Tesla Targets 20,000 Weekly Output for Optimus Robots

Tesla has begun assembling Optimus robots on a high-volume line inside its Fremont factory. The company states these units are the first to meet the standard for a commercial product. A third quarter shareholder letter confirms the manufacturing process is now suitable for a real product.

The production line occupies space that previously built the Model S and Model X. Tesla removed the old vehicle equipment in 46 days this spring. The facility is sized for one million units per year. Most of the necessary equipment for that capacity is already installed in Fremont.

Elon Musk addressed the ramp-up during Wednesday's earnings call. He stated that weekly output will climb toward roughly 20,000 robots. Musk described the growth trajectory as a classic S curve. He noted that the initial phase involves a shallow slope before volume increases sharply.

The engineering team is currently refining the robot's artificial intelligence and electromechanical design. They are also working on reliability and service life. Musk expects this work to be complete by the end of next year. Tesla plans to start delivering units to outside companies at that time. Those robots will handle dangerous, repetitive, and around the clock jobs.

Tesla faces a unique challenge in training these robots. It lacks the vast fleet of vehicles that provides driving data for its cars. Musk acknowledged this difference on the call. The company developed a solution called the Optimus Academy. Between 10,000 and 20,000 robots will practice a wide range of tasks in this environment.

Every unit produced on the Fremont line will feed data into the Academy. This process allows the robots to teach themselves and improve the next version of the model. Tesla also plans to install its AI5 chip into the robots next year. Musk said a new video will arrive within one or two months. That video will show the robot completing tasks driven by artificial intelligence rather than preprogrammed motions.

Musk criticized rival humanoid demonstrations that rely on scripted routines. He argued those demos would not hold up in the real world. The current focus remains on making the hardware and software robust enough for commercial deployment.