The $400 humanoid, Digit the forklift, Germany's robot gyms, and 6 chip companies focusing on humanoid robots
Today in Humanoid Daily ... a new $400 humanoid powered by a smartphone, Digit 5 is the forklift of humanoids, Germany's robot billionaire is building robot gyms, China unveils a 13-foot humanoid robot that lifts half a ton, and the number of the day is 6 ... 6 chip companies focusing on robots.
UCLA built a $400 humanoid powered by a smartphone
What if you could build a humanoid robot for just $400? Researchers at UCLA have developed PhoneBot, a miniature humanoid that uses an ordinary Android smartphone as its brain, eyes, and sensors.
It's about 48 centimeters tall, has 13 motors, and can walk, follow people, and interact with humans.
And instead of requiring expensive dedicated processors, cameras, and computing hardware, PhoneBot uses technology that's already inside millions of smartphones.
The $400 price doesn't include the phone, and this robot doesn't have arms or hands. It's not going to replace a warehouse worker.
But that's not the point.
The researchers have made the hardware and software open source, giving universities, schools, and researchers an inexpensive platform for experimenting with humanoid robotics.
And I think that's significant, because getting more robots into more hands means more people experimenting, developing software, and solving problems. We've seen how inexpensive computers democratized software development.
Could inexpensive humanoids do something similar for physical AI?
Digit 5 is the forklift of humanoid robots
Here's something you don't hear very often from a robotics founder.
Jonathan Hurst, co-founder of Agility Robotics, has written a remarkably personal essay for IEEE Spectrum about Digit 5, the company's newest humanoid. And he admits he's not entirely happy with what he's built.
Hurst's dream was to create robots that moved with the elegance and agility of animals and humans, and his early robots, ATRIAS and Cassie, were designed around that vision. They could balance, run, and recover from disturbances with remarkably natural movements.
But Digit 5?
Hurst says it feels more like a forklift than the graceful humanoid he originally envisioned. Turns out, customers don't care how elegantly a robot walks. They care whether it can move heavy bins, work all day, and operate safely around people.
So Digit 5 weighs more than 120 kilograms. It's designed to lift 23-kilogram bins onto shelves 1.75 meters high. It has a massive battery, sophisticated cooling, and enough sensors and safety systems to work near human employees. In fact, the battery and cooling systems occupy about 75% of its torso.
All that added weight means Digit 5 can't run or jump like its predecessors ...
There's another interesting compromise. If a human approaches too closely, Digit is designed to put down its load, sit on the floor, and shut off its motors, which is not particularly elegant. But ... it is safe.
Hurst wanted to build a beautiful machine that mastered humanlike movement. Instead, he ended up building something that might actually make money. That's probalby an important moment in the humanoid industry: the transition from impressive demos to ugly machines engineered around real customer requirements.
Hurst says Digit 5 may not be the robot he dreamed of, but it could be the foundation that makes his original dream possible.
Germany's robot billionaire is building humanoid gyms
Germany has a new robotics billionaire, and his plan to win the humanoid race involves something that sounds like a fitness club for robots. Or something straight out of China.
David Reger is the founder of Neura Robotics, which is now reportedly valued at about $7 billion.
Neura is building dedicated facilities where robots can learn to perform real-world jobs. Businesses can bring robots into these training environments and teach them specific tasks. The company already has a facility at Munich Airport, developed with partners at a reported cost of $19 million. Another is being developed at RWTH Aachen University.
Neura is targeting five operating gyms by the end of this year, with ten more in development.
The skills learned in these gyms will eventually be uploaded to Neura's cloud platform, Neuraverse, and transferred to other robots. Teach 1 robot a complicated manufacturing task, and you can downloadi that skill into hundreds of others ... or offer it on a marketplace where businesses buy and sell robot skills.
Building humanoid robots is hard, sure.
But teaching them enough useful skills to make them economically valuable might be even harder.
China's new humanoid is 13 feet tall and lifts half a ton
Not all humanoid robots are human-sized. Some are much scarier. China's NeodraDynamics has unveiled Neo1, an absolutely massive humanoid robot standing four meters tall — about 13 feet — and weighing 1.2 tons.
This thing is built for heavy lifting.
(Maybe it's the real forklift of humanoid robots, not Digit 5!)
The company says its two arms can carry a combined 500 kilograms, or roughly 1,100 pounds, and Neo1 has 41 active joints, 39 of which use hydraulic actuators. That's a very different approach from the electric motors powering most of today's humanoids ... but heavy industrial equipment does still use hydraulics: it's strong and reliable.
And yeah ... the specifications get even crazier.
NeodraDynamics says its hydraulic technology can tolerate temperatures from minus 50 to plus 250 degrees Celsius and pressures equivalent to depths of 10,000 meters underwater.
(!!!)
Neo1 isn't commercially deployed yet, so let's not get too excited until others can test it. But the potential applications are fascinating: mining, nuclear facilities, deep-sea operations, disaster response ... any places where sending humans can be extremely dangerous.
Most humanoid companies are trying to build robots that can do what people do. NeodraDynamics is pursuing things people physically cannot.
Number of the day: 6
6.
That's how many major semiconductor and electronics companies have recently made moves targeting humanoid robots, and it's a sign that the industry is starting to build a specialized supply chain.
- Renesas has established a dedicated physical AI team.
- Micron is developing new memory technology for humanoids.
- Intel is targeting 700 TOPS of on-robot computing power by 2028.
- TDK InvenSense is developing magnetic sensors for robot joints.
- Namuga is supplying camera modules to humanoid manufacturers.
- Infineon is expanding into semiconductor systems for robotic hands.
A humanoid is essentially a walking collection of processors, sensors, motors, memory, and power electronics, and most of those components are still expensive. But as companies start designing chips and other components specifically for humanoids, they can become cheaper, smaller, more powerful, and more efficient.
We've already seen this happen with smartphones and electric vehicles: specialized components become standardized, standardized components get cheaper, and cheaper components enable scale.
Some similar is likely to happen here ... depending on the level of scaling we see in humanoids.
Sources
- HumanoidMegaHub — The $400 humanoid | UCLA — PhoneBot research paper
- IEEE Spectrum — The Robot I Didn't Expect to Build, by Jonathan Hurst
- Forbes — Germany's new robot billionaire is betting big on humanoid gyms
- NeodraDynamics — Official Neo1 announcement
- The Elec — Renesas launches physical AI team | Micron's humanoid memory initiative
Written by
I tell stories with data and build startups. I've raised angel and VC funds, led software companies and startups, and chronicled the rise of the mobile economy. Currently I write for Forbes, host several podcasts, and am starting up HumanoidMegaHub. I was TUNE's mobile economist and Singular's VP of Insights, built the research team at VentureBeat, and managed teams creating software for partners like Intel and Disney. In past lives I built websites, e-learning platforms, and mobile apps, and, a long time ago, was named to Folio's top 100 of the media industry's "most innovative entrepreneurs and market shaker-uppers.”