According to a report by *The Information*, production of a few hundred robots per week began in Fremont back in August.

That is roughly ten times the level seen in the second quarter.

So, where are these robots? Who is buying them?

They aren't being sold externally yet. Tesla is using them in its own factories to test their durability and gather data for their learning process.

To me, this is the most interesting part of the story.

Tesla is ramping up production without waiting for the robots to become capable of performing every task autonomously.

It appears the strategy is to first master production and lower costs, then refine the AI ​​using data collected from robots operating in the field.

The report also notes that the former Model S and Model X production area has been allocated to Optimus. The goal for the automated assembly line—scheduled to go online in late 2026—is to produce over 1,000 robots per week.

The long-term target is around 20,000 per week.

These are targets, of course; let’s not confuse them with current production figures.

The plan is to eventually transfer the manufacturing methods developed in Fremont to the large-scale lines at Giga Texas.

Naturally, there are challenges involved.

Take the hands, for instance.

Assembling each hand and forearm involves over 100 tiny screws and precision parts that are still put together by hand.

Production issues at supplier facilities also mean some parts require manual adjustment after assembly.

It takes a significant amount of human labor to manufacture the very robot hand designed to reduce human labor. :)

The learning aspect isn't easy, either.

According to the report, Tesla has reassigned a substantial portion of the team that previously labeled data for Autopilot to work on the robots instead. They are trying to increase their training data to over a million hours using motion-tracking suits and remote-control systems.

In other words, scaling up the robot's hardware is progressing faster than getting it to reliably perform various tasks.

There are two things I’ll be watching here:

How much is the cost of manufacturing a robot dropping, and for how many hours can it perform useful work without human assistance?

Yes, seeing thousands of robots in a factory is impressive.

But for an investor, the real issue is how many of those robots are actually getting work done and creating value.