Technology

Tesla Starts Optimus Production at Fremont as Profit Falls 43%

Martin HollowayPublished 7m ago4 min readBased on 19 sources
Reading level
Tesla Starts Optimus Production at Fremont as Profit Falls 43%
Photo by U.S. Air Force / Trevor Cokley / Public domain

Tesla said on Wednesday, October 21, 2026, that it had started production of its Optimus humanoid robot at its Fremont, California factory, a plant that previously built electric vehicles. The announcement came with third-quarter results in which revenue held roughly flat and net income fell sharply.

According to the company's third-quarter 2026 shareholder letter, the Optimus units coming off the Fremont line are the first production-quality bots Tesla has manufactured (Tesla Q3 2026 Update). The Q2 2026 update, in its robotics section, described the decommissioning of the Model S and X manufacturing lines at Fremont and the installation of first-generation equipment there. Tesla's July 22 earnings report had said it would "start production soon" of Optimus and that the initial robots would be used for training data collection, meaning the data a robot's AI learns from (CNBC).

Total third-quarter revenue was $28.04 billion. That is roughly flat against $28.09 billion a year earlier and about $200 million below the prior quarter's $28.23 billion. Net income was $787 million, down 43% from $1.37 billion in the third quarter of 2025. Tesla attributed the pressure on profit to lower sales of regulatory credits, which automakers earn for meeting emissions rules and can sell to other manufacturers, and to rising spending on AI projects and the Optimus robot, while overall expenses increased (TechCrunch).

The segment picture was mixed. EV sales revenue was flat year over year at about $20 billion. Services grew 33% to $4.6 billion, and the services and other segment posted a record $734 million in gross profit, the margin left after the direct cost of delivering the service. Solar and energy storage revenue fell 16% to $2.8 billion, which Tesla attributed to tariff-related customer refunds.

Vehicle volumes were steady at the margin. Tesla delivered 486,532 vehicles in the quarter, about 6,000 more than in the second quarter but fewer than in the third quarter of 2025. The October 2 production and deliveries release had put output at over 464,000 vehicles and deliveries at over 486,000 (Tesla IR). On the energy side, Tesla deployed 13.7 GWh of storage products, up from 13.5 GWh in the second quarter.

Capital intensity runs through the whole quarter. Capital expenditure, or capex, is money spent on factories, equipment and infrastructure rather than on day-to-day operations. In April, CFO Vaibhav Taneja raised Tesla's 2026 capex forecast to $25 billion, from $9 billion in 2025 (New York Post). Tesla's Investor Relations site lists the third-quarter earnings event for October 21, 2026, with a press release, Q&A and webcast.

The broader context here is the mix of the numbers. Spending commitments have risen far faster than the automotive revenue line, and the quarter's growth came from services rather than from vehicles. For now, the Fremont announcement reads best as a cost-side and capacity-side event. Tesla describes the early robots as slated for training data collection, which points to internal use before any external revenue.

In my view, the metrics to watch over the next few quarters are unit yield, meaning the share of robots that pass quality checks on the first attempt, the repeatability of the build, and the cost per robot at the Fremont line. Those figures will determine whether Optimus becomes a revenue line or remains an expense the company has to justify. Tesla has not yet attached a specific output rate to the Fremont start, which makes the early ramp the more informative signal.

A more hopeful reading is also available. Retooling a high-volume car plant to build a different class of machine is a concrete example of manufacturing capacity being repurposed. Think of a bakery converting its ovens and benches to make furniture: the floor space stays the same, but the equipment and the skills change. That kind of step is what determines how quickly a humanoid robot can move from demonstration to production. Whether the Fremont line achieves it will take more than one quarter to show.