Tesla can build hundreds of Optimus robots a week—but not yet reliable hands

Tesla produced several hundred Optimus robots a week in August, up from a few dozen during second-quarter pilot production, according to The Information. Managers want a continuously running line capable of more than 1,000 a week by year-end; the eventual ambition is approximately 20,000. The robots now leaving Fremont are mainly used internally for testing and data collection, and are not yet the commercial version Tesla intends to offer customers.

The hand is the clearest physical bottleneck. Its forearm and hand contain more than 100 screws and other small parts that workers still assemble manually. Touch sensors have had reliability problems, prompting Tesla to develop a replaceable sensing glove rather than replace the whole hand. Fixtures that align joints and other components sometimes miss the precision required, while new automated equipment becomes unreliable when the line speeds up, producing robots that need rework.

Software has not escaped the same gap between demonstration and deployment. Tesla has accumulated more than 500,000 hours of training data and aims to double that by year-end, using remote-operation suits, first-person cameras and collection hubs across several US states. Yet the robots still rely heavily on training for specific tasks, may take days to learn even a basic operation and behave unpredictably outside familiar situations. Their current factory work therefore occurs in tightly controlled areas.

Tesla reportedly plans to lease the first commercial machines so it can retrieve, upgrade and refurbish them while collecting customer-factory data. That is a sensible way to manage an immature product, but it also changes what the production numbers mean. Several hundred bodies a week are not several hundred general-purpose workers: dexterous hardware, dependable suppliers and transferable behaviour all have to work simultaneously before Optimus becomes the business Musk describes. Tesla did not comment to The Information; Musk has publicly acknowledged that scaling the robot is an unusually difficult manufacturing problem.