Tesla starts Optimus production in Fremont as Q3 net income falls 43%
Tesla said in its Q3 2026 update, filed with the SEC on Oct 21, that Optimus production has begun on its high-volume lines in Fremont, California, and that the robots coming off the line are the first production-quality units it has built; the company listed the Fremont Optimus line as "Early Ramp" and Texas as "Construction" but disclosed no unit count, rate or customer.
GAAP net income attributable to shareholders fell 43% year on year to $787 million on flat revenue of $28.05 billion, while R&D rose 69% to $2.76 billion and capex reached $4.73 billion.
Tesla has started building Optimus humanoid robots on a production line, and it still will not say how many. In its third-quarter 2026 update, filed with the SEC as an exhibit to an 8-K on Oct 21, the company said Optimus production began in Fremont on its high-volume lines and that the bots coming off this line are the first production-quality bots it has manufactured. It added that it will keep iterating on hardware and process but has proven that the current manufacturing process is suitable for a commercial product.
The same document shows what the robot programme costs the rest of the company. GAAP net income attributable to common stockholders fell to $787 million from $1,373 million a year earlier, a 43% drop, on revenue that was essentially flat at $28,046 million. The question for investors is whether the line in Fremont is the start of a product business or another stage of a capital programme with no disclosed output. ([MEKAWIRE explainer](/news/tesla-optimus-explained-2026-what-tesla-has-filed-1-million-robot-line-10-million-texas-plan-no-units-disclosed-20261017))
What the filing says, and what it leaves out
Tesla's capacity table now lists Optimus in California as "Early Ramp" and Optimus in Texas as "Construction". In the Q2 update the Fremont entry described lines still being installed, with initial builds going to an internal Optimus Academy for data collection. The outlook section repeats the point in one line: production has begun for Tesla Semi in Nevada and Optimus in Fremont.
The update gives no number of robots built, no weekly rate, no external customer, no price and no date for sales. Capacity figures for Optimus, which earlier updates gave as about 1 million a year for the first-generation line in Fremont and 10 million for a second-generation line in Texas, are not restated in the table; the Optimus rows carry a status but no capacity. The results call on Oct 21 is not used here because Tesla does not publish a transcript.
The money behind the robot
Operating income fell to $718 million from $1,624 million, and the operating margin dropped to 2.6%. Tesla attributed the decline mainly to higher operating expenses for super intelligence and other R&D projects, stock-based compensation including the 2025 CEO award, and SG&A, along with lower regulatory credit and energy profits. R&D spending reached $2,761 million, up 69% from $1,630 million, or 9.8% of revenue. Capital expenditure was $4,732 million in the quarter, after $5,789 million in Q2.
- Q3 2025
- 28,095
- 1,624
- 1,630
- 2,248
- 1,373
- Not listed
- Q1 2026
- 22,387
- 941
- 1,946
- 2,493
- 477
- Line planned to replace Model S/X
- Q2 2026
- 28,236
- 398
- 2,371
- 5,789
- 1,114
- Lines being installed
- Q3 2026
- 28,046
- 718
- 2,761
- 4,732
- 787
- Early Ramp, production begun
Tesla quarterly updates (Ex. 99.1). Net income change = 787 / 1,373 = 0.573 (-42.7%); R&D change = 2,761 / 1,630 = 1.694 (+69.4%); operating margin = 718 / 28,046 = 2.6%.
As of Oct 21, 2026
Tesla said it plans to reach 400 MW of training compute by year end to develop autonomy software for both vehicles and humanoid robots; Cortex 2 in Texas is now listed above 120 MW. Free cash flow turned positive again at $1,531 million and cash and short-term investments rose to $45.4 billion, so the spending is funded from the balance sheet, not from new capital.
How the claim compares with rivals' disclosed output
Tesla is late to disclosed volume. In MEKAWIRE's verified records, Unitree reported more than 5,500 humanoids shipped in 2025 in its STAR Market filings, UBTech reported 921 humanoid units sold in the first half of 2026 in its HKEX interim results, Figure said on Apr 29 that it had delivered more than 350 robots and was building one per hour, and Agility Robotics has disclosed more than $300 million of Digit orders. All of those are counts or contract values. Tesla's is a status label.
- Unitree
- China
- >5,500 humanoids shipped in 2025
- STAR Market filing
- UBTech
- China
- 921 humanoid units sold, H1 2026
- HKEX interim results
- Figure AI
- US
- >350 robots delivered, 1 per hour (Apr 29, 2026)
- Company post
- Agility Robotics
- US
- >$300M of Digit orders
- Company release
- Tesla
- US
- Production begun, Early Ramp; no units
- SEC filing
Disclosed figures only, not estimates; metrics differ by company and are not directly comparable.
As of Oct 22, 2026
MEKAWIRE analysis
Tesla has moved Optimus from construction to production in its own filing, but until it discloses units, Optimus remains a capital story that is now visibly weighing on group profit.
The filing changes one thing: Tesla now asserts in an SEC exhibit that its process is suitable for a commercial product, a stronger statement than "in anticipation of production" in Q2. It does not change the evidence base. A line in early ramp can produce a handful of robots a week or hundreds; the document does not let anyone tell which. Our calculation from the filed figures shows the R&D increase alone, $1,131 million year on year, is larger than the entire quarter's operating income.
The strongest counter-argument is that Tesla gives no production counts for other early products in this update either; Cybercab and Semi appear with capacity and status only, so silence on Optimus units is consistent with how it reports a ramp. Against that, Chinese rivals with smaller balance sheets already report unit sales, and buyers comparing humanoids will ask for deployments, not status labels.
Bull case: Tesla reports internal deployment numbers in its Q4 update and names a first external customer in 2027, and the Fremont line proves the cost curve that its car business is known for. Bear case: the Q4 update still lists "Early Ramp" with no count, R&D keeps growing faster than revenue, and the market compares a 2.6% operating margin with rivals' growing shipments.
Three signals to watch: Tesla's Q3 10-Q, due in the coming weeks, for Optimus-specific capex or inventory; the Q4 2026 update in January for a first unit count; and the Texas Optimus status moving from Construction to Early Ramp.
- $2,761M, up 69% from $1,630M
- $4,732M in Q3 (Q2: $5,789M)
- Tesla Q3 2026 update, Exhibit 99.1 to 8-K filed Oct 21, 2026
- $28,046M (Q3 2025: $28,095M)
- Fremont, California: Early Ramp; Texas: Construction
- None (no count, rate, customer or price)
- $718M, down 56% from $1,624M; margin 2.6%
- Target 400 MW by end of 2026; Cortex 2 >120 MW
- First production-quality bots built; process suitable for a commercial product
- $787M, down 43% from $1,373M