Tesla is finally building Optimus robots. Musk spent the call lowering expectations

Tesla has started building humanoid robots in a California car factory. That sentence has been "coming soon" for so long that it's easy to miss when it finally becomes true, and it became true this week, buried in the same shareholder letter that reported a 43% drop in profit.

In its third-quarter letter, published October 21, Tesla said Optimus units are now rolling off a line in Fremont, in space that used to build electric vehicles. The company called them "the first production-quality bots that we have manufactured," according to TechCrunch, and wrote that "we have proven that the current manufacturing process for Optimus is suitable for a commercial product."

Then Elon Musk got on the earnings call and spent a fair bit of it lowering expectations.

The robot, by Musk's own account

Musk described this as "early production" and warned the ramp would be slow and uneven, because nobody has built this product at scale before. His target, per TechCrunch, is 20,000 robots a week by the end of 2027. He also said the robot still has to be made "safe and reliable and has long duration," which is a polite way of listing the three things a factory robot most needs to be.

According to Shacknews' notes from the call, Musk said the hard part is mostly the hand and fingers, which have to be soft enough to handle objects and tough enough to survive wear and tear. He said the robot has around 10,000 components and that production can only grow as fast as the slowest one. He expects the AI, the electromechanical design and the reliability problems to be solved by the end of next year, and said outside companies should expect deliveries around the end of 2027, starting with industrial customers. At one point, per Shacknews, he told listeners he wanted to temper expectations around Optimus.

He also promised a video in a month or so of Optimus doing things that aren't preprogrammed. Keep that one on file.

Why the skepticism is earned

Optimus has a long history of demos that looked more autonomous than they were. In January, on the fourth-quarter 2025 call, Electrek reported that Musk admitted no Optimus robots were doing useful work in Tesla's factories, saying: "It's not in usage in our factories in a material way." Back then, Tesla's own update said production was planned to start before the end of 2026. It has, just about, with ten weeks to spare.

Tesla gave no price and no delivery date for customers. What it did give was a factory line, and that matters. Building a few hundred hand-tuned prototypes is a science project. Building them on a line, with suppliers, yields and a cost per unit, is a business, or at least the start of one. Tesla knows how to do that second thing better than almost any company in robotics. The question has always been whether the robot itself would be good enough to deserve the line.

The bill for the bet

Meanwhile, the car company that pays for all of this had a flat quarter. Revenue was $28.04 billion, a hair under the $28.09 billion of a year earlier, per TechCrunch. Net income fell 43% to $787 million from $1.37 billion. Vehicle revenue held at roughly $20 billion. Deliveries were 486,532, up about 6,000 on last quarter but down on the same quarter of 2025. Solar and storage revenue fell 16% to $2.8 billion, which Tesla blamed on tariff-related refunds to customers.

The bright spot was services, up 33% to $4.6 billion with a record $734 million in gross profit. The drag, according to TechCrunch, was mostly spending on AI and Optimus, plus fewer regulatory credit sales.

Put plainly, Tesla is taking profit from a business that works and pouring it into one that might. Shareholders signed up for that. They also signed up for timelines, and those keep moving. On the same call, Tesla's VP of AI software, Ashok Elluswamy, gave a fresh estimate for unsupervised Full Self-Driving: "We suspect somewhere around Q2 next year is when we can turn on some of the first customer vehicles." Back in April, Musk had said that would "probably" come in the fourth quarter of 2026. That is now.

Who actually wins the humanoid race

Tesla is not alone on this. Agility, Figure and a long list of Chinese firms including Unitree and AgiBot are chasing the same factory customers. Tesla's pitch has always been scale: if anyone can bring a humanoid's cost down to a car-like number, it's the company that already makes millions of cars.

That pitch got a little more real this week. The first robots off the Fremont line exist. By Musk's own estimate, deliveries to outside companies are more than a year out. Twenty thousand a week is a number for 2027 slides.

For now, the most honest line from the whole call might be the one Musk said, according to Shacknews, about the strategy: build imperfect robots, then train them. It's a reasonable plan. It's also, word for word, what every robotics startup without a car business has been saying for a decade.

Sources: TechCrunch: Tesla starts Optimus robot production in California as profit falls 43%, Shacknews: Tesla (TSLA) Q3 2026 earnings results and conference call, Electrek: Musk admits no Optimus robots are doing 'useful work' at Tesla.