Tesla's Optimus robot prototype walked on stage

Tesla held its AI Day event last week. The star of the show was supposed to be the Optimus humanoid robot. The Tesla Bot.

It walked out on stage. Slowly. Carefully. Arms slightly raised for balance. No tether, which is something. But the movements were stiff. Tentative. It waved at the audience. The audience cheered, but there was a quality to the cheering that felt like encouragement more than amazement.

It’s fair to say the demo was underwhelming compared to what Boston Dynamics has been showing for years. Atlas does parkour. Optimus shuffles.

And yet.

The Tesla pattern

Tesla has a pattern. Show a terrible first version. Get mocked. Iterate. Ship something that works. Get mocked less. Iterate more. Ship something impressive.

The original Tesla Roadster was a Lotus Elise with batteries. The Model S was a luxury sedan that couldn’t be serviced anywhere. The Model 3 had build quality issues that would have sunk a traditional automaker. The Gigafactories were supposed to be impossible.

None of those things are mockable anymore.

I’m not saying Optimus will follow the same trajectory. Building a car and building a humanoid robot are different problems with different physics. Cars don’t need to balance. Cars don’t need to grasp objects or navigate stairs or interact with humans at arm’s length.

But the pattern of “terrible first version from a company with enormous resources and a willingness to iterate in public” is worth watching.

The mass production claim

Here’s the thing that caught my attention more than the demo itself: Elon said Tesla intends to produce Optimus in the millions at a target price under $20,000.

Millions. Under $20,000.

Boston Dynamics makes a few hundred units of Spot, its quadruped robot, at roughly $75,000 each. Atlas is a research platform, not a product. No company has ever mass-produced a humanoid robot. The infrastructure to do it doesn’t exist.

But Tesla makes 1.3 million cars a year. They have factories. They have supply chains. They have battery manufacturing at scale. They have experience building complex electromechanical systems in volume.

If anyone could figure out mass production of a humanoid robot, it would be a company that already mass-produces complex machines. I don’t know if they will. But the gap between “a car company that makes millions of robots on wheels” and “a car company that makes millions of robots on legs” is smaller than the gap between “a robotics lab that makes a few hundred units” and “a factory that makes millions.”

My questions

Can Optimus do useful work? The demo showed it carrying a box and watering a plant. Those are party tricks, not tasks. A useful humanoid robot needs to handle unstructured environments: pick up objects of varying sizes, navigate cluttered spaces, respond to unexpected situations. Atlas can do some of this in controlled settings. Can Optimus?

What’s the timeline? Tesla timelines are famously optimistic. Full Self-Driving has been “next year” for several years. If Optimus follows a similar trajectory, “millions of units” might mean “a few thousand eventually.”

Who’s it for? A $20,000 humanoid robot would be cheaper than a year of minimum wage labor. If it can do even basic physical work (warehouse picking, simple assembly, janitorial tasks), the economics are immediate. But “if” is doing a lot of heavy lifting in that sentence.

I have questions. I don’t have answers. But I’ve learned not to write off the company that made electric cars mainstream while everyone said it couldn’t be done.

The robot shuffled across a stage. It was unimpressive. It was also the beginning of something. I’m watching.

Related thinking: