The Robotics Industry Part 3, Humanoids TO THE MOON
Humanoids are having their Crypto moment.
Money is flowing into humanoid robot startups. Companies are raising hundreds of millions at crazy valuations.
Imagine, a robot that can replace all human labor.
I am hearing accounts from established robotics companies whose investors are asking them — “but can you make it a humanoid?”
As a former robotics researcher, I’m here to tell you that making your hardware as complicated as possible is not going to help you automate things more effectively.
Please, draw your own conclusion from the accounts collected bellow.
The Story of Boston dynamics:
With the guiding principle that you must build something useful to make money, I have long watched the Boston dynamics dancing robot videos, and wondered, how the heck do they make money. Let’s unravel the myth.
Boston Dynamics was a spinout of a lab at MIT in 1992 working on legged robots funded by military agencies like DARPA. In 2004, Boston Dynamics created Big Dog, a quadruped robot capable of carrying heavy loads over rough terrain.
There is no public information about how many Big Dogs were sold, but the sole application was in the military for carrying supplies.
In 2013, Google acquired Boston Dynamics for $500 million, but the company struggled to fit into Google’s plans.
Boston Dynamics’ Atlas humanoid robot was created around 2015 and was never intended to be sold to the public. It was initially developed for the DARPA robotics challenge to enhance the role of robots in disaster relief and was never sold, always described as a research platform.
In 2017, SoftBank acquired Boston Dynamics at a 60% discount for $165 million, and the company began focusing more on commercializing its technology.
In 2020 Boston dynamics released Spot, its first commercial robot charging $75,000 for the quadruped and sold a couple hundred over the next few years. This puts a rough estimate of revenue $10 million.
That same year Hyundai acquired Boston dynamics for nearly $1 billion.
In 2024 Boston Dynamics unveiled plans to sell a newer, fully electric generation of the Atlas humanoid robot designed for real-world applications like moving objects in factories. The cost per robot is estimated to be around $150k.
The story of Boston dynamics is a quite interesting one.
They are a state-of-the-art, money-losing robotics research company.
Their evolution is a metaphor for the challenges faced by robots companies generally. Robots are an epic technology, but they’ve struggled to find product market fit in industry, and much like Boston Dynamics, robotics found their home (or their use) to be inside a car company.
The Age of the Humanoid:
For the past few decades Boston Dynamics was THE humanoid company, now there is a proliferation of humanoids and VC investment flowing into the space, and I think everyone should take a breath.
It may seem surprising, given how frequently we see humanoid robots in YouTube videos, but not one company has actually sold a humanoid for any use other than entertainment or research. And not one humanoid is being used as a true resource to drive efficiency.
Obviously there are many who believe that humanoid robots are the solution to all our automation problems and maybe someday we’ll live in a world like this…
Well a few weeks ago I went to a Robotics in Industry summit and spent a lot of time talking to the leaders in commercialization, this was my experience:
I went to an event that was mostly populated with robotics researchers, and a majority of the discussion revolved around humanoids & AI foundation models for robotics.
Later that day I went to the industry expo for companies in robotics, and there was not a humanoid to be seen. I asked around at the 100's of booths, and each company laughed when I asked them about humanoids telling me that they had enough complexity with 6 degrees-of-freedom.
What this confirms in my mind is the propensity for academic & media interests to gravitate to what’s possible, and industry to gravitate to what’s useful.
It’s easy to conflate what you see often, with the true prevalence of an idea.
In pondering the value of all this media attention and investment with the head of engineering of Verve Motion a robotic exoskeleton startup, he said something that I quite liked:
Going to the moon didn’t solve any problems or generate any revenue, but it did drive innovation, and create many valuable technologies that went on to find broad market penetration.
So in theory, the development of humanoid robots would have the same effect on the robotic technology, whether or not there is an immediate use case for them.
The same day I had a discussion with various industry veterans from Guardian Ag and Realtime Robotics, on choosing the right hardware platform. I found this comment funny and insightful:
Let’s say you are choosing whether to put wheels or legs on your robotic system. Wheels are about 1/100 of the bill of materials cost, so you would nearly always choose wheels if you’re thinking practically.
What if there are stairs? Well, why are there stairs? and can you not change the surroundings that this robot will operate in. Navigating a ramp is going to be much more reliable than climbing a set of stairs.
Reliability is the single most important thing in robotics. Simplify the problem. Don’t over engineer the solution.
In summary, humanoid robots are an amazing intellectual pursuit for the good of humanity and I’m so glad that the best minds are focusing on it.
Given the rate of change of technology today, I wouldn’t bet against anything. I remain reserved on the proximity of monetization and I prefer to see robotics companies solving a task, not trying to automate a human.
I would love to be proven wrong.