Paxton says Figure and Optimus cannot lift 50 lb
One researcher reads the published payload specs and says the heavy humanoid field is tiny, and almost entirely American.
Chris Paxton posted that there really are very few heavy-industry humanoids, and that neither Figure nor Optimus is strong enough for much of the work humans do every day, which he puts at around a 50 lb payload.
So if you want a heavy humanoid today, you literally only have American options
His argument is about where the line sits. Fifty pounds is the standard OSHA limit for manual material handling, so it is the number that decides whether a robot can do the job at all. And the step from 40 lb to 50 lb is bigger than it sounds. A higher payload robot is heavier, control gets harder, and falls do more damage.
The China part
Paxton used the same post to push back on the idea that China is ahead in robotics. On his list of heavy humanoids, he says there are essentially no Chinese companies and no European ones. The one exception he names is Agibot's new A2 Max, and he discounts it because as far as he can tell the robot does not exist yet and is still listed as coming soon on the company website.
"So if you want a heavy humanoid today, you literally only have American options," he wrote.
Shown versus claimed
None of this is a test. It is one researcher reading published specs, and Paxton says the chart he posted came from ChatGPT and tells people not to take it too seriously.
The thread started from a line Humanoids daily flagged. Asked by analysts to name his competitors, Agility co-founder Jonathan Hurst said almost none of the humanoid robots that exist compete with Agility. They have two legs and two arms maybe, Hurst said, but they are not picking up bins and they do not have any safety. The only rival he named was Boston Dynamics.
Almost none of the humanoid robots that exist compete with us. They have two legs and two arms maybe, but they aren't picking up bins... They don't have any safety.
Steve Crowe pushed back on the framing, asking why people always compare Atlas, Digit and Figure to the Unitrees of the world, calling it apples and oranges. Paxton partly agreed. He said Unitree's H series is a fair comparison for Figure but the G1 is not, and added that Figure comes out looking better than H1 and H2 anyway.
The useful takeaway is the threshold rather than the leaderboard. Most humanoid demos you have seen move light objects. The payload number that matters for warehouse work is one almost nobody has published a robot that clears.
Why do people always compare Atlas, Digit, Figure to the Unitrees of the world? Apples & oranges.