Tesla Must Build 1,000 Robots a Week, and the Hands Keep Failing

The number that decides Tesla's year is a weekly count at Fremont: Optimus humanoids built per week. Managers want more than 1,000 a week by 31 December, on a line that until May built the Model S ¹²³. Output reached several hundred a week in August, up from a few dozen in the spring ²³.

The car business kept 1.4 cents of every dollar

Second-quarter revenue set a record at $28.2 billion, up 26%, yet operating income fell 57% to $398 million: a 1.4% margin against 4.1% a year earlier ⁴⁵. Capex guidance went above $25 billion in April↗, roughly triple the $8.5 billion spent in 2025, and second-quarter free cash flow ran at minus $1.09 billion ⁴⁶. CFO Vaibhav Taneja already books "preproduction ramp costs" for the Semi, Optimus, Cybercab and other AI initiatives inside operating expenses, and expects negative free cash flow through year-end ⁶⁷.

A 155-times multiple rests on the robot

Tesla trades at roughly 155 times forward earnings against about 15 for Nvidia on one tally, near 195 on another ¹⁴. Morningstar's Seth Goldstein states the condition plainly: the capex case depends on Optimus becoming Tesla's most value-creating platform, and skeptical investors may find the spending difficult to support ⁴.

The hands are the hard part

Each hand and forearm holds more than 100 screws and small parts, still fitted by workers; touch sensors fail often enough that a replaceable "sensing glove" is planned for next year; and suppliers, many in China, struggle to hold quality at volume ²³. Most robots built so far stay inside Tesla for testing and data collection, and the software needs days to learn a basic task ². Tesla's next platform is being screwed together by hand. No public figure establishes the weekly rate beyond "several hundred," and no robot revenue is broken out ⁴.

Musk's own forecast is the flat part

"Optimus will follow the sort of normal S-curve of a manufacturing ramp, but the initial portion of the S-curve will be quite flat and long," Musk said on the Q2 call↗ ¹. He calls the spending "well justified for a substantially increased future revenue stream," with useful work outside Tesla "probably" a 2027 event ⁶. Output did rise about tenfold in a quarter ³.

Rivals raise billions on the same thesis

FieldAI has signed a term sheet for $700 million at a $10 billion valuation↗, five times its worth a year ago, on more than $135 million of revenue and customer contracts, with Jeff Bezos's family office, Nvidia's NVentures and Intel Capital among its backers ⁸⁹. Skild AI stands above $14 billion and Physical Intelligence near $11 billion ⁸. The private market pays for the robot brain without a car business financing it.

The verdict watch: 21 October

Third-quarter earnings arrive around 21 October ¹. The tell is disclosure: a build rate compounding toward 1,000 a week, or the same multiple resting on a program with no separable revenue.