of Insurance: Mainstream insurance companies refuse to insure humanoid robot product liability.
📌 Real Impact: All commercialization scenarios are forced to be limited toclosed, low-risk environments (such as substations and warehouses), greatly limiting market space.
Challenge 4:Insufficient Software Intelligence – “Can See” but “Cannot Understand”
- Visual recognition has a 99% accuracy in the lab, but drops to 85% in factory bright light and dust environments;
- Poor multi-task scheduling ability: cannot simultaneously handle “transport + obstacle avoidance + voice interaction”;
- Lack of support from general large models: existing AI is mostly task-specific, with poor generalization ability.
💡 Typical Case: In a pilot factory, a robot misjudged equipment failure because it could not recognize a reflective dashboard, leading to a production line shutdown.
Challenge 5:Unproven Business Model – Customers Are Unwilling to Pay for the “Future”
- Factory owners calculate: one robot saves 2 workers, saving 180,000 yuan per year, but the purchase cost exceeds 400,000 yuan, with a payback period of >2.5 years;
- Compared to AGVs/robotic arms: dedicated equipment costs only 50,000–100,000 yuan, with ROI < 1 year;
- Result: Humanoid robots have become “image projects”, making it difficult to achieve repurchase.
3. Path to Breakthrough: Which Links Are Accelerating Progress?
✅ Path 1:Prioritize Industrial Applications, Abandon the “All-Purpose Fantasy”
- Focus onsingle high-value tasks: such as distribution cabinet inspection and high-voltage equipment operation;
- Reduce degree of freedom requirements (from over 40 to within 20), directly reducing costs by over 30%;
- Utree H1 and UBTECH Walker X have validated that this path is feasible.
✅ Path 2:Accelerate Domestic Substitution of Core Components
- Haozhi Electromechanical’s hollow cup motor yield has reached 93%;
- Green’s harmonic and dual-flying bearings have passed industrial durability tests;
- Huahai Chengke and Lianrui New Materials have made breakthroughs in advanced packaging materials.
📌 It is expected that by 2026–2027, the BOM domestic production rate will exceed 70%.
✅ Path 3:Policy Promotion of Standards and Pilot Projects
- The Ministry of Industry and Information Technology has released the “Safety Specifications for Humanoid Robots (Draft for Comments)”;
- The National Supercomputing Center is leading the construction of a “Embodied Intelligent Large Model”;
- BYD and CATL have committed to purchasing ≥1,000 units each by 2026.
4. A Framework for Ordinary People: How to Track Real Progress?
Don’t listen to press conferences, focus on these 3 signals:
- Cost Signal: Is the total BOM cost below 300,000 yuan? (Pay attention to Haozhi and Green’s quarterly gross margins)
- Order Signal: Are there non-government customersself-purchasing (not pilot)? (Check announcements from Yijiahe and Shenhao Technology)
- Standard Signal: Has the state issuedmandatory safety certification? (Updated monthly on the Ministry of Industry and Information Technology’s official website)
⏳ Reasonable Expectations:
- 2026: Small batch volume in industrial scenarios (annual shipments of 10,000–20,000 units);
- 2027–2028: Costs drop below 200,000 yuan, entering a scaling inflection point;
- Before 2030: Household scenarios will still be difficult to popularize.
5. Conclusion: Slow is Fast
Humanoid robots are not the next smartphone, but ahard technology long march that requires ten years of patience.
The real opportunity does not belong to those who shout “mass production next year”,but to those who, in the laboratory, refine every gram of weight, every watt of power consumption, and every millisecond of delay into certainty.
Smart machines do not serve as a megaphone for quick success myths,but as long-term partners.
Disclaimer: This article does not constitute any investment advice. The market has risks, and decisions must be made independently.
🔖 Tags:#HumanoidRobot #EmbodiedIntelligence #HardTech #SmartManufacturing #AIInvestment #Industrialization #UtreeTechnology #TechFrontier #InvestmentPractice #SmartMachine