Source: LatePost
Author: Li Zinan
"Old Ma's credibility score feels like it can't even borrow a power bank anymore," said a Tesla supply chain insider. Tesla had originally planned to finalize the technical solution for its third-generation robot at the beginning of this year and start mass production, which would allow this supplier to secure orders worth hundreds of millions of dollars. For this purpose, they have invested hundreds of millions of yuan and hundreds of personnel into research and development over nearly three years.
Musk confirmed during yesterday’s Q3 earnings call that the finalized design of the third-generation humanoid robot, Optimus, would be postponed until the first quarter of next year. On July 14th, Musk posted on X that he had just left Tesla’s design studio and that there would be an epic, groundbreaking demonstration by the end of the year. According to two supply chain insiders, Tesla originally intended to complete the development of the third-generation Optimus by November this year and hold a launch event.
"Creating a hand as dexterous and efficient as a human hand is an exceedingly difficult task," said Musk. We understand that$Tesla (TSLA.US)$the technical route for the dexterous hand was determined just half a month ago, but numerous engineering challenges still lie ahead.
The dexterous hands of the Optimus robot are already among the most sophisticated in the industry. However, when training to sort packages, their lifespan is only six weeks. The cost of a single hand exceeds $6,000, and considering other easily damaged components, excluding electricity costs, replacing parts for one robot to continue operating amounts to nearly $100,000 per year. This sum could hire two automotive assembly workers in the U.S.
The delay in R&D progress has not yet impacted Tesla's mass production plans for Optimus as communicated to suppliers. However, the aforementioned suppliers all indicated that they do not fully trust Tesla’s guidance because the engineering difficulties faced by Optimus extend beyond just the hands; everything remains in rapid flux.
Most components have been finalized in design, but all are waiting for the hand.
In July this year,$Tesla (TSLA.US)$Tesla temporarily suspended the procurement of Optimus components and redesigned some parts. According to supply chain sources, Tesla has largely finalized suppliers for all components except for the dexterous hands and forearms. In September this year, Tesla resumed component procurement and began producing Optimus units without hands, with a weekly production capacity exceeding 200 units.
The immediate goal of these changes is to enable Optimus to deliver an impressive live demonstration, rather than achieving the level of intelligence and flexibility previously promised by Musk that would rival human capabilities.
“All humanoid robot companies frequently release update videos. Therefore, Musk needs a flawless product and an extraordinary launch event to make Optimus significantly outperform its competitors. For instance, having the robot dance on stage with him or interact face-to-face with investors, rather than just releasing a video,” said a senior executive from one of Tesla's suppliers.
Since the beginning of this year, Musk has made numerous promises regarding Optimus. He claimed that 80% of Tesla’s future value will come from robots and subsequently purchased $1 billion worth of Tesla stock.
Supply chain information indicates that to enhance the flexibility and sensory capabilities of the dexterous hands, Tesla increased the number of actuators in the dexterous hands from 17 to 22 in the July redesign and added electronic skin sensors to the palm area.
These modifications still face significant engineering challenges.
Optimus's dexterous hands continue to encounter issues such as insufficient motor cooling and short service life for certain components. One source stated that the lifespan of Optimus’s dexterous hands does not exceed two months, with the flexible electronic skin covering the fingers and palms wearing out faster when Optimus touches objects. Additionally, the tendon ropes driving the fingers are prone to aging and breaking.
Musk noted that to simulate the capabilities of the human hand, Tesla engineers even consulted with surgeons to study the muscles and force mechanisms of the human hand.
After increasing the degrees of freedom, arranging actuators within the dexterous hand becomes more challenging. Tesla needs to redesign the mechanical structure and circuits to accommodate five additional miniature ball screws and motors in a palm already packed with 17 pen-cap-sized micro ball screws and dozens of motors. The precision metal gears inside the dexterous hand wear out during high-frequency resonance and rotation, losing accuracy.
A supplier once told us that the dexterous hand could also be damaged due to the failure of motion control algorithms, such as when the robot might lose control and smash its hand on a tabletop. Once broken, the intricately structured dexterous hand can only be replaced, not repaired, with each unit costing over $6,000.
The main body continues to undergo improvements, as its lifespan and efficiency are still insufficient for factory work.
Elon Musk stated that some design issues may only surface after mass production begins. Therefore, during the current pre-production phase, Optimus’ design is continuously being adjusted and will continue to improve even after large-scale production starts.
This is because Optimus’s performance has yet to meet Tesla's design goals and remains quite far from the objective of replacing humans.
Optimus weighs less than most full-sized humanoid robots and has a slimmer and more delicate build. This is due to the higher integration level of the robot's actuators, allowing engineers to pack more components into a smaller volume. A supplier estimated that the key actuator size of the third-generation Optimus is approximately 10% smaller than the previous generation. However, it has yet to meet Musk's target.
Optimus’s body underwent several adjustments — the four harmonic reducers in the waist actuator were replaced with new types of reducers to enhance Optimus's load-bearing capacity; lighter reducers and motors were redesigned to reduce weight.
Musk expects the third-generation Optimus to shed 10 kilograms from the second-generation model (weighing 56.6 kilograms) to extend battery life. Currently, the engineering team has achieved only 70% of this goal, reducing the weight by 7.3 kilograms.
These improvements might make Optimus slightly superior to robots from Figure AI and various Chinese companies, but they remain insufficient to support Musk's narrative of transforming the global economic structure through robots. During yesterday’s earnings call, Musk reiterated that deploying Optimus at scale means 'infinite money' as Optimus could be five times more productive annually than a human, operating 24/7 without needing recharging, simply plugged into a power source.
However, according to multiple suppliers, apart from the hands, the service life and stability of multiple joint actuators in Optimus’s arms and legs also fall short of expectations.
From walking to picking up objects, various movements of Optimus will wear out the precision components required for joint activity, and unlike the human body, they cannot self-repair. The service life of precision devices such as reducers is generally 8,000 hours, which means less than three years under an 8-hour work schedule; once broken, they can only be replaced. The service life of planetary roller screws is even lower.
Such potential engineering issues still require a significant amount of time for debugging and correction. 'The Tesla team believes that as long as the hardware design approach is correct, improvements in accuracy and reliability are foreseeable. However, engineers sometimes underestimate the time and resources required,' the aforementioned person said.
An executive from a robotics company stated that the average working speed of the second-generation Optimus in automotive factories is only 20% to 30% of that of humans, with only a few simple actions reaching 60% of human speed. The efficiency of the third-generation Optimus in moving objects has doubled but remains significantly slower than workers.
Development hurdles do not affect Tesla's upward adjustment of production capacity plans.
What we are in contact with$Tesla (TSLA.US)$The suppliers we spoke with only learned from Musk early yesterday morning that Tesla plans to ramp up Optimus production to one million units annually by the end of next year.
These suppliers all believe that the plan will inevitably change. In just the past few weeks, Tesla’s procurement department has been assessing supplier production readiness and has already requested some suppliers to increase their component production capacity to 1,000 units per week by October next year and to 2,000 units per week by December next year — approximately 100,000 units annually.
Musk once described Optimus as a technical challenge more difficult than the Model X but simpler than sending humans to Mars via Starship. The difficulty with Optimus lies in the fact that Tesla has no reference models and must redesign almost every component, achieving true vertical integration. When Tesla started making electric vehicles 20 years ago, it at least didn’t need to redesign the steering wheel or tires. A humanoid robot consists of about 7,000 parts, 30% fewer than a car, but most of them have to be redesigned.
But at least this time, Tesla is receiving much stronger supply chain support than it did 20 years ago.
Even though suppliers have many complaints about the technical challenges and the never-realized mass production schedules, no one would refuse Tesla. Supplying to Tesla means not only orders but also technical support. According to an informed source, a supplier selected by Tesla had worked on frameless torque motors for two years but still could not meet Tesla's requirements; Tesla solved the problem themselves and licensed the blueprints and technical solutions to the supplier for contract manufacturing.
The boost to suppliers' stock prices from entering Tesla's supply chain is also a driving factor. A secondary market investor said, “Making plastic robot masks has no technical content. But if you can do business with Tesla, it’s a different story.” Currently, at least 10 companies are working with Tesla to develop dexterous hands in hopes of becoming suppliers.
Based on the most conservative mass production plan, Tesla will still produce approximately 50,000 humanoid robots next year, several times more than companies like Figure. Tesla has yet to prove that Optimus is capable of commercial use or generating economic value. However, we understand that several leading global automotive parts companies are in discussions with Tesla regarding intention orders totaling over ten thousand units, allowing Optimus to begin work in factories.
Editor/melody
