Tesla is upgrading Optimus to several hundred units per week, but hands and generalization are blocking

Production increased tenfold between the second quarter and August. The internal target is more than a thousand robots per week by the end of the year, while the V3s remain confined to internal use.

Tesla increased the assembly of its humanoid Optimus from a few dozen units per week, in small test batches in the second quarter, to several hundred per week in August — a roughly tenfold increase. The teams are reportedly aiming for a continuous automated line capable of surpassing a thousand robots per week by the end of the year, with a longer-term target of about 20,000 units per week.

Almost all of the machines produced remain in-house, for testing, training and data collection. The V3s that come off the Fremont line are not the version intended for wide commercialization: Tesla still has to cross stricter thresholds of durability and reliability before renting the robots to outside customers.

The hard points

Three obstacles return: hands, whose mechanical complexity remains the most cited bottleneck; the automated manufacturing equipment itself; and the availability of suppliers. In addition, there is a software limitation that is less visible in production figures — the difficulty of robots in generalizing from one learned task to a related one, which conditions the real interest of a fleet of several thousand units.