Tesla Starts Initial Optimus Production at Fremont Factory; Original Model S/X Production Lines Have Been Removed

Sure enough, Tesla has recently confirmed that the Optimus humanoid robot production line at its Fremont, California factory has begun initial production, and the first batch of robots meeting mass-production manufacturing standards has officially rolled off the line.

You should know that this was a classic Tesla vehicle production line that had been running for more than a decade, and just a few months ago it was still building the Tesla Model S/X. To make room for the robots, Tesla simply tore out the original car production line.

And that's not all.

Although Optimus's hands are not yet fully mature, Musk still plans to first build 10,000 to 20,000 mass-production-grade robots.

But once these robots roll off the line, they won't be going to work just yet. Musk plans to send them to a dedicated school first, like raising children.

Tesla's latest third-quarter financial report shows that Optimus robots have begun initial production on a high-volume production line at the Fremont, California factory, and the first batch of robots meeting mass-production manufacturing quality standards has rolled off the line.

Tesla said that although the hardware design and manufacturing process are still being iterated, the existing production process has preliminarily proven that it can meet the manufacturing requirements for commercial products.

In the official capacity table, the status of the Fremont Optimus project has also changed from the previous construction phase to Early Ramp (initial capacity ramp-up).

And at the subsequent earnings call, Musk revealed more details:

At present, the Fremont factory's production line planned for an annual output of 1 million Optimus units has most of its equipment already in place. However, to truly reach a production rate of about 20,000 units per week, it is expected to take at least another year, or even longer.

Musk said that Optimus's capacity ramp-up will follow a typical S-curve, with progress relatively slow and even somewhat bumpy in the early stages. The reason is that this is a brand-new, complex hardware product.

According to him, robot manufacturing requires thousands of steps to operate normally at the same time; even if 9,999 steps are ready, a problem in just one remaining step could slow down the entire production rhythm.

What's more, some of Optimus's designs still need adjustment, and the supply chain and manufacturing processes are also being improved in parallel.

But Musk is clearly impatient.

He expects that by around the end of 2027, Tesla is expected to resolve Optimus's main issues in AI capabilities, electromechanical design, reliability and service life, and begin delivering robots to external companies.

The first batch of robots is expected to mainly enter the industrial sector, taking on dangerous, highly repetitive tasks or tasks that require long periods of execution. As product capabilities and output further improve, they will gradually enter more scenarios such as homes.

Beyond Fremont, Tesla has also prepared a larger-scale production base for Optimus.

According to previously announced plans, Tesla is building a second-generation Optimus production line at its Texas Gigafactory in the United States, with a long-term target annual capacity of 10 million units.

There is one more thing worth noting here—Fremont is one of Tesla's most historically significant factories.

In 2012, the Model S began production here, and in 2015 the Model X joined it. The former, with its range, performance and intelligent experience, broke many people's preconceptions about pure electric vehicles; the latter further strengthened Tesla's tech brand image with unique designs such as falcon-wing doors.

Before the Model 3/Y became the sales mainstay, these two high-end models could be said to be important contributors that supported Tesla's development.

However, in January this year, Musk suddenly announced that Model S/X production would end in the second quarter. After the last batch of Model S/X rolled off the line in May, in just 46 days, this automotive production line with more than a decade of history was dismantled.

Tesla even added a caption: End of an era, the end of an era.

The reason for reorganizing the Fremont production line was to free up capacity for the robots. In the eyes of outsiders, Tesla has two considerations:

On the one hand, it is eliminating old products with limited sales to improve the efficiency of factory resource utilization.

In 2025, other models including the Model S/X and Cybertruck delivered a total of about 50,900 vehicles for the full year. According to Electrek's estimate, the combined annual sales of the Model S/X have already fallen to about 30,000 vehicles, while the original production line had a planned annual capacity of 100,000 vehicles.

Of course, judging from the million-unit-level capacity plan, Tesla clearly wants far more than just filling an idle car production line.

So on the other hand, Tesla also wants to apply the production, supply chain and engineering experience accumulated over years in the automotive business to what Musk sees as a new business with greater growth potential.

However, before Optimus robots officially enter large-scale mass production, there is actually one issue that cannot be ignored—

Some of Optimus's technologies have not yet been solved, especially the robot's hands, which have still not been fully finalized. Why is Tesla in such a hurry to move it into the production stage?

At the latest earnings call, analysts put this question to Musk.

Musk immediately gave a clear answer:

In other words, Tesla wants to build robots at scale first, and then let them provide their own training data.

This is actually closely related to the FSD development model that Tesla is familiar with.

In the field of autonomous driving, Tesla has a huge installed base of real vehicles. Millions of cars are constantly running on real roads, providing a large number of driving scenarios and training data for the FSD model.

But at present, when it comes to robots, Tesla neither has a data network made up of millions of humanoid robots, nor can it easily obtain from existing car data the experience of manipulating objects, using tools, and performing complex labor tasks—after all, the environments and work content that humanoid robots face are far more complex than those of cars.

Therefore, Optimus needs to establish a new method of accumulating data.

In fact, Tesla had already invested a large amount of manpower in collecting training data. According to a September report by The Information, Tesla had already accumulated more than 500,000 hours of robot training data at that time, and planned to double that scale by the end of this year.

To obtain this data, Tesla recruited dedicated collection personnel who, using tools such as head-mounted cameras, motion capture equipment, and tactile gloves, recorded human operation processes such as folding clothes, wiping tabletops, and sorting batteries.

No matter how many soccer matches a person watches, it does not mean they can play soccer well. To truly master this skill, one still has to practice on the field in person.

The same goes for robots. After understanding tasks through human demonstrations, robots still need to continuously accumulate experience of success and failure through their own actual operations.

So Musk arranged a special school for Optimus—Optimus Academy.

According to the latest plan, Tesla intends to first manufacture approximately 10,000 to 20,000 robots, let them try various tasks in training facilities, and continuously improve their capabilities through autonomous practice in simulated environments and the real world.

Musk even compared this process to "children growing up." Humans, from toddling to mastering various life skills, rely to a large extent on constant interaction, trial, and practice with the real world. And he hopes Optimus can also follow a similar learning path.

In this way, manufacturing a large number of robots in advance is not only for validating the product and production process, but also responsible for accumulating training experience for the next generation of robots.

On closer thought, this approach is actually very much in Musk's style.

Previously, Tesla insisted on pure vision autonomous driving, and one important reason Musk gave was that humans mainly rely on their eyes and brain to drive, and machines should also be able to use cameras and neural networks to accomplish the same work; the current robots are likewise seeking inspiration from the way humans learn skills.

However, to have tens of thousands of robots learning at the same time, Tesla first has to solve the problem of how to manufacture them stably.

According to a previous report by The Information citing people familiar with the matter, in August of this year, Optimus's weekly output had reached several hundred units, about a 10-fold increase compared with the level of a few dozen units per week in the second quarter. However, there is still a large gap from Musk's envisioned 20,000 units per week.

One of the most difficult links to solve is precisely the robot's hands.

The report mentioned that Optimus's hands and forearms contain more than 100 screws and other small parts, some components still require manual assembly by workers, and the tactile sensors also have reliability issues.

Musk further explained at the latest conference call that human fingers are soft and flexible, have a large number of tactile receptors, and also possess a certain self-repair capability.

But robots must both perform fine grasping and work repeatedly for long periods, and must not easily wear out or damage internal sensors—achieving this is far more complicated than imagined. Even the small mechanical structures inside the fingers will affect the robot's lifespan and energy consumption.

In his view, a set of pre-programmed movements, no matter how spectacular the performance, cannot truly prove that a robot possesses the intelligent capability to adapt to the real environment.

In addition, Tesla also plans to integrate the next-generation AI5 chip into Optimus around April 2027. According to Musk, its computing power is expected to reach about 10 times that of AI4, providing support for the robot to handle more complex tasks.

He even compared the three to multipliers that stack on one another, believing that as these technologies continue to advance, the robot's capabilities are also expected to accelerate.

And this is still far from the endpoint he envisions.

At the conference call, Musk once again made a bold claim: Optimus could become the largest product on Earth.

He even began to imagine that in the future Optimus would manufacture the next generation of Optimus itself, and then the new generation of robots would continue production, forming an ever-expanding manufacturing cycle.

According to his prediction, the number of humanoid robots on Earth may eventually even exceed the number of humans. Production capacity plans of millions or tens of millions of units seem to be only a starting point.

Although Musk's previous product plans have often slipped, this time it is rare for him to be so swift and positive in recent years, which shows he is truly serious about robots...