Tesla Says Optimus Is on the Fremont Line, With $4.73 Billion of Capex and No Unit Count

The number that matters in Tesla's third-quarter update is not the claim that Optimus is in production but the one it leaves blank. Tesla's installed-capacity table lists California as "Early Ramp" and Texas as "Construction", and neither row carries a figure, according to TeslaNorth's reading of the update. A line with no published rate cannot be compared with the one million units a year Elon Musk named as the Fremont goal on the Q4 2025 call.

What Tesla actually said about Optimus on October 21

Tesla wrote that "Optimus production began in Fremont on our high-volume lines" and that "The bots coming off this line are the first production-quality bots that we have manufactured," as quoted by TeslaNorth. It added that it has "proven that the current manufacturing process for Optimus is suitable for a commercial product."

That is a claim about the process, not about output. The update gives no production rate, no unit count, no price, no date for sales outside Tesla and no statement of which Optimus version is being built. Digital Today's account of the shareholder letter likewise gives no volumes or pricing.

Item	Disclosed in Q3 2026 update?
Optimus production started in Fremont	Yes
Units built or per-month rate	No
Installed capacity, California	"Early Ramp", no figure
Installed capacity, Texas	"Construction", no figure
Price or outside-sales date	No

Where the money is going: $4.73 billion of capex in one quarter

The cost side is easier to read than the output side. TeslaNorth's recap of the quarter puts revenue at $28.05 billion, flat on the year, with operating expenses up 39% to $4.78 billion. Operating income was $718 million, a 2.6% margin, against $1.62 billion a year earlier.

GAAP net income was $787 million, down 43% from $1.37 billion in the same quarter of 2025, per Digital Today. Digital Today attributes the drop to heavier spending on Optimus, A.I. and autonomy, and to lower regulatory credit sales. TeslaNorth's recap does not give a cause.

Capex was $4.73 billion against operating cash flow of $6.26 billion, leaving free cash flow of $1.53 billion. In the second quarter free cash flow was negative $1.09 billion on $5.79 billion of capex, so the swing came mostly from cash flow, not from spending less.

1. Operating cash flow6.26 $ billion
2. Capex4.73 $ billion
3. Free cash flow1.53 $ billion

Source: TeslaNorth recap of Tesla's Q3 2026 update, accessed 2026-10-23

Capex is not split by program in the material TeslaNorth summarised. On the Q1 call, CFO Vaibhav Taneja said "our current expectation for 2026 is over $25 billion of CapEx", a figure that LetsDataScience reported covers the Fremont Optimus line, the Giga Texas Optimus factory, the Cybercab ramp and a $3 billion semiconductor research fab. Optimus is one line item among four.

Why the ramp is slow by design

Musk told the Q1 call the ramp "will be quite slow for us as we iron out the 10,000-plus unique items" and that it "will move as fast as the least lucky, slowest, dumbest part in the entire 10,000," per LetsDataScience. That is a supply-chain statement. A humanoid with more than 10,000 distinct parts is gated by its worst supplier, and a first line on high-volume floor space tells you the tooling exists, not that parts arrive.

The floor space was cleared on purpose. Musk said in January that Model S and Model X would end to make room for Optimus, and TeslaNorth reports the S and X lines were decommissioned by Q2. Early builds go to what Tesla calls the "Optimus Academy", which collects training data.

The track record is the denominator here. LetsDataScience notes Musk's January 2025 forecast of roughly 10,000 robots that year went unmet, and that in January 2026 he acknowledged no Optimus robots were doing useful work at Tesla. A production-quality label on a line is a step beyond that, but it is one step.

What the competition has put on paper

On the same Q1 call coverage, LetsDataScience lists Boston Dynamics shipping electric Atlas units to Hyundai with a planned 30,000-unit-a-year facility, and Agility's Digit deployed at a Toyota warehouse. The Terminal has covered the Atlas GR3 hand and its 100,000-a-year target, and Mind Robotics, which has raised $1.015 billion with no public robot. Each of them sets a target; none is a measured rate either, and a target is the cheapest thing in a robotics plan to write down.

What would change my read

Two disclosures would turn this from a process claim into a production claim: a units-per-month figure for the Fremont line, and a capacity number in the Texas row. Tesla has promised neither, TeslaNorth reports the Texas factory's steel frame was nearing completion in September, and LetsDataScience says the second factory is targeted for around summer 2027. Until the table carries a number, the 4.73 billion dollars is the only quantity in the update that can be checked.