Tesla Is Building Hundreds of Optimus Robots a Week, but Hands Are the Bottleneck
Tesla built several hundred Optimus robots per week last month, working toward a stated goal of 20,000 per week, with hand assembly and tight part fitting now setting the pace. The Verge
That output follows a decision earlier this year to reuse Tesla's Model S and Model X lines for Optimus. The change gave Tesla floor space and line equipment for humanoids, but the production problems since then have been specific to humanoids, not leftover car-making issues.
The hands are the constraint. The V3 hands and forearms hold more than 100 screws and other small parts that a worker must still put together by hand. Tesla must also align parts with a precision its car operations did not need, because Optimus parts are much smaller and must fit together more closely than car parts.
V3 hardware is lighter and has more cameras than earlier versions. It still lacks a sensing glove now in development, which would improve durability by letting Tesla replace worn touch sensors without replacing the whole hand. That missing part affects service. Replacing a hand means swapping a complex device at the end of the arm, what engineers call an end-effector. Replacing a glove would be a smaller, lower-level repair.
Use remains narrow. Tesla uses Optimus internally for specific jobs in limited areas, not for general work and not under remote human control, what the industry calls teleoperation. The reported plan is to lease Optimus units at first rather than sell them, and to use the training data they collect to improve the robots. For people building robots, that points to a closed learning loop, with data staying inside Tesla's fleet before any wider release.
From small batches to line production
In the second quarter, Tesla made a few dozen Optimus robots per week during small-batch testing, according to background reporting from The Information. The move to several hundred per week last month suggests line work at its Fremont factory has moved past bench builds, the stage where Tesla says it develops manufacturing lines and processes for Optimus.
That shift has not been steady. Tesla paused buying Optimus parts while it changed the robot's hardware and software, a pause reported in July 2025. The Information Separate coverage in September 2026 cited a report that Optimus hands break and suppliers cannot keep up with production. Electrek
Job listings point to the same bottleneck. Tesla lists a Manufacturing Engineer, Optimus Hands Launch role to lead the launch and ramp of production for robot hands, and a Robotics Manufacturing Engineer role to launch production for full robots including Full Bot Validation Testing.
Milan Kovac, the head of the Optimus program, said in a post on X in June 2025 that he planned to step down. Reuters
Timelines and product definition
Elon Musk has said Optimus will be Tesla's biggest product of all time. Public dates for outside availability have shifted. In April 2024, Musk said Tesla could start selling Optimus by the end of next year. Reuters In January 2026, Musk said Tesla will probably sell Optimus to the public by the end of next year. Bloomberg The current reported plan to lease first rather than sell would keep early units under Tesla control while training data builds up.
Tesla describes Optimus as a "general-purpose, bi-pedal, autonomous humanoid robot capable of performing unsafe, repetitive or boring tasks." Tesla Current internal use for specific tasks in limited areas fits that description as a goal, not as a present capability. The recent reporting describes no general-purpose deployment and no teleoperated operation.
The broader context here is a change in manufacturing type, not only higher volume. Car assembly allows millimeter-level variation across large metal parts, with fasteners and seals that absorb small differences. Dexterous hands reverse that pattern. Small errors across dozens of screws, tiny bearings, cable-like tendons and thin sensor wiring add up quickly, a problem engineers call tolerance stack-up, and hand assembly hides variation that automated stations will later need to control or design out.
In my view, the lease-first detail is the clearest operational signal. It keeps failure analysis, touch-sensor wear data and grasp failures inside Tesla's engineering loop, and it avoids building a service network for a machine whose most failure-prone part, the hand, still needs manual assembly and full-hand replacement. The gap is large. Several hundred per week to 20,000 per week is nearly a hundredfold increase, and added labor alone will not close it.
Worth flagging for humanoid teams watching this, Tesla's problems will look familiar to anyone who has shipped precise small-scale hardware. Design for easy assembly, fewer screws, modular sensor packaging and tight supplier control for tiny mechanical parts often set schedules more than AI models or fleet size. If Tesla delivers the sensing glove, automates hand assembly and holds supplier yield, leased fleets doing limited jobs could produce the hands-on data that general-purpose claims will eventually need. That would make Optimus factory equipment first and a public product second, which over the long term is still how useful robots usually reach people.
