1X Prioritizes Fleet Learning Over Manufacturing
1X Technologies
This split is really a bet on where value will sit in humanoids, in the robot body or in the learning loop. 1X and similar teams are using each deployment as a data collection machine, where teleoperators show the robot how to move, cameras and sensors record edge cases, and those traces feed back into general control models. That is a very different race from building large factories early, because the goal is to own the behavior stack before hardware gets cheaper and more standardized.
-
At 1X, that loop is already built into the product. EVE can be supervised through a browser dashboard, and new behaviors can be taught with VR headsets and haptic gloves. NEO is also being introduced with heavy teleoperation support, which turns commercial use into labeled training data collection.
-
The contrast with manufacturing first players is concrete. Agility built RoboFab with peak capacity of 10,000 robots per year, while 1X is still centered on model improvement and fleet learning. In practice, one side optimizes factory throughput, the other optimizes how fast each robot gets smarter from real work.
-
This strategy also reflects pricing pressure. As lower cost humanoid hardware spreads, including Unitree G1 at a starting price around $16,000, basic robot bodies become easier to source. That makes proprietary data, control software, and adaptation to messy real environments more important than pure manufacturing scale.
The likely end state is a humanoid market where hardware looks more interchangeable and advantage comes from the company with the best real world learning engine. If 1X keeps expanding deployments across security, logistics, and home settings, each new robot can improve the whole fleet and push the company toward software like economics on top of hardware.