Optimus V3 arrives this summer, and Fremont production starts within weeks

Updated 2026-06-12.

Tesla's humanoid robot is about to become a production program. The Optimus V3 unveiling is expected in the middle of this year, with the Fremont production line starting up in late July or August.

What happened

Elon Musk laid out the timeline at an investor presentation, per Electrek's April coverage. The Fremont Optimus line takes over the space where the last Model S and X rolled off in early May 2026, a real changing-of-the-guard moment: the floor that built Tesla's first mass-market flagships now builds its first robot. Initial output will start slow by design. Optimus has 10,000 unique parts on an entirely new line, and Musk was candid that the first-year production rate is hard to predict.

The long-term targets are enormous. Tesla is aiming for 1 million robots per year of capacity, with first production before the end of 2026 and consumer sales targeted for end of 2027, according to a timeline summary published by Tesery. A second Optimus factory is already under construction at Giga Texas with production expected around summer 2027.

The robot rides on Tesla's broader AI build-out. The Cortex 2.0 supercomputer comes online in phases starting April 2026, reaching full 500MW capacity by mid-2026, deliberately timed to support the summer production launch. The same compute that trains FSD trains Optimus, which is the whole thesis: one AI stack, many bodies.

Our take

Humanoid robots aren't FrunkLab's market. We build a Tesla wrap design tool. But the underlying shift powering Optimus is the same one powering us, and it's worth spelling out.

The generative AI layer that lets a robot interpret a kitchen and pour a drink is the same layer that lets you describe a wrap in plain English and get a usable starting design in five seconds. You type "matte navy with copper pinstripes," and the model returns a top-down design that maps to your specific Tesla template. That's possible because AI has finally gotten cheap and fast enough to be ambient, not a feature you pay $50 a month to access.

The honest part: shipping AI into the physical world is hard, and the hard part isn't the model. It's the integration into real workflows with tolerances, failure modes, and users who don't read documentation. Tesla is taking the time to get a 10,000-part machine right, and that's the correct trade. The same dynamic shows up in AI car customization, which we covered in AI in car customization 2026: the demos got good fast, and the products that win are the ones that sweat the boring integration details.

A robot that builds itself a job and a car that designs its own look are downstream of the same bet, and 2026 is the year both get real. If you want to try the wrap half of that future today, the studio is open. It costs nothing to start.

The robots can wait. Your car can't. Start in the studio.