Latest Updates

Reporter Qiu Siyu and Yu Liyan

 

Artificial intelligence is moving from the digital world to the physical space, and humanoid robots are considered one of the most important landing vehicles for physical AI. With Tesla revealing that its third-generation humanoid robot Optimus is ready for mass production, market attention is once again focused on this track.

 

Shanghai Securities News reporters interviewed industry experts and learned that as a long-term growth direction for Tesla's key layout, if Optimus smoothly enters the production ramp up stage, it is expected to drive the release of demand for components such as actuators, reducers, and sensors in China. Many upstream manufacturers have already prepared production capacity in advance. However, the road to industrialization is not smooth. Hardware reliability, physical AI algorithm generalization ability, cost control, industry compliance standards, and data closed-loop construction are still the core constraints for the development of the humanoid robot industry.

 

Optimus ready for mass production

 

Upstream components are expected to receive orders

 

Recently, there has been a flurry of news about the production progress of Tesla's third-generation humanoid robot Optimus. On October 19th Eastern Time, Tesla announced in writing in its annual report to shareholders that the third-generation humanoid robot Optimus is ready for mass production. The chairman of the board of directors of the company, Robin Dehom, wrote in a letter to shareholders that Tesla continues to carry out product research and development and invest in building the infrastructure required for large-scale delivery, including the construction of the Semi (Tesla semi-trailer heavy-duty truck) factory in Nevada and the first batch of Optimus production lines. Previously, Tesla's core robot supplier, New Sword Transmission, stated in a response to the listing audit inquiry letter that Tesla's robot design scheme is close to finalization and will soon enter the mass production cycle.

 

The humanoid robot Optimus is one of Tesla's key business directions for layout. On October 19, Elon Musk, CEO of Tesla, said on the social platform that the company's Optimus humanoid robot and Cybercab autonomous vehicle would constitute the "vast majority" of Tesla's value, and reiterated his focus on robots and artificial intelligence.

 

Industry experts told reporters that with the development of intelligence, many industrial chain enterprises in China have expanded from automotive business to robotics business. With the increase in Tesla's robot production capacity, its demand for robot components will also significantly increase.

 

The reporter previously learned that the Tesla robot team has started factory audits at listed companies headquartered in Ningbo, Zhejiang, such as Top Group, Sanhua Intelligent Control, and Junsheng Electronics. Tesla's factory audit leads to the mass production of the humanoid robot Optimus, which is expected to distribute related orders to Chinese supply chain enterprises.

 

Based on experience, if Tesla's third-generation humanoid robot Optimus enters mass production, the demand for related components will also increase significantly, and the company is ready to ramp up production capacity. ”On October 20th, the head of an industrial chain company in Zhejiang Province told reporters that the company is an upstream supplier of components for Tesla's humanoid robots, as well as an upstream supplier for multiple domestic humanoid robots.

 

The hardware progress of humanoid robots is remarkable, which is also the basis for the market's hope for Tesla's third-generation humanoid robot Optimus. ”Roland Berger Global Partner Shi Shuai told reporters that high torque density actuators, advanced reducers, visual sensors, edge AI chips, and battery management systems have entered the quasi commercial stage.

 

Xiao Qunxi, Chief Analyst of Guotai Haitong Machinery Industry, believes that the landing of Tesla Optimus production line is accelerating, and the mass production of humanoid robots has entered a critical verification period. This marks the gradual shift of the humanoid robot industry chain from thematic catalysis to the stages of capacity building, supply chain introduction, and order verification, which is expected to accelerate the performance of core components and automation equipment.

 

Cost reduction is the primary hurdle

 

Shortcomings in software and hardware still need to be addressed

 

In recent years, the development of the humanoid robot industry has accelerated, and domestic and foreign markets have gradually formed distinctive development paths. Shi Shuai believes that Tesla's North American market is more inclined towards the "AI first" path: based on basic models, visual language models, and proprietary data as the core, pursuing the general operational capabilities of robots in unstructured environments. The domestic track presents a distinct feature of "deployment priority, manufacturing driven". Chinese companies are accelerating the deployment of robots in entertainment, logistics, and other clearly defined and controllable work scenarios, relying on high-frequency deployment to accumulate data, using large-scale manufacturing to compress costs, and further expanding deployment with cost advantages.

 

Cost control is also the primary hurdle that Optimus needs to overcome for large-scale production. What truly drives the humanoid robot industry out of the laboratory is not just the cool technology, but the increasingly clear unit economy, "said Shi Shuai.

 

The Roland Berger research report predicts that the cost of advanced humanoid robots for complex tasks is expected to decrease to $20000 to $30000 per unit around 2035, including training costs; The entry-level product for basic automation and service scenarios may be priced at $8000 to $10000.

 

More importantly, the operating cost of humanoid robots is expected to be as low as about $2 per hour, "Shi Shuai said. This means that humanoid robots will not only serve to" replace expensive labor ", but may also become an important lever to alleviate structural labor shortages, improve night and weekend production capacity, and lower the threshold for flexible automation. However, the key subsystems such as actuators of humanoid robots still need to significantly reduce costs, and the cost reduction of some links needs to reach 50% to 90%.

 

From the perspective of hardware maturity, the industry has generally expected that hardware design will initially stabilize around 2028 to 2029. However, 'being able to run' does not necessarily mean 'being able to deploy on a large scale'. The biggest challenge at present has shifted from mechanical design to AI architecture, training data, durability, supply chain, and security supervision. The maturity of software and ecosystem still lags behind hardware by three to five years overall. ”Shi Shuai said.

 

Li Yifan, co-founder and CEO of Hesai Technology, mentioned to reporters the importance of training data. It is reported that real-world data is currently the scarcest "fuel" in the embodied intelligence industry. High quality 3D data is scarce in the real world, and the cost of collection is also high.

 

Analysts say that the bottlenecks in the industrial chain are concentrated in three aspects: firstly, insufficient hardware reliability, and the lifespan and stability of joints and dexterous hands need to be continuously verified under long-term and multi working conditions; Secondly, the generalization ability of physical AI algorithms is insufficient, and the success rate of actions in complex and unstructured scenarios is low; Thirdly, there is a lack of industry safety and certification standard systems, and there is a lack of mature norms for human-machine collaboration scenarios, which hinders commercialization implementation. Overall, cost, reliability, compliance, and data closure jointly constrain the pace of the industry.

 

(Editor in charge: He Xin)