BMW’s strategy for humanoid robots is overlooked

BMW is deploying humanoid robots on its factory floor in Leipzig this summer. There are two Aeon models from Hexagon Robotics. BMW is not doing it because humanoids are the best tool for the job. Nope, they are doing it because the factory already exists.

The manufacturing lines already exist with tooling, workstations, and processes that are all built for human-shaped workers. In other words, humanoid shapes fit, and specialized optimized robot shapes don’t.

Rethinking an entire production line around a new form factor costs a fortune. But dropping in a robot that can navigate the same space a human worker uses costs a fraction of that. Robot costs are falling fast, which means the ROI tipping point is coming soon.

You don’t need a humanoid robot to be perfect. It just needs to be good enough, cheap enough, and compatible with what you already built. And it will be ok if it makes mistakes along the way.

One, a robot may not injure a human being or, through inaction, allow a human being to come to harm….

Isaac Asimov, I, Robot

That’s BMW’s bet. And I think it’s the right one. Companies that win the next decade of manufacturing won’t be the ones who designed the most optimized robotic factory from scratch. They’ll be the ones who figure out how to deploy capable robots in their existing factories.

We’re building robots that are great at demos and terrible at reality

NASA’s former robotics chief warned this week that the US is ‘building the wrong kind of robots.’ Designers are over-optimizing for demos that give investors a wow-moment, but are unprepared for real-world deployment.

Real life is not a demo. And there’s a perverse incentive at play.

Demos raise money, get media press coverage, and make founders look like geniuses. So founders optimize for what gets them to the next funding round. But not the thing that actually works in the wild.

This isn’t a new dynamic, and we’ve seen it in AI broadly. Systems that look incredible under curated conditions but fall apart in the real world when users touch them. The robotics companies that will matter in ten years are the ones chasing deployment data right now. The companies suffering through mountains of problems as they tackle messy edge cases. Like the flooded roads that made Waymo pause service this week. You can’t demo your way to reliability. You have to earn it the hard way.

Tesla’s Optimus Factory is the biggest bet in robotics right now

Tesla is reportedly building a dedicated Optimus factory at Gigafactory Texas with a target production of 10 million units per year. But keep in mind that’s an Elon Musk target.

But ten million humanoid robots per year is significantly more than the entire global industrial robot output of 500,000 units. The important thing to consider is that Tesla does not have to hit ten million to disrupt the manufacturing industry. They only need to deliver a number that drives unit economics to a point where humanoid robots become a commodity. When a capable humanoid robot costs less than a year of human labor, the business case for robots changes quickly.

Of course, that change does not apply to every job. But it does apply to enough jobs to entirely restructure how we think about the workforce. Hugging Face just dropped a 3D-printable humanoid for $2,500. Tesla is laser-focused on mass production, which is something they know about now thanks to Tesla cars. The race to the cost floor is on.